--- old/src/java.base/share/classes/java/net/IDN.java 2020-01-10 15:57:02.000000000 -0800 +++ new/src/java.base/share/classes/java/net/IDN.java 2020-01-10 15:57:02.000000000 -0800 @@ -1,5 +1,5 @@ /* - * Copyright (c) 2005, 2013, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2005, 2020, Oracle and/or its affiliates. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it @@ -29,9 +29,9 @@ import java.security.AccessController; import java.security.PrivilegedAction; -import sun.net.idn.StringPrep; -import sun.net.idn.Punycode; -import sun.text.normalizer.UCharacterIterator; +import jdk.internal.icu.impl.Punycode; +import jdk.internal.icu.text.StringPrep; +import jdk.internal.icu.text.UCharacterIterator; /** * Provides methods to convert internationalized domain names (IDNs) between @@ -226,7 +226,7 @@ InputStream stream = null; try { - final String IDN_PROFILE = "uidna.spp"; + final String IDN_PROFILE = "/sun/net/idn/uidna.spp"; if (System.getSecurityManager() != null) { stream = AccessController.doPrivileged(new PrivilegedAction<>() { public InputStream run() { --- old/src/java.base/share/classes/java/text/Bidi.java 2020-01-10 15:57:03.000000000 -0800 +++ new/src/java.base/share/classes/java/text/Bidi.java 2020-01-10 15:57:03.000000000 -0800 @@ -1,5 +1,5 @@ /* - * Copyright (c) 2000, 2019, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2000, 2020, Oracle and/or its affiliates. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it @@ -35,7 +35,7 @@ package java.text; -import sun.text.bidi.BidiBase; +import jdk.internal.icu.text.BidiBase; /** * This class implements the Unicode Bidirectional Algorithm. --- old/src/java.base/share/classes/java/text/CollationElementIterator.java 2020-01-10 15:57:04.000000000 -0800 +++ new/src/java.base/share/classes/java/text/CollationElementIterator.java 2020-01-10 15:57:04.000000000 -0800 @@ -1,5 +1,5 @@ /* - * Copyright (c) 1996, 2019, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 1996, 2020, Oracle and/or its affiliates. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it @@ -41,7 +41,7 @@ import java.lang.Character; import java.util.Vector; import sun.text.CollatorUtilities; -import sun.text.normalizer.NormalizerBase; +import jdk.internal.icu.text.NormalizerBase; /** * The {@code CollationElementIterator} class is used as an iterator --- old/src/java.base/share/classes/java/text/Normalizer.java 2020-01-10 15:57:05.000000000 -0800 +++ new/src/java.base/share/classes/java/text/Normalizer.java 2020-01-10 15:57:05.000000000 -0800 @@ -1,5 +1,5 @@ /* - * Copyright (c) 2005, 2019, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2005, 2020, Oracle and/or its affiliates. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it @@ -37,7 +37,7 @@ package java.text; -import sun.text.normalizer.NormalizerBase; +import jdk.internal.icu.text.NormalizerBase; /** * This class provides the method {@code normalize} which transforms Unicode --- old/src/java.base/share/classes/java/text/RBTableBuilder.java 2020-01-10 15:57:06.000000000 -0800 +++ new/src/java.base/share/classes/java/text/RBTableBuilder.java 2020-01-10 15:57:06.000000000 -0800 @@ -1,5 +1,5 @@ /* - * Copyright (c) 1999, 2019, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 1999, 2020, Oracle and/or its affiliates. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it @@ -42,8 +42,7 @@ import sun.text.UCompactIntArray; import sun.text.IntHashtable; import sun.text.ComposedCharIter; -import sun.text.CollatorUtilities; -import sun.text.normalizer.NormalizerImpl; +import jdk.internal.icu.impl.NormalizerImpl; /** * This class contains all the code to parse a RuleBasedCollator pattern --- old/src/java.base/share/classes/sun/text/CollatorUtilities.java 2020-01-10 15:57:08.000000000 -0800 +++ new/src/java.base/share/classes/sun/text/CollatorUtilities.java 2020-01-10 15:57:07.000000000 -0800 @@ -1,5 +1,5 @@ /* - * Copyright (c) 2000, 2005, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2000, 2020, Oracle and/or its affiliates. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it @@ -25,7 +25,7 @@ package sun.text; -import sun.text.normalizer.NormalizerBase; +import jdk.internal.icu.text.NormalizerBase; public class CollatorUtilities { --- old/src/java.base/share/classes/sun/text/ComposedCharIter.java 2020-01-10 15:57:09.000000000 -0800 +++ new/src/java.base/share/classes/sun/text/ComposedCharIter.java 2020-01-10 15:57:09.000000000 -0800 @@ -1,5 +1,5 @@ /* - * Copyright (c) 2001, 2015, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2001, 2020, Oracle and/or its affiliates. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it @@ -25,8 +25,8 @@ package sun.text; -import sun.text.normalizer.NormalizerBase; -import sun.text.normalizer.NormalizerImpl; +import jdk.internal.icu.impl.NormalizerImpl; +import jdk.internal.icu.text.NormalizerBase; public final class ComposedCharIter { /** --- old/src/java.base/share/classes/sun/text/Normalizer.java 2020-01-10 15:57:10.000000000 -0800 +++ new/src/java.base/share/classes/sun/text/Normalizer.java 2020-01-10 15:57:10.000000000 -0800 @@ -1,5 +1,5 @@ /* - * Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2005, 2020, Oracle and/or its affiliates. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it @@ -25,8 +25,8 @@ package sun.text; -import sun.text.normalizer.NormalizerBase; -import sun.text.normalizer.UCharacter; +import jdk.internal.icu.lang.UCharacter; +import jdk.internal.icu.text.NormalizerBase; /** * This Normalizer is for Unicode 3.2 support for IDNA only. --- old/test/jdk/java/text/Bidi/Bug6850113.java 2020-01-10 15:57:11.000000000 -0800 +++ new/test/jdk/java/text/Bidi/Bug6850113.java 2020-01-10 15:57:11.000000000 -0800 @@ -1,5 +1,5 @@ /* - * Copyright (c) 2009, 2012, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2009, 2020, Oracle and/or its affiliates. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it @@ -22,14 +22,14 @@ */ /* * @test - * @bug 6850113 + * @bug 6850113 8174270 * @summary Verify the return value of digit() for some digits. - * @modules java.base/sun.text.normalizer + * @modules java.base/jdk.internal.icu.lang * @compile -XDignore.symbol.file=true Bug6850113.java * @run main Bug6850113 */ -import sun.text.normalizer.UCharacter; +import jdk.internal.icu.lang.UCharacter; public class Bug6850113 { --- old/test/jdk/java/text/Bidi/Bug7051769.java 2020-01-10 15:57:12.000000000 -0800 +++ new/test/jdk/java/text/Bidi/Bug7051769.java 2020-01-10 15:57:12.000000000 -0800 @@ -1,5 +1,5 @@ /* - * Copyright (c) 2011, 2016, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2011, 2020, Oracle and/or its affiliates. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it @@ -23,7 +23,7 @@ /* * @test - * @bug 7051769 8038092 + * @bug 7051769 8038092 8174270 * @summary verify that Bidi.toString() returns the corect result. * The second run is intended to test lazy SharedSectets init for 8038092 * @modules java.desktop @@ -40,7 +40,7 @@ if (System.getProperty("preloadBidi", "").equals("true")) { // Make sure the SharedSecret is lazily initialized correctly try { - Class.forName("sun.text.bidi.BidiBase"); + Class.forName("jdk.internal.icu.text.BidiBase"); System.out.println("BidiBase class has been pre-loaded."); } catch (ClassNotFoundException e) { System.out.println("BidiBase class could not be pre-loaded."); @@ -68,7 +68,7 @@ TextAttribute.RUN_DIRECTION_RTL); String text = "\u0623\u0643\u062a\u0648\u0628\u0631 10"; - String expected = "sun.text.bidi.BidiBase[dir: 2 baselevel: 1 length: 9 runs: [1 1 1 1 1 1 1 2 2] text: [0x623 0x643 0x62a 0x648 0x628 0x631 0x20 0x661 0x660]]"; + String expected = "jdk.internal.icu.text.BidiBase[dir: 2 baselevel: 1 length: 9 runs: [1 1 1 1 1 1 1 2 2] text: [0x623 0x643 0x62a 0x648 0x628 0x631 0x20 0x661 0x660]]"; AttributedString as = new AttributedString(text, attrNS); AttributedCharacterIterator itr = as.getIterator(); --- old/test/jdk/java/text/Normalizer/ConformanceTest.java 2020-01-10 15:57:14.000000000 -0800 +++ new/test/jdk/java/text/Normalizer/ConformanceTest.java 2020-01-10 15:57:13.000000000 -0800 @@ -1,5 +1,5 @@ /* - * Copyright (c) 2019, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2019, 2020, Oracle and/or its affiliates. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it @@ -22,10 +22,10 @@ */ /* * @test - * @bug 4221795 6565620 6959267 7070436 7198195 8032446 8221431 + * @bug 4221795 6565620 6959267 7070436 7198195 8032446 8174270 8221431 * @summary Confirm Normalizer's fundamental behavior * @library /lib/testlibrary/java/lang - * @modules java.base/sun.text java.base/sun.text.normalizer + * @modules java.base/sun.text java.base/jdk.internal.icu.text * @compile -XDignore.symbol.file ConformanceTest.java * @run main/timeout=3000 ConformanceTest */ @@ -34,14 +34,12 @@ import java.io.File; import java.io.FileInputStream; import java.io.InputStreamReader; -import java.lang.reflect.Method; import java.nio.charset.Charset; import java.nio.charset.CharsetDecoder; import java.util.BitSet; import java.util.StringTokenizer; -import sun.text.normalizer.NormalizerBase; -import sun.text.normalizer.NormalizerImpl; +import jdk.internal.icu.text.NormalizerBase; /* * Conformance test for java.text.Normalizer and sun.text.Normalizer. --- old/test/jdk/java/text/Normalizer/ICUBasicTest.java 2020-01-10 15:57:15.000000000 -0800 +++ new/test/jdk/java/text/Normalizer/ICUBasicTest.java 2020-01-10 15:57:15.000000000 -0800 @@ -1,5 +1,5 @@ /* - * Copyright (c) 2019, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2019, 2020, Oracle and/or its affiliates. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it @@ -22,10 +22,10 @@ */ /* * @test - * @bug 4221795 8032446 + * @bug 4221795 8032446 8174270 * @summary Confirm Normalizer's fundamental behavior. Imported from ICU4J 3.2's * src/com/ibm/icu/dev/test and modified. - * @modules java.base/sun.text java.base/sun.text.normalizer + * @modules java.base/sun.text java.base/jdk.internal.icu.text * @library /java/text/testlib * @compile -XDignore.symbol.file ICUBasicTest.java * @run main/timeout=30 ICUBasicTest @@ -39,11 +39,9 @@ */ import sun.text.Normalizer; -import sun.text.normalizer.NormalizerBase; -import sun.text.normalizer.NormalizerImpl; +import jdk.internal.icu.text.NormalizerBase; import static java.text.Normalizer.Form.*; -import static sun.text.normalizer.NormalizerBase.Mode.*; public class ICUBasicTest extends IntlTest { --- old/test/jdk/java/text/Normalizer/NormalizerAPITest.java 2020-01-10 15:57:16.000000000 -0800 +++ new/test/jdk/java/text/Normalizer/NormalizerAPITest.java 2020-01-10 15:57:16.000000000 -0800 @@ -1,5 +1,5 @@ /* - * Copyright (c) 2018, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2018, 2020, Oracle and/or its affiliates. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it @@ -23,9 +23,9 @@ /* * @test - * @bug 4221795 + * @bug 4221795 8174270 * @summary Confirm Normalizer's fundamental behavior - * @modules java.base/sun.text java.base/sun.text.normalizer + * @modules java.base/sun.text java.base/jdk.internal.icu.text * @library /java/text/testlib * @compile -XDignore.symbol.file NormalizerAPITest.java * @run main/timeout=30 NormalizerAPITest @@ -64,8 +64,8 @@ static final int[] options = { 0x00, sun.text.Normalizer.UNICODE_3_2, - sun.text.normalizer.NormalizerBase.UNICODE_3_2, - sun.text.normalizer.NormalizerBase.UNICODE_LATEST, + jdk.internal.icu.text.NormalizerBase.UNICODE_3_2, + jdk.internal.icu.text.NormalizerBase.UNICODE_LATEST, }; static final String nonNullStr = "testdata"; @@ -319,7 +319,7 @@ in.getClass().getSimpleName() + ") failed."); } out = sun.text.Normalizer.normalize(in, NFD, - sun.text.normalizer.NormalizerBase.UNICODE_LATEST); + jdk.internal.icu.text.NormalizerBase.UNICODE_LATEST); if (!out.equals(expected.toString())) { errln("sun.text.Normalizer.normalize(" + in.getClass().getSimpleName() + ") failed."); @@ -330,7 +330,7 @@ in.getClass().getSimpleName() + ") failed."); } if (!sun.text.Normalizer.isNormalized(expected, NFD, - sun.text.normalizer.NormalizerBase.UNICODE_LATEST)) { + jdk.internal.icu.text.NormalizerBase.UNICODE_LATEST)) { errln("sun.text.Normalizer.isNormalize(" + in.getClass().getSimpleName() + ") failed."); } --- old/test/jdk/java/text/Normalizer/ThreadSafeTest.java 2020-01-10 15:57:17.000000000 -0800 +++ new/test/jdk/java/text/Normalizer/ThreadSafeTest.java 2020-01-10 15:57:17.000000000 -0800 @@ -1,5 +1,5 @@ /* - * Copyright (c) 2018, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2018, 2020, Oracle and/or its affiliates. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it @@ -23,10 +23,10 @@ /* * @test - * @bug 4221795 8032446 - * @summary Confirm that java.text.Normalizer and sun.text.Normalize are + * @bug 4221795 8032446 8174270 + * @summary Confirm that java.text.Normalizer and sun.text.Normalizer are * thread-safe. - * @modules java.base/sun.text java.base/sun.text.normalizer + * @modules java.base/sun.text java.base/jdk.internal.icu.text * @compile -XDignore.symbol.file ThreadSafeTest.java * @run main/othervm -esa ThreadSafeTest 5 10 */ @@ -115,7 +115,7 @@ sun.text.Normalizer.UNICODE_3_2); testJavaNormalize(2, java.text.Normalizer.Form.NFKD); testSunNormalize(3, java.text.Normalizer.Form.NFC, - sun.text.normalizer.NormalizerBase.UNICODE_LATEST); + jdk.internal.icu.text.NormalizerBase.UNICODE_LATEST); testJavaNormalize(4, java.text.Normalizer.Form.NFD); testIsNormalized(0, java.text.Normalizer.Form.NFKC); --- old/test/jdk/sun/net/idn/NFS4StringPrep.java 2020-01-10 15:57:18.000000000 -0800 +++ new/test/jdk/sun/net/idn/NFS4StringPrep.java 2020-01-10 15:57:18.000000000 -0800 @@ -1,5 +1,5 @@ /* - * Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2005, 2020, Oracle and/or its affiliates. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it @@ -32,8 +32,8 @@ import java.io.UnsupportedEncodingException; import java.text.ParseException; -import sun.net.idn.StringPrep; -import sun.text.normalizer.UCharacterIterator; +import jdk.internal.icu.text.StringPrep; +import jdk.internal.icu.text.UCharacterIterator; /** * @author ram --- old/test/jdk/sun/net/idn/PunycodeTest.java 2020-01-10 15:57:20.000000000 -0800 +++ new/test/jdk/sun/net/idn/PunycodeTest.java 2020-01-10 15:57:20.000000000 -0800 @@ -1,5 +1,5 @@ /* - * Copyright (c) 2005, 2006, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2005, 2020, Oracle and/or its affiliates. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it @@ -23,17 +23,16 @@ /* * @test - * @summary Unit test for sun.net.idn.Punycode - * @bug 4737170 - * @modules java.base/sun.net.idn + * @summary Unit test for jdk.internal.icu.impl.Punycode + * @bug 4737170 8174270 + * @modules java.base/jdk.internal.icu.impl * @compile -XDignore.symbol.file PunycodeTest.java * @run main/othervm -ea PunycodeTest * @author Edward Wang */ import java.util.Scanner; -import java.text.ParseException; -import sun.net.idn.Punycode; +import jdk.internal.icu.impl.Punycode; /** * unit test for Punycode that is also originated from the sample code --- old/test/jdk/sun/net/idn/TestStringPrep.java 2020-01-10 15:57:21.000000000 -0800 +++ new/test/jdk/sun/net/idn/TestStringPrep.java 2020-01-10 15:57:21.000000000 -0800 @@ -1,5 +1,5 @@ /* - * Copyright (c) 2005, 2016, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2005, 2020, Oracle and/or its affiliates. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it @@ -23,10 +23,9 @@ /* * @test - * @summary Unit test for sun.net.idn.Punycode - * @bug 4737170 8060097 - * @modules java.base/sun.net.idn:+open - * java.base/sun.text.normalizer + * @summary Unit test for jdk.internal.icu.text.StringPrep + * @bug 4737170 8060097 8174270 + * @modules java.base/jdk.internal.icu.text * @library . * @compile -XDignore.symbol.file TestStringPrep.java NFS4StringPrep.java * TestData.java @@ -44,8 +43,8 @@ import java.io.InputStream; import java.util.Locale; -import sun.net.idn.StringPrep; -import sun.text.normalizer.UCharacterIterator; +import jdk.internal.icu.text.StringPrep; +import jdk.internal.icu.text.UCharacterIterator; public class TestStringPrep { public static void main(String[] args) throws Exception { --- old/src/java.base/share/classes/sun/text/normalizer/BMPSet.java 2020-01-10 15:57:23.000000000 -0800 +++ /dev/null 2020-01-10 15:57:23.000000000 -0800 @@ -1,526 +0,0 @@ -/* - * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved. - * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. - * - * This code is free software; you can redistribute it and/or modify it - * under the terms of the GNU General Public License version 2 only, as - * published by the Free Software Foundation. Oracle designates this - * particular file as subject to the "Classpath" exception as provided - * by Oracle in the LICENSE file that accompanied this code. - * - * This code is distributed in the hope that it will be useful, but WITHOUT - * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or - * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License - * version 2 for more details (a copy is included in the LICENSE file that - * accompanied this code). - * - * You should have received a copy of the GNU General Public License version - * 2 along with this work; if not, write to the Free Software Foundation, - * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. - * - * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA - * or visit www.oracle.com if you need additional information or have any - * questions. - */ - -/* - ****************************************************************************** - * - * Copyright (C) 2009-2014, International Business Machines - * Corporation and others. All Rights Reserved. - * - ****************************************************************************** - */ - -package sun.text.normalizer; - -import sun.text.normalizer.UnicodeSet.SpanCondition; - -/** - * Helper class for frozen UnicodeSets, implements contains() and span() optimized for BMP code points. - * - * Latin-1: Look up bytes. - * 2-byte characters: Bits organized vertically. - * 3-byte characters: Use zero/one/mixed data per 64-block in U+0000..U+FFFF, with mixed for illegal ranges. - * Supplementary characters: Call contains() on the parent set. - */ -final class BMPSet { - - /** - * One boolean ('true' or 'false') per Latin-1 character. - */ - private boolean[] latin1Contains; - - /** - * One bit per code point from U+0000..U+07FF. The bits are organized vertically; consecutive code points - * correspond to the same bit positions in consecutive table words. With code point parts lead=c{10..6} - * trail=c{5..0} it is set.contains(c)==(table7FF[trail] bit lead) - * - * Bits for 0..7F (non-shortest forms) are set to the result of contains(FFFD) for faster validity checking at - * runtime. - */ - private int[] table7FF; - - /** - * One bit per 64 BMP code points. The bits are organized vertically; consecutive 64-code point blocks - * correspond to the same bit position in consecutive table words. With code point parts lead=c{15..12} - * t1=c{11..6} test bits (lead+16) and lead in bmpBlockBits[t1]. If the upper bit is 0, then the lower bit - * indicates if contains(c) for all code points in the 64-block. If the upper bit is 1, then the block is mixed - * and set.contains(c) must be called. - * - * Bits for 0..7FF (non-shortest forms) and D800..DFFF are set to the result of contains(FFFD) for faster - * validity checking at runtime. - */ - private int[] bmpBlockBits; - - /** - * Inversion list indexes for restricted binary searches in findCodePoint(), from findCodePoint(U+0800, U+1000, - * U+2000, .., U+F000, U+10000). U+0800 is the first 3-byte-UTF-8 code point. Code points below U+0800 are - * always looked up in the bit tables. The last pair of indexes is for finding supplementary code points. - */ - private int[] list4kStarts; - - /** - * The inversion list of the parent set, for the slower contains() implementation for mixed BMP blocks and for - * supplementary code points. The list is terminated with list[listLength-1]=0x110000. - */ - private final int[] list; - private final int listLength; // length used; list may be longer to minimize reallocs - - public BMPSet(final int[] parentList, int parentListLength) { - list = parentList; - listLength = parentListLength; - latin1Contains = new boolean[0x100]; - table7FF = new int[64]; - bmpBlockBits = new int[64]; - list4kStarts = new int[18]; - - /* - * Set the list indexes for binary searches for U+0800, U+1000, U+2000, .., U+F000, U+10000. U+0800 is the - * first 3-byte-UTF-8 code point. Lower code points are looked up in the bit tables. The last pair of - * indexes is for finding supplementary code points. - */ - list4kStarts[0] = findCodePoint(0x800, 0, listLength - 1); - int i; - for (i = 1; i <= 0x10; ++i) { - list4kStarts[i] = findCodePoint(i << 12, list4kStarts[i - 1], listLength - 1); - } - list4kStarts[0x11] = listLength - 1; - - initBits(); - } - - public boolean contains(int c) { - if (c <= 0xff) { - return (latin1Contains[c]); - } else if (c <= 0x7ff) { - return ((table7FF[c & 0x3f] & (1 << (c >> 6))) != 0); - } else if (c < 0xd800 || (c >= 0xe000 && c <= 0xffff)) { - int lead = c >> 12; - int twoBits = (bmpBlockBits[(c >> 6) & 0x3f] >> lead) & 0x10001; - if (twoBits <= 1) { - // All 64 code points with the same bits 15..6 - // are either in the set or not. - return (0 != twoBits); - } else { - // Look up the code point in its 4k block of code points. - return containsSlow(c, list4kStarts[lead], list4kStarts[lead + 1]); - } - } else if (c <= 0x10ffff) { - // surrogate or supplementary code point - return containsSlow(c, list4kStarts[0xd], list4kStarts[0x11]); - } else { - // Out-of-range code points get false, consistent with long-standing - // behavior of UnicodeSet.contains(c). - return false; - } - } - - /** - * Span the initial substring for which each character c has spanCondition==contains(c). It must be - * spanCondition==0 or 1. - * - * @param start The start index - * @param outCount If not null: Receives the number of code points in the span. - * @return the limit (exclusive end) of the span - * - * NOTE: to reduce the overhead of function call to contains(c), it is manually inlined here. Check for - * sufficient length for trail unit for each surrogate pair. Handle single surrogates as surrogate code points - * as usual in ICU. - */ - public final int span(CharSequence s, int start, SpanCondition spanCondition, - OutputInt outCount) { - char c, c2; - int i = start; - int limit = s.length(); - int numSupplementary = 0; - if (SpanCondition.NOT_CONTAINED != spanCondition) { - // span - while (i < limit) { - c = s.charAt(i); - if (c <= 0xff) { - if (!latin1Contains[c]) { - break; - } - } else if (c <= 0x7ff) { - if ((table7FF[c & 0x3f] & (1 << (c >> 6))) == 0) { - break; - } - } else if (c < 0xd800 || - c >= 0xdc00 || (i + 1) == limit || (c2 = s.charAt(i + 1)) < 0xdc00 || c2 >= 0xe000) { - int lead = c >> 12; - int twoBits = (bmpBlockBits[(c >> 6) & 0x3f] >> lead) & 0x10001; - if (twoBits <= 1) { - // All 64 code points with the same bits 15..6 - // are either in the set or not. - if (twoBits == 0) { - break; - } - } else { - // Look up the code point in its 4k block of code points. - if (!containsSlow(c, list4kStarts[lead], list4kStarts[lead + 1])) { - break; - } - } - } else { - // surrogate pair - int supplementary = UCharacterProperty.getRawSupplementary(c, c2); - if (!containsSlow(supplementary, list4kStarts[0x10], list4kStarts[0x11])) { - break; - } - ++numSupplementary; - ++i; - } - ++i; - } - } else { - // span not - while (i < limit) { - c = s.charAt(i); - if (c <= 0xff) { - if (latin1Contains[c]) { - break; - } - } else if (c <= 0x7ff) { - if ((table7FF[c & 0x3f] & (1 << (c >> 6))) != 0) { - break; - } - } else if (c < 0xd800 || - c >= 0xdc00 || (i + 1) == limit || (c2 = s.charAt(i + 1)) < 0xdc00 || c2 >= 0xe000) { - int lead = c >> 12; - int twoBits = (bmpBlockBits[(c >> 6) & 0x3f] >> lead) & 0x10001; - if (twoBits <= 1) { - // All 64 code points with the same bits 15..6 - // are either in the set or not. - if (twoBits != 0) { - break; - } - } else { - // Look up the code point in its 4k block of code points. - if (containsSlow(c, list4kStarts[lead], list4kStarts[lead + 1])) { - break; - } - } - } else { - // surrogate pair - int supplementary = UCharacterProperty.getRawSupplementary(c, c2); - if (containsSlow(supplementary, list4kStarts[0x10], list4kStarts[0x11])) { - break; - } - ++numSupplementary; - ++i; - } - ++i; - } - } - if (outCount != null) { - int spanLength = i - start; - outCount.value = spanLength - numSupplementary; // number of code points - } - return i; - } - - /** - * Symmetrical with span(). - * Span the trailing substring for which each character c has spanCondition==contains(c). It must be s.length >= - * limit and spanCondition==0 or 1. - * - * @return The string index which starts the span (i.e. inclusive). - */ - public final int spanBack(CharSequence s, int limit, SpanCondition spanCondition) { - char c, c2; - - if (SpanCondition.NOT_CONTAINED != spanCondition) { - // span - for (;;) { - c = s.charAt(--limit); - if (c <= 0xff) { - if (!latin1Contains[c]) { - break; - } - } else if (c <= 0x7ff) { - if ((table7FF[c & 0x3f] & (1 << (c >> 6))) == 0) { - break; - } - } else if (c < 0xd800 || - c < 0xdc00 || 0 == limit || (c2 = s.charAt(limit - 1)) < 0xd800 || c2 >= 0xdc00) { - int lead = c >> 12; - int twoBits = (bmpBlockBits[(c >> 6) & 0x3f] >> lead) & 0x10001; - if (twoBits <= 1) { - // All 64 code points with the same bits 15..6 - // are either in the set or not. - if (twoBits == 0) { - break; - } - } else { - // Look up the code point in its 4k block of code points. - if (!containsSlow(c, list4kStarts[lead], list4kStarts[lead + 1])) { - break; - } - } - } else { - // surrogate pair - int supplementary = UCharacterProperty.getRawSupplementary(c2, c); - if (!containsSlow(supplementary, list4kStarts[0x10], list4kStarts[0x11])) { - break; - } - --limit; - } - if (0 == limit) { - return 0; - } - } - } else { - // span not - for (;;) { - c = s.charAt(--limit); - if (c <= 0xff) { - if (latin1Contains[c]) { - break; - } - } else if (c <= 0x7ff) { - if ((table7FF[c & 0x3f] & (1 << (c >> 6))) != 0) { - break; - } - } else if (c < 0xd800 || - c < 0xdc00 || 0 == limit || (c2 = s.charAt(limit - 1)) < 0xd800 || c2 >= 0xdc00) { - int lead = c >> 12; - int twoBits = (bmpBlockBits[(c >> 6) & 0x3f] >> lead) & 0x10001; - if (twoBits <= 1) { - // All 64 code points with the same bits 15..6 - // are either in the set or not. - if (twoBits != 0) { - break; - } - } else { - // Look up the code point in its 4k block of code points. - if (containsSlow(c, list4kStarts[lead], list4kStarts[lead + 1])) { - break; - } - } - } else { - // surrogate pair - int supplementary = UCharacterProperty.getRawSupplementary(c2, c); - if (containsSlow(supplementary, list4kStarts[0x10], list4kStarts[0x11])) { - break; - } - --limit; - } - if (0 == limit) { - return 0; - } - } - } - return limit + 1; - } - - /** - * Set bits in a bit rectangle in "vertical" bit organization. start> 6; // Named for UTF-8 2-byte lead byte with upper 5 bits. - int trail = start & 0x3f; // Named for UTF-8 2-byte trail byte with lower 6 bits. - - // Set one bit indicating an all-one block. - int bits = 1 << lead; - if ((start + 1) == limit) { // Single-character shortcut. - table[trail] |= bits; - return; - } - - int limitLead = limit >> 6; - int limitTrail = limit & 0x3f; - - if (lead == limitLead) { - // Partial vertical bit column. - while (trail < limitTrail) { - table[trail++] |= bits; - } - } else { - // Partial vertical bit column, - // followed by a bit rectangle, - // followed by another partial vertical bit column. - if (trail > 0) { - do { - table[trail++] |= bits; - } while (trail < 64); - ++lead; - } - if (lead < limitLead) { - bits = ~((1 << lead) - 1); - if (limitLead < 0x20) { - bits &= (1 << limitLead) - 1; - } - for (trail = 0; trail < 64; ++trail) { - table[trail] |= bits; - } - } - // limit<=0x800. If limit==0x800 then limitLead=32 and limitTrail=0. - // In that case, bits=1<= 0x100) { - break; - } - do { - latin1Contains[start++] = true; - } while (start < limit && start < 0x100); - } while (limit <= 0x100); - - // Set table7FF[]. - while (start < 0x800) { - set32x64Bits(table7FF, start, limit <= 0x800 ? limit : 0x800); - if (limit > 0x800) { - start = 0x800; - break; - } - - start = list[listIndex++]; - if (listIndex < listLength) { - limit = list[listIndex++]; - } else { - limit = 0x110000; - } - } - - // Set bmpBlockBits[]. - int minStart = 0x800; - while (start < 0x10000) { - if (limit > 0x10000) { - limit = 0x10000; - } - - if (start < minStart) { - start = minStart; - } - if (start < limit) { // Else: Another range entirely in a known mixed-value block. - if (0 != (start & 0x3f)) { - // Mixed-value block of 64 code points. - start >>= 6; - bmpBlockBits[start & 0x3f] |= 0x10001 << (start >> 6); - start = (start + 1) << 6; // Round up to the next block boundary. - minStart = start; // Ignore further ranges in this block. - } - if (start < limit) { - if (start < (limit & ~0x3f)) { - // Multiple all-ones blocks of 64 code points each. - set32x64Bits(bmpBlockBits, start >> 6, limit >> 6); - } - - if (0 != (limit & 0x3f)) { - // Mixed-value block of 64 code points. - limit >>= 6; - bmpBlockBits[limit & 0x3f] |= 0x10001 << (limit >> 6); - limit = (limit + 1) << 6; // Round up to the next block boundary. - minStart = limit; // Ignore further ranges in this block. - } - } - } - - if (limit == 0x10000) { - break; - } - - start = list[listIndex++]; - if (listIndex < listLength) { - limit = list[listIndex++]; - } else { - limit = 0x110000; - } - } - } - - /** - * Same as UnicodeSet.findCodePoint(int c) except that the binary search is restricted for finding code - * points in a certain range. - * - * For restricting the search for finding in the range start..end, pass in lo=findCodePoint(start) and - * hi=findCodePoint(end) with 0<=lo<=hi= hi || c >= list[hi - 1]) - return hi; - // invariant: c >= list[lo] - // invariant: c < list[hi] - for (;;) { - int i = (lo + hi) >>> 1; - if (i == lo) { - break; // Found! - } else if (c < list[i]) { - hi = i; - } else { - lo = i; - } - } - return hi; - } - - private final boolean containsSlow(int c, int lo, int hi) { - return (0 != (findCodePoint(c, lo, hi) & 1)); - } -} - --- /dev/null 2020-01-10 15:57:23.000000000 -0800 +++ new/src/java.base/share/classes/jdk/internal/icu/impl/BMPSet.java 2020-01-10 15:57:22.000000000 -0800 @@ -0,0 +1,527 @@ +/* + * Copyright (c) 2015, 2020, Oracle and/or its affiliates. All rights reserved. + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This code is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License version 2 only, as + * published by the Free Software Foundation. Oracle designates this + * particular file as subject to the "Classpath" exception as provided + * by Oracle in the LICENSE file that accompanied this code. + * + * This code is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License + * version 2 for more details (a copy is included in the LICENSE file that + * accompanied this code). + * + * You should have received a copy of the GNU General Public License version + * 2 along with this work; if not, write to the Free Software Foundation, + * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. + * + * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA + * or visit www.oracle.com if you need additional information or have any + * questions. + */ + +/* + ****************************************************************************** + * + * Copyright (C) 2009-2014, International Business Machines + * Corporation and others. All Rights Reserved. + * + ****************************************************************************** + */ + +package jdk.internal.icu.impl; + +import jdk.internal.icu.text.UnicodeSet.SpanCondition; +import jdk.internal.icu.util.OutputInt; + +/** + * Helper class for frozen UnicodeSets, implements contains() and span() optimized for BMP code points. + * + * Latin-1: Look up bytes. + * 2-byte characters: Bits organized vertically. + * 3-byte characters: Use zero/one/mixed data per 64-block in U+0000..U+FFFF, with mixed for illegal ranges. + * Supplementary characters: Call contains() on the parent set. + */ +public final class BMPSet { + + /** + * One boolean ('true' or 'false') per Latin-1 character. + */ + private boolean[] latin1Contains; + + /** + * One bit per code point from U+0000..U+07FF. The bits are organized vertically; consecutive code points + * correspond to the same bit positions in consecutive table words. With code point parts lead=c{10..6} + * trail=c{5..0} it is set.contains(c)==(table7FF[trail] bit lead) + * + * Bits for 0..7F (non-shortest forms) are set to the result of contains(FFFD) for faster validity checking at + * runtime. + */ + private int[] table7FF; + + /** + * One bit per 64 BMP code points. The bits are organized vertically; consecutive 64-code point blocks + * correspond to the same bit position in consecutive table words. With code point parts lead=c{15..12} + * t1=c{11..6} test bits (lead+16) and lead in bmpBlockBits[t1]. If the upper bit is 0, then the lower bit + * indicates if contains(c) for all code points in the 64-block. If the upper bit is 1, then the block is mixed + * and set.contains(c) must be called. + * + * Bits for 0..7FF (non-shortest forms) and D800..DFFF are set to the result of contains(FFFD) for faster + * validity checking at runtime. + */ + private int[] bmpBlockBits; + + /** + * Inversion list indexes for restricted binary searches in findCodePoint(), from findCodePoint(U+0800, U+1000, + * U+2000, .., U+F000, U+10000). U+0800 is the first 3-byte-UTF-8 code point. Code points below U+0800 are + * always looked up in the bit tables. The last pair of indexes is for finding supplementary code points. + */ + private int[] list4kStarts; + + /** + * The inversion list of the parent set, for the slower contains() implementation for mixed BMP blocks and for + * supplementary code points. The list is terminated with list[listLength-1]=0x110000. + */ + private final int[] list; + private final int listLength; // length used; list may be longer to minimize reallocs + + public BMPSet(final int[] parentList, int parentListLength) { + list = parentList; + listLength = parentListLength; + latin1Contains = new boolean[0x100]; + table7FF = new int[64]; + bmpBlockBits = new int[64]; + list4kStarts = new int[18]; + + /* + * Set the list indexes for binary searches for U+0800, U+1000, U+2000, .., U+F000, U+10000. U+0800 is the + * first 3-byte-UTF-8 code point. Lower code points are looked up in the bit tables. The last pair of + * indexes is for finding supplementary code points. + */ + list4kStarts[0] = findCodePoint(0x800, 0, listLength - 1); + int i; + for (i = 1; i <= 0x10; ++i) { + list4kStarts[i] = findCodePoint(i << 12, list4kStarts[i - 1], listLength - 1); + } + list4kStarts[0x11] = listLength - 1; + + initBits(); + } + + public boolean contains(int c) { + if (c <= 0xff) { + return (latin1Contains[c]); + } else if (c <= 0x7ff) { + return ((table7FF[c & 0x3f] & (1 << (c >> 6))) != 0); + } else if (c < 0xd800 || (c >= 0xe000 && c <= 0xffff)) { + int lead = c >> 12; + int twoBits = (bmpBlockBits[(c >> 6) & 0x3f] >> lead) & 0x10001; + if (twoBits <= 1) { + // All 64 code points with the same bits 15..6 + // are either in the set or not. + return (0 != twoBits); + } else { + // Look up the code point in its 4k block of code points. + return containsSlow(c, list4kStarts[lead], list4kStarts[lead + 1]); + } + } else if (c <= 0x10ffff) { + // surrogate or supplementary code point + return containsSlow(c, list4kStarts[0xd], list4kStarts[0x11]); + } else { + // Out-of-range code points get false, consistent with long-standing + // behavior of UnicodeSet.contains(c). + return false; + } + } + + /** + * Span the initial substring for which each character c has spanCondition==contains(c). It must be + * spanCondition==0 or 1. + * + * @param start The start index + * @param outCount If not null: Receives the number of code points in the span. + * @return the limit (exclusive end) of the span + * + * NOTE: to reduce the overhead of function call to contains(c), it is manually inlined here. Check for + * sufficient length for trail unit for each surrogate pair. Handle single surrogates as surrogate code points + * as usual in ICU. + */ + public final int span(CharSequence s, int start, SpanCondition spanCondition, + OutputInt outCount) { + char c, c2; + int i = start; + int limit = s.length(); + int numSupplementary = 0; + if (SpanCondition.NOT_CONTAINED != spanCondition) { + // span + while (i < limit) { + c = s.charAt(i); + if (c <= 0xff) { + if (!latin1Contains[c]) { + break; + } + } else if (c <= 0x7ff) { + if ((table7FF[c & 0x3f] & (1 << (c >> 6))) == 0) { + break; + } + } else if (c < 0xd800 || + c >= 0xdc00 || (i + 1) == limit || (c2 = s.charAt(i + 1)) < 0xdc00 || c2 >= 0xe000) { + int lead = c >> 12; + int twoBits = (bmpBlockBits[(c >> 6) & 0x3f] >> lead) & 0x10001; + if (twoBits <= 1) { + // All 64 code points with the same bits 15..6 + // are either in the set or not. + if (twoBits == 0) { + break; + } + } else { + // Look up the code point in its 4k block of code points. + if (!containsSlow(c, list4kStarts[lead], list4kStarts[lead + 1])) { + break; + } + } + } else { + // surrogate pair + int supplementary = UCharacterProperty.getRawSupplementary(c, c2); + if (!containsSlow(supplementary, list4kStarts[0x10], list4kStarts[0x11])) { + break; + } + ++numSupplementary; + ++i; + } + ++i; + } + } else { + // span not + while (i < limit) { + c = s.charAt(i); + if (c <= 0xff) { + if (latin1Contains[c]) { + break; + } + } else if (c <= 0x7ff) { + if ((table7FF[c & 0x3f] & (1 << (c >> 6))) != 0) { + break; + } + } else if (c < 0xd800 || + c >= 0xdc00 || (i + 1) == limit || (c2 = s.charAt(i + 1)) < 0xdc00 || c2 >= 0xe000) { + int lead = c >> 12; + int twoBits = (bmpBlockBits[(c >> 6) & 0x3f] >> lead) & 0x10001; + if (twoBits <= 1) { + // All 64 code points with the same bits 15..6 + // are either in the set or not. + if (twoBits != 0) { + break; + } + } else { + // Look up the code point in its 4k block of code points. + if (containsSlow(c, list4kStarts[lead], list4kStarts[lead + 1])) { + break; + } + } + } else { + // surrogate pair + int supplementary = UCharacterProperty.getRawSupplementary(c, c2); + if (containsSlow(supplementary, list4kStarts[0x10], list4kStarts[0x11])) { + break; + } + ++numSupplementary; + ++i; + } + ++i; + } + } + if (outCount != null) { + int spanLength = i - start; + outCount.value = spanLength - numSupplementary; // number of code points + } + return i; + } + + /** + * Symmetrical with span(). + * Span the trailing substring for which each character c has spanCondition==contains(c). It must be s.length >= + * limit and spanCondition==0 or 1. + * + * @return The string index which starts the span (i.e. inclusive). + */ + public final int spanBack(CharSequence s, int limit, SpanCondition spanCondition) { + char c, c2; + + if (SpanCondition.NOT_CONTAINED != spanCondition) { + // span + for (;;) { + c = s.charAt(--limit); + if (c <= 0xff) { + if (!latin1Contains[c]) { + break; + } + } else if (c <= 0x7ff) { + if ((table7FF[c & 0x3f] & (1 << (c >> 6))) == 0) { + break; + } + } else if (c < 0xd800 || + c < 0xdc00 || 0 == limit || (c2 = s.charAt(limit - 1)) < 0xd800 || c2 >= 0xdc00) { + int lead = c >> 12; + int twoBits = (bmpBlockBits[(c >> 6) & 0x3f] >> lead) & 0x10001; + if (twoBits <= 1) { + // All 64 code points with the same bits 15..6 + // are either in the set or not. + if (twoBits == 0) { + break; + } + } else { + // Look up the code point in its 4k block of code points. + if (!containsSlow(c, list4kStarts[lead], list4kStarts[lead + 1])) { + break; + } + } + } else { + // surrogate pair + int supplementary = UCharacterProperty.getRawSupplementary(c2, c); + if (!containsSlow(supplementary, list4kStarts[0x10], list4kStarts[0x11])) { + break; + } + --limit; + } + if (0 == limit) { + return 0; + } + } + } else { + // span not + for (;;) { + c = s.charAt(--limit); + if (c <= 0xff) { + if (latin1Contains[c]) { + break; + } + } else if (c <= 0x7ff) { + if ((table7FF[c & 0x3f] & (1 << (c >> 6))) != 0) { + break; + } + } else if (c < 0xd800 || + c < 0xdc00 || 0 == limit || (c2 = s.charAt(limit - 1)) < 0xd800 || c2 >= 0xdc00) { + int lead = c >> 12; + int twoBits = (bmpBlockBits[(c >> 6) & 0x3f] >> lead) & 0x10001; + if (twoBits <= 1) { + // All 64 code points with the same bits 15..6 + // are either in the set or not. + if (twoBits != 0) { + break; + } + } else { + // Look up the code point in its 4k block of code points. + if (containsSlow(c, list4kStarts[lead], list4kStarts[lead + 1])) { + break; + } + } + } else { + // surrogate pair + int supplementary = UCharacterProperty.getRawSupplementary(c2, c); + if (containsSlow(supplementary, list4kStarts[0x10], list4kStarts[0x11])) { + break; + } + --limit; + } + if (0 == limit) { + return 0; + } + } + } + return limit + 1; + } + + /** + * Set bits in a bit rectangle in "vertical" bit organization. start> 6; // Named for UTF-8 2-byte lead byte with upper 5 bits. + int trail = start & 0x3f; // Named for UTF-8 2-byte trail byte with lower 6 bits. + + // Set one bit indicating an all-one block. + int bits = 1 << lead; + if ((start + 1) == limit) { // Single-character shortcut. + table[trail] |= bits; + return; + } + + int limitLead = limit >> 6; + int limitTrail = limit & 0x3f; + + if (lead == limitLead) { + // Partial vertical bit column. + while (trail < limitTrail) { + table[trail++] |= bits; + } + } else { + // Partial vertical bit column, + // followed by a bit rectangle, + // followed by another partial vertical bit column. + if (trail > 0) { + do { + table[trail++] |= bits; + } while (trail < 64); + ++lead; + } + if (lead < limitLead) { + bits = ~((1 << lead) - 1); + if (limitLead < 0x20) { + bits &= (1 << limitLead) - 1; + } + for (trail = 0; trail < 64; ++trail) { + table[trail] |= bits; + } + } + // limit<=0x800. If limit==0x800 then limitLead=32 and limitTrail=0. + // In that case, bits=1<= 0x100) { + break; + } + do { + latin1Contains[start++] = true; + } while (start < limit && start < 0x100); + } while (limit <= 0x100); + + // Set table7FF[]. + while (start < 0x800) { + set32x64Bits(table7FF, start, limit <= 0x800 ? limit : 0x800); + if (limit > 0x800) { + start = 0x800; + break; + } + + start = list[listIndex++]; + if (listIndex < listLength) { + limit = list[listIndex++]; + } else { + limit = 0x110000; + } + } + + // Set bmpBlockBits[]. + int minStart = 0x800; + while (start < 0x10000) { + if (limit > 0x10000) { + limit = 0x10000; + } + + if (start < minStart) { + start = minStart; + } + if (start < limit) { // Else: Another range entirely in a known mixed-value block. + if (0 != (start & 0x3f)) { + // Mixed-value block of 64 code points. + start >>= 6; + bmpBlockBits[start & 0x3f] |= 0x10001 << (start >> 6); + start = (start + 1) << 6; // Round up to the next block boundary. + minStart = start; // Ignore further ranges in this block. + } + if (start < limit) { + if (start < (limit & ~0x3f)) { + // Multiple all-ones blocks of 64 code points each. + set32x64Bits(bmpBlockBits, start >> 6, limit >> 6); + } + + if (0 != (limit & 0x3f)) { + // Mixed-value block of 64 code points. + limit >>= 6; + bmpBlockBits[limit & 0x3f] |= 0x10001 << (limit >> 6); + limit = (limit + 1) << 6; // Round up to the next block boundary. + minStart = limit; // Ignore further ranges in this block. + } + } + } + + if (limit == 0x10000) { + break; + } + + start = list[listIndex++]; + if (listIndex < listLength) { + limit = list[listIndex++]; + } else { + limit = 0x110000; + } + } + } + + /** + * Same as UnicodeSet.findCodePoint(int c) except that the binary search is restricted for finding code + * points in a certain range. + * + * For restricting the search for finding in the range start..end, pass in lo=findCodePoint(start) and + * hi=findCodePoint(end) with 0<=lo<=hi= hi || c >= list[hi - 1]) + return hi; + // invariant: c >= list[lo] + // invariant: c < list[hi] + for (;;) { + int i = (lo + hi) >>> 1; + if (i == lo) { + break; // Found! + } else if (c < list[i]) { + hi = i; + } else { + lo = i; + } + } + return hi; + } + + private final boolean containsSlow(int c, int lo, int hi) { + return (0 != (findCodePoint(c, lo, hi) & 1)); + } +} + --- old/src/java.base/share/classes/sun/text/normalizer/CharTrie.java 2020-01-10 15:57:24.000000000 -0800 +++ /dev/null 2020-01-10 15:57:24.000000000 -0800 @@ -1,175 +0,0 @@ -/* - * Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved. - * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. - * - * This code is free software; you can redistribute it and/or modify it - * under the terms of the GNU General Public License version 2 only, as - * published by the Free Software Foundation. Oracle designates this - * particular file as subject to the "Classpath" exception as provided - * by Oracle in the LICENSE file that accompanied this code. - * - * This code is distributed in the hope that it will be useful, but WITHOUT - * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or - * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License - * version 2 for more details (a copy is included in the LICENSE file that - * accompanied this code). - * - * You should have received a copy of the GNU General Public License version - * 2 along with this work; if not, write to the Free Software Foundation, - * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. - * - * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA - * or visit www.oracle.com if you need additional information or have any - * questions. - */ - -/* - ****************************************************************************** - * Copyright (C) 1996-2014, International Business Machines Corporation and - * others. All Rights Reserved. - ****************************************************************************** - */ - -package sun.text.normalizer; - -import java.io.DataInputStream; -import java.io.InputStream; -import java.io.IOException; - -/** - * Trie implementation which stores data in char, 16 bits. - * @author synwee - * @see com.ibm.icu.impl.Trie - * @since release 2.1, Jan 01 2002 - */ - - // note that i need to handle the block calculations later, since chartrie - // in icu4c uses the same index array. -public class CharTrie extends Trie -{ - // public constructors --------------------------------------------- - - /** - *

Creates a new Trie with the settings for the trie data.

- *

Unserialize the 32-bit-aligned input stream and use the data for the - * trie.

- * @param inputStream file input stream to a ICU data file, containing - * the trie - * @param dataManipulate object which provides methods to parse the char - * data - * @throws IOException thrown when data reading fails - * @draft 2.1 - */ - public CharTrie(InputStream inputStream, - DataManipulate dataManipulate) throws IOException - { - super(inputStream, dataManipulate); - - if (!isCharTrie()) { - throw new IllegalArgumentException( - "Data given does not belong to a char trie."); - } - } - - // public methods -------------------------------------------------- - - /** - * Gets the value associated with the codepoint. - * If no value is associated with the codepoint, a default value will be - * returned. - * @param ch codepoint - * @return offset to data - */ - public final char getCodePointValue(int ch) - { - int offset; - - // fastpath for U+0000..U+D7FF - if(0 <= ch && ch < UTF16.LEAD_SURROGATE_MIN_VALUE) { - // copy of getRawOffset() - offset = (m_index_[ch >> INDEX_STAGE_1_SHIFT_] << INDEX_STAGE_2_SHIFT_) - + (ch & INDEX_STAGE_3_MASK_); - return m_data_[offset]; - } - - // handle U+D800..U+10FFFF - offset = getCodePointOffset(ch); - - // return -1 if there is an error, in this case we return the default - // value: m_initialValue_ - return (offset >= 0) ? m_data_[offset] : m_initialValue_; - } - - /** - * Gets the value to the data which this lead surrogate character points - * to. - * Returned data may contain folding offset information for the next - * trailing surrogate character. - * This method does not guarantee correct results for trail surrogates. - * @param ch lead surrogate character - * @return data value - */ - public final char getLeadValue(char ch) - { - return m_data_[getLeadOffset(ch)]; - } - - // protected methods ----------------------------------------------- - - /** - *

Parses the input stream and stores its trie content into a index and - * data array

- * @param inputStream data input stream containing trie data - * @exception IOException thrown when data reading fails - */ - protected final void unserialize(InputStream inputStream) - throws IOException - { - DataInputStream input = new DataInputStream(inputStream); - int indexDataLength = m_dataOffset_ + m_dataLength_; - m_index_ = new char[indexDataLength]; - for (int i = 0; i < indexDataLength; i ++) { - m_index_[i] = input.readChar(); - } - m_data_ = m_index_; - m_initialValue_ = m_data_[m_dataOffset_]; - } - - /** - * Gets the offset to the data which the surrogate pair points to. - * @param lead lead surrogate - * @param trail trailing surrogate - * @return offset to data - * @draft 2.1 - */ - protected final int getSurrogateOffset(char lead, char trail) - { - if (m_dataManipulate_ == null) { - throw new NullPointerException( - "The field DataManipulate in this Trie is null"); - } - - // get fold position for the next trail surrogate - int offset = m_dataManipulate_.getFoldingOffset(getLeadValue(lead)); - - // get the real data from the folded lead/trail units - if (offset > 0) { - return getRawOffset(offset, (char)(trail & SURROGATE_MASK_)); - } - - // return -1 if there is an error, in this case we return the default - // value: m_initialValue_ - return -1; - } - - // private data members -------------------------------------------- - - /** - * Default value - */ - private char m_initialValue_; - /** - * Array of char data - */ - private char m_data_[]; -} --- /dev/null 2020-01-10 15:57:24.000000000 -0800 +++ new/src/java.base/share/classes/jdk/internal/icu/impl/CharTrie.java 2020-01-10 15:57:24.000000000 -0800 @@ -0,0 +1,177 @@ +/* + * Copyright (c) 2005, 2020, Oracle and/or its affiliates. All rights reserved. + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This code is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License version 2 only, as + * published by the Free Software Foundation. Oracle designates this + * particular file as subject to the "Classpath" exception as provided + * by Oracle in the LICENSE file that accompanied this code. + * + * This code is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License + * version 2 for more details (a copy is included in the LICENSE file that + * accompanied this code). + * + * You should have received a copy of the GNU General Public License version + * 2 along with this work; if not, write to the Free Software Foundation, + * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. + * + * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA + * or visit www.oracle.com if you need additional information or have any + * questions. + */ + +/* + ****************************************************************************** + * Copyright (C) 1996-2014, International Business Machines Corporation and + * others. All Rights Reserved. + ****************************************************************************** + */ + +package jdk.internal.icu.impl; + +import jdk.internal.icu.text.UTF16; + +import java.io.DataInputStream; +import java.io.InputStream; +import java.io.IOException; + +/** + * Trie implementation which stores data in char, 16 bits. + * @author synwee + * @see com.ibm.icu.impl.Trie + * @since release 2.1, Jan 01 2002 + */ + + // note that i need to handle the block calculations later, since chartrie + // in icu4c uses the same index array. +public class CharTrie extends Trie +{ + // public constructors --------------------------------------------- + + /** + *

Creates a new Trie with the settings for the trie data.

+ *

Unserialize the 32-bit-aligned input stream and use the data for the + * trie.

+ * @param inputStream file input stream to a ICU data file, containing + * the trie + * @param dataManipulate object which provides methods to parse the char + * data + * @throws IOException thrown when data reading fails + * @draft 2.1 + */ + public CharTrie(InputStream inputStream, + DataManipulate dataManipulate) throws IOException + { + super(inputStream, dataManipulate); + + if (!isCharTrie()) { + throw new IllegalArgumentException( + "Data given does not belong to a char trie."); + } + } + + // public methods -------------------------------------------------- + + /** + * Gets the value associated with the codepoint. + * If no value is associated with the codepoint, a default value will be + * returned. + * @param ch codepoint + * @return offset to data + */ + public final char getCodePointValue(int ch) + { + int offset; + + // fastpath for U+0000..U+D7FF + if(0 <= ch && ch < UTF16.LEAD_SURROGATE_MIN_VALUE) { + // copy of getRawOffset() + offset = (m_index_[ch >> INDEX_STAGE_1_SHIFT_] << INDEX_STAGE_2_SHIFT_) + + (ch & INDEX_STAGE_3_MASK_); + return m_data_[offset]; + } + + // handle U+D800..U+10FFFF + offset = getCodePointOffset(ch); + + // return -1 if there is an error, in this case we return the default + // value: m_initialValue_ + return (offset >= 0) ? m_data_[offset] : m_initialValue_; + } + + /** + * Gets the value to the data which this lead surrogate character points + * to. + * Returned data may contain folding offset information for the next + * trailing surrogate character. + * This method does not guarantee correct results for trail surrogates. + * @param ch lead surrogate character + * @return data value + */ + public final char getLeadValue(char ch) + { + return m_data_[getLeadOffset(ch)]; + } + + // protected methods ----------------------------------------------- + + /** + *

Parses the input stream and stores its trie content into a index and + * data array

+ * @param inputStream data input stream containing trie data + * @exception IOException thrown when data reading fails + */ + protected final void unserialize(InputStream inputStream) + throws IOException + { + DataInputStream input = new DataInputStream(inputStream); + int indexDataLength = m_dataOffset_ + m_dataLength_; + m_index_ = new char[indexDataLength]; + for (int i = 0; i < indexDataLength; i ++) { + m_index_[i] = input.readChar(); + } + m_data_ = m_index_; + m_initialValue_ = m_data_[m_dataOffset_]; + } + + /** + * Gets the offset to the data which the surrogate pair points to. + * @param lead lead surrogate + * @param trail trailing surrogate + * @return offset to data + * @draft 2.1 + */ + protected final int getSurrogateOffset(char lead, char trail) + { + if (m_dataManipulate_ == null) { + throw new NullPointerException( + "The field DataManipulate in this Trie is null"); + } + + // get fold position for the next trail surrogate + int offset = m_dataManipulate_.getFoldingOffset(getLeadValue(lead)); + + // get the real data from the folded lead/trail units + if (offset > 0) { + return getRawOffset(offset, (char)(trail & SURROGATE_MASK_)); + } + + // return -1 if there is an error, in this case we return the default + // value: m_initialValue_ + return -1; + } + + // private data members -------------------------------------------- + + /** + * Default value + */ + private char m_initialValue_; + /** + * Array of char data + */ + private char m_data_[]; +} --- old/src/java.base/share/classes/sun/text/normalizer/CharacterIteratorWrapper.java 2020-01-10 15:57:26.000000000 -0800 +++ /dev/null 2020-01-10 15:57:26.000000000 -0800 @@ -1,145 +0,0 @@ -/* - * Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved. - * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. - * - * This code is free software; you can redistribute it and/or modify it - * under the terms of the GNU General Public License version 2 only, as - * published by the Free Software Foundation. Oracle designates this - * particular file as subject to the "Classpath" exception as provided - * by Oracle in the LICENSE file that accompanied this code. - * - * This code is distributed in the hope that it will be useful, but WITHOUT - * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or - * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License - * version 2 for more details (a copy is included in the LICENSE file that - * accompanied this code). - * - * You should have received a copy of the GNU General Public License version - * 2 along with this work; if not, write to the Free Software Foundation, - * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. - * - * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA - * or visit www.oracle.com if you need additional information or have any - * questions. - */ - -/* - ******************************************************************************* - * (C) Copyright IBM Corp. 1996-2005 - All Rights Reserved * - * * - * The original version of this source code and documentation is copyrighted * - * and owned by IBM, These materials are provided under terms of a License * - * Agreement between IBM and Sun. This technology is protected by multiple * - * US and International patents. This notice and attribution to IBM may not * - * to removed. * - ******************************************************************************* - */ - -package sun.text.normalizer; - -import java.text.CharacterIterator; - -/** - * This class is a wrapper around CharacterIterator and implements the - * UCharacterIterator protocol - * @author ram - */ - -class CharacterIteratorWrapper extends UCharacterIterator { - - private CharacterIterator iterator; - - public CharacterIteratorWrapper(CharacterIterator iter){ - if(iter==null){ - throw new IllegalArgumentException(); - } - iterator = iter; - } - - /** - * @see UCharacterIterator#current() - */ - public int current() { - int c = iterator.current(); - if(c==CharacterIterator.DONE){ - return DONE; - } - return c; - } - - /** - * @see UCharacterIterator#getLength() - */ - public int getLength() { - return (iterator.getEndIndex() - iterator.getBeginIndex()); - } - - /** - * @see UCharacterIterator#getIndex() - */ - public int getIndex() { - return iterator.getIndex(); - } - - /** - * @see UCharacterIterator#next() - */ - public int next() { - int i = iterator.current(); - iterator.next(); - if(i==CharacterIterator.DONE){ - return DONE; - } - return i; - } - - /** - * @see UCharacterIterator#previous() - */ - public int previous() { - int i = iterator.previous(); - if(i==CharacterIterator.DONE){ - return DONE; - } - return i; - } - - /** - * @see UCharacterIterator#setIndex(int) - */ - public void setIndex(int index) { - iterator.setIndex(index); - } - - /** - * @see UCharacterIterator#getText(char[]) - */ - public int getText(char[] fillIn, int offset){ - int length =iterator.getEndIndex() - iterator.getBeginIndex(); - int currentIndex = iterator.getIndex(); - if(offset < 0 || offset + length > fillIn.length){ - throw new IndexOutOfBoundsException(Integer.toString(length)); - } - - for (char ch = iterator.first(); ch != CharacterIterator.DONE; ch = iterator.next()) { - fillIn[offset++] = ch; - } - iterator.setIndex(currentIndex); - - return length; - } - - /** - * Creates a clone of this iterator. Clones the underlying character iterator. - * @see UCharacterIterator#clone() - */ - public Object clone(){ - try { - CharacterIteratorWrapper result = (CharacterIteratorWrapper) super.clone(); - result.iterator = (CharacterIterator)this.iterator.clone(); - return result; - } catch (CloneNotSupportedException e) { - return null; // only invoked if bad underlying character iterator - } - } -} --- /dev/null 2020-01-10 15:57:26.000000000 -0800 +++ new/src/java.base/share/classes/jdk/internal/icu/impl/CharacterIteratorWrapper.java 2020-01-10 15:57:25.000000000 -0800 @@ -0,0 +1,147 @@ +/* + * Copyright (c) 2005, 2020, Oracle and/or its affiliates. All rights reserved. + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This code is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License version 2 only, as + * published by the Free Software Foundation. Oracle designates this + * particular file as subject to the "Classpath" exception as provided + * by Oracle in the LICENSE file that accompanied this code. + * + * This code is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License + * version 2 for more details (a copy is included in the LICENSE file that + * accompanied this code). + * + * You should have received a copy of the GNU General Public License version + * 2 along with this work; if not, write to the Free Software Foundation, + * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. + * + * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA + * or visit www.oracle.com if you need additional information or have any + * questions. + */ + +/* + ******************************************************************************* + * (C) Copyright IBM Corp. 1996-2005 - All Rights Reserved * + * * + * The original version of this source code and documentation is copyrighted * + * and owned by IBM, These materials are provided under terms of a License * + * Agreement between IBM and Sun. This technology is protected by multiple * + * US and International patents. This notice and attribution to IBM may not * + * to removed. * + ******************************************************************************* + */ + +package jdk.internal.icu.impl; + +import java.text.CharacterIterator; + +import jdk.internal.icu.text.UCharacterIterator; + +/** + * This class is a wrapper around CharacterIterator and implements the + * UCharacterIterator protocol + * @author ram + */ + +public class CharacterIteratorWrapper extends UCharacterIterator { + + private CharacterIterator iterator; + + public CharacterIteratorWrapper(CharacterIterator iter){ + if(iter==null){ + throw new IllegalArgumentException(); + } + iterator = iter; + } + + /** + * @see UCharacterIterator#current() + */ + public int current() { + int c = iterator.current(); + if(c==CharacterIterator.DONE){ + return DONE; + } + return c; + } + + /** + * @see UCharacterIterator#getLength() + */ + public int getLength() { + return (iterator.getEndIndex() - iterator.getBeginIndex()); + } + + /** + * @see UCharacterIterator#getIndex() + */ + public int getIndex() { + return iterator.getIndex(); + } + + /** + * @see UCharacterIterator#next() + */ + public int next() { + int i = iterator.current(); + iterator.next(); + if(i==CharacterIterator.DONE){ + return DONE; + } + return i; + } + + /** + * @see UCharacterIterator#previous() + */ + public int previous() { + int i = iterator.previous(); + if(i==CharacterIterator.DONE){ + return DONE; + } + return i; + } + + /** + * @see UCharacterIterator#setIndex(int) + */ + public void setIndex(int index) { + iterator.setIndex(index); + } + + /** + * @see UCharacterIterator#getText(char[]) + */ + public int getText(char[] fillIn, int offset){ + int length =iterator.getEndIndex() - iterator.getBeginIndex(); + int currentIndex = iterator.getIndex(); + if(offset < 0 || offset + length > fillIn.length){ + throw new IndexOutOfBoundsException(Integer.toString(length)); + } + + for (char ch = iterator.first(); ch != CharacterIterator.DONE; ch = iterator.next()) { + fillIn[offset++] = ch; + } + iterator.setIndex(currentIndex); + + return length; + } + + /** + * Creates a clone of this iterator. Clones the underlying character iterator. + * @see UCharacterIterator#clone() + */ + public Object clone(){ + try { + CharacterIteratorWrapper result = (CharacterIteratorWrapper) super.clone(); + result.iterator = (CharacterIterator)this.iterator.clone(); + return result; + } catch (CloneNotSupportedException e) { + return null; // only invoked if bad underlying character iterator + } + } +} --- old/src/java.base/share/classes/sun/text/normalizer/ICUBinary.java 2020-01-10 15:57:27.000000000 -0800 +++ /dev/null 2020-01-10 15:57:27.000000000 -0800 @@ -1,323 +0,0 @@ -/* - * Copyright (c) 2003, 2019, Oracle and/or its affiliates. All rights reserved. - * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. - * - * This code is free software; you can redistribute it and/or modify it - * under the terms of the GNU General Public License version 2 only, as - * published by the Free Software Foundation. Oracle designates this - * particular file as subject to the "Classpath" exception as provided - * by Oracle in the LICENSE file that accompanied this code. - * - * This code is distributed in the hope that it will be useful, but WITHOUT - * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or - * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License - * version 2 for more details (a copy is included in the LICENSE file that - * accompanied this code). - * - * You should have received a copy of the GNU General Public License version - * 2 along with this work; if not, write to the Free Software Foundation, - * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. - * - * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA - * or visit www.oracle.com if you need additional information or have any - * questions. - */ - -/* - ******************************************************************************* - * Copyright (C) 1996-2014, International Business Machines Corporation and - * others. All Rights Reserved. - ******************************************************************************* - */ - -package sun.text.normalizer; - -import java.io.BufferedInputStream; -import java.io.DataInputStream; -import java.io.InputStream; -import java.io.IOException; -import java.io.UncheckedIOException; -import java.nio.ByteBuffer; -import java.nio.ByteOrder; -import java.util.Arrays; -import java.security.AccessController; -import java.security.PrivilegedAction; - -public final class ICUBinary { - - private static final class IsAcceptable implements Authenticate { - @Override - public boolean isDataVersionAcceptable(byte version[]) { - return version[0] == 1; - } - } - - // public inner interface ------------------------------------------------ - - /** - * Special interface for data authentication - */ - public static interface Authenticate - { - /** - * Method used in ICUBinary.readHeader() to provide data format - * authentication. - * @param version version of the current data - * @return true if dataformat is an acceptable version, false otherwise - */ - public boolean isDataVersionAcceptable(byte version[]); - } - - // public methods -------------------------------------------------------- - - /** - * Loads an ICU binary data file and returns it as a ByteBuffer. - * The buffer contents is normally read-only, but its position etc. can be modified. - * - * @param itemPath Relative ICU data item path, for example "root.res" or "coll/ucadata.icu". - * @return The data as a read-only ByteBuffer. - */ - public static ByteBuffer getRequiredData(String itemPath) { - final Class root = ICUBinary.class; - - try (InputStream is = AccessController.doPrivileged(new PrivilegedAction() { - public InputStream run() { - return root.getResourceAsStream(itemPath); - } - })) { - - // is.available() may return 0, or 1, or the total number of bytes in the stream, - // or some other number. - // Do not try to use is.available() == 0 to find the end of the stream! - byte[] bytes; - int avail = is.available(); - if (avail > 32) { - // There are more bytes available than just the ICU data header length. - // With luck, it is the total number of bytes. - bytes = new byte[avail]; - } else { - bytes = new byte[128]; // empty .res files are even smaller - } - // Call is.read(...) until one returns a negative value. - int length = 0; - for(;;) { - if (length < bytes.length) { - int numRead = is.read(bytes, length, bytes.length - length); - if (numRead < 0) { - break; // end of stream - } - length += numRead; - } else { - // See if we are at the end of the stream before we grow the array. - int nextByte = is.read(); - if (nextByte < 0) { - break; - } - int capacity = 2 * bytes.length; - if (capacity < 128) { - capacity = 128; - } else if (capacity < 0x4000) { - capacity *= 2; // Grow faster until we reach 16kB. - } - bytes = Arrays.copyOf(bytes, capacity); - bytes[length++] = (byte) nextByte; - } - } - return ByteBuffer.wrap(bytes, 0, length); - } - catch (IOException e) { - throw new UncheckedIOException(e); - } - } - - /** - * Same as readHeader(), but returns a VersionInfo rather than a compact int. - */ - public static VersionInfo readHeaderAndDataVersion(ByteBuffer bytes, - int dataFormat, - Authenticate authenticate) - throws IOException { - return getVersionInfoFromCompactInt(readHeader(bytes, dataFormat, authenticate)); - } - - private static final byte BIG_ENDIAN_ = 1; - public static final byte[] readHeader(InputStream inputStream, - byte dataFormatIDExpected[], - Authenticate authenticate) - throws IOException - { - DataInputStream input = new DataInputStream(inputStream); - char headersize = input.readChar(); - int readcount = 2; - //reading the header format - byte magic1 = input.readByte(); - readcount ++; - byte magic2 = input.readByte(); - readcount ++; - if (magic1 != MAGIC1 || magic2 != MAGIC2) { - throw new IOException(MAGIC_NUMBER_AUTHENTICATION_FAILED_); - } - - input.readChar(); // reading size - readcount += 2; - input.readChar(); // reading reserved word - readcount += 2; - byte bigendian = input.readByte(); - readcount ++; - byte charset = input.readByte(); - readcount ++; - byte charsize = input.readByte(); - readcount ++; - input.readByte(); // reading reserved byte - readcount ++; - - byte dataFormatID[] = new byte[4]; - input.readFully(dataFormatID); - readcount += 4; - byte dataVersion[] = new byte[4]; - input.readFully(dataVersion); - readcount += 4; - byte unicodeVersion[] = new byte[4]; - input.readFully(unicodeVersion); - readcount += 4; - if (headersize < readcount) { - throw new IOException("Internal Error: Header size error"); - } - input.skipBytes(headersize - readcount); - - if (bigendian != BIG_ENDIAN_ || charset != CHAR_SET_ - || charsize != CHAR_SIZE_ - || !Arrays.equals(dataFormatIDExpected, dataFormatID) - || (authenticate != null - && !authenticate.isDataVersionAcceptable(dataVersion))) { - throw new IOException(HEADER_AUTHENTICATION_FAILED_); - } - return unicodeVersion; - } - - /** - * Reads an ICU data header, checks the data format, and returns the data version. - * - *

Assumes that the ByteBuffer position is 0 on input. - * The buffer byte order is set according to the data. - * The buffer position is advanced past the header (including UDataInfo and comment). - * - *

See C++ ucmndata.h and unicode/udata.h. - * - * @return dataVersion - * @throws IOException if this is not a valid ICU data item of the expected dataFormat - */ - public static int readHeader(ByteBuffer bytes, int dataFormat, Authenticate authenticate) - throws IOException { - assert bytes.position() == 0; - byte magic1 = bytes.get(2); - byte magic2 = bytes.get(3); - if (magic1 != MAGIC1 || magic2 != MAGIC2) { - throw new IOException(MAGIC_NUMBER_AUTHENTICATION_FAILED_); - } - - byte isBigEndian = bytes.get(8); - byte charsetFamily = bytes.get(9); - byte sizeofUChar = bytes.get(10); - if (isBigEndian < 0 || 1 < isBigEndian || - charsetFamily != CHAR_SET_ || sizeofUChar != CHAR_SIZE_) { - throw new IOException(HEADER_AUTHENTICATION_FAILED_); - } - bytes.order(isBigEndian != 0 ? ByteOrder.BIG_ENDIAN : ByteOrder.LITTLE_ENDIAN); - - int headerSize = bytes.getChar(0); - int sizeofUDataInfo = bytes.getChar(4); - if (sizeofUDataInfo < 20 || headerSize < (sizeofUDataInfo + 4)) { - throw new IOException("Internal Error: Header size error"); - } - // TODO: Change Authenticate to take int major, int minor, int milli, int micro - // to avoid array allocation. - byte[] formatVersion = new byte[] { - bytes.get(16), bytes.get(17), bytes.get(18), bytes.get(19) - }; - if (bytes.get(12) != (byte)(dataFormat >> 24) || - bytes.get(13) != (byte)(dataFormat >> 16) || - bytes.get(14) != (byte)(dataFormat >> 8) || - bytes.get(15) != (byte)dataFormat || - (authenticate != null && !authenticate.isDataVersionAcceptable(formatVersion))) { - throw new IOException(HEADER_AUTHENTICATION_FAILED_ + - String.format("; data format %02x%02x%02x%02x, format version %d.%d.%d.%d", - bytes.get(12), bytes.get(13), bytes.get(14), bytes.get(15), - formatVersion[0] & 0xff, formatVersion[1] & 0xff, - formatVersion[2] & 0xff, formatVersion[3] & 0xff)); - } - - bytes.position(headerSize); - return // dataVersion - ((int)bytes.get(20) << 24) | - ((bytes.get(21) & 0xff) << 16) | - ((bytes.get(22) & 0xff) << 8) | - (bytes.get(23) & 0xff); - } - - public static void skipBytes(ByteBuffer bytes, int skipLength) { - if (skipLength > 0) { - bytes.position(bytes.position() + skipLength); - } - } - - public static byte[] getBytes(ByteBuffer bytes, int length, int additionalSkipLength) { - byte[] dest = new byte[length]; - bytes.get(dest); - if (additionalSkipLength > 0) { - skipBytes(bytes, additionalSkipLength); - } - return dest; - } - - public static String getString(ByteBuffer bytes, int length, int additionalSkipLength) { - CharSequence cs = bytes.asCharBuffer(); - String s = cs.subSequence(0, length).toString(); - skipBytes(bytes, length * 2 + additionalSkipLength); - return s; - } - - public static char[] getChars(ByteBuffer bytes, int length, int additionalSkipLength) { - char[] dest = new char[length]; - bytes.asCharBuffer().get(dest); - skipBytes(bytes, length * 2 + additionalSkipLength); - return dest; - } - - public static int[] getInts(ByteBuffer bytes, int length, int additionalSkipLength) { - int[] dest = new int[length]; - bytes.asIntBuffer().get(dest); - skipBytes(bytes, length * 4 + additionalSkipLength); - return dest; - } - - /** - * Returns a VersionInfo for the bytes in the compact version integer. - */ - public static VersionInfo getVersionInfoFromCompactInt(int version) { - return VersionInfo.getInstance( - version >>> 24, (version >> 16) & 0xff, (version >> 8) & 0xff, version & 0xff); - } - - // private variables ------------------------------------------------- - - /** - * Magic numbers to authenticate the data file - */ - private static final byte MAGIC1 = (byte)0xda; - private static final byte MAGIC2 = (byte)0x27; - - /** - * File format authentication values - */ - private static final byte CHAR_SET_ = 0; - private static final byte CHAR_SIZE_ = 2; - - /** - * Error messages - */ - private static final String MAGIC_NUMBER_AUTHENTICATION_FAILED_ = - "ICUBinary data file error: Magic number authentication failed"; - private static final String HEADER_AUTHENTICATION_FAILED_ = - "ICUBinary data file error: Header authentication failed"; -} --- /dev/null 2020-01-10 15:57:27.000000000 -0800 +++ new/src/java.base/share/classes/jdk/internal/icu/impl/ICUBinary.java 2020-01-10 15:57:27.000000000 -0800 @@ -0,0 +1,324 @@ +/* + * Copyright (c) 2003, 2020, Oracle and/or its affiliates. All rights reserved. + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This code is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License version 2 only, as + * published by the Free Software Foundation. Oracle designates this + * particular file as subject to the "Classpath" exception as provided + * by Oracle in the LICENSE file that accompanied this code. + * + * This code is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License + * version 2 for more details (a copy is included in the LICENSE file that + * accompanied this code). + * + * You should have received a copy of the GNU General Public License version + * 2 along with this work; if not, write to the Free Software Foundation, + * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. + * + * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA + * or visit www.oracle.com if you need additional information or have any + * questions. + */ + +/* + ******************************************************************************* + * Copyright (C) 1996-2014, International Business Machines Corporation and + * others. All Rights Reserved. + ******************************************************************************* + */ + +package jdk.internal.icu.impl; + +import java.io.DataInputStream; +import java.io.InputStream; +import java.io.IOException; +import java.io.UncheckedIOException; +import java.nio.ByteBuffer; +import java.nio.ByteOrder; +import java.util.Arrays; +import java.security.AccessController; +import java.security.PrivilegedAction; + +import jdk.internal.icu.util.VersionInfo; + +public final class ICUBinary { + + private static final class IsAcceptable implements Authenticate { + @Override + public boolean isDataVersionAcceptable(byte version[]) { + return version[0] == 1; + } + } + + // public inner interface ------------------------------------------------ + + /** + * Special interface for data authentication + */ + public static interface Authenticate + { + /** + * Method used in ICUBinary.readHeader() to provide data format + * authentication. + * @param version version of the current data + * @return true if dataformat is an acceptable version, false otherwise + */ + public boolean isDataVersionAcceptable(byte version[]); + } + + // public methods -------------------------------------------------------- + + /** + * Loads an ICU binary data file and returns it as a ByteBuffer. + * The buffer contents is normally read-only, but its position etc. can be modified. + * + * @param itemPath Relative ICU data item path, for example "root.res" or "coll/ucadata.icu". + * @return The data as a read-only ByteBuffer. + */ + public static ByteBuffer getRequiredData(String itemPath) { + final Class root = ICUBinary.class; + + try (InputStream is = AccessController.doPrivileged(new PrivilegedAction() { + public InputStream run() { + return root.getResourceAsStream(itemPath); + } + })) { + + // is.available() may return 0, or 1, or the total number of bytes in the stream, + // or some other number. + // Do not try to use is.available() == 0 to find the end of the stream! + byte[] bytes; + int avail = is.available(); + if (avail > 32) { + // There are more bytes available than just the ICU data header length. + // With luck, it is the total number of bytes. + bytes = new byte[avail]; + } else { + bytes = new byte[128]; // empty .res files are even smaller + } + // Call is.read(...) until one returns a negative value. + int length = 0; + for(;;) { + if (length < bytes.length) { + int numRead = is.read(bytes, length, bytes.length - length); + if (numRead < 0) { + break; // end of stream + } + length += numRead; + } else { + // See if we are at the end of the stream before we grow the array. + int nextByte = is.read(); + if (nextByte < 0) { + break; + } + int capacity = 2 * bytes.length; + if (capacity < 128) { + capacity = 128; + } else if (capacity < 0x4000) { + capacity *= 2; // Grow faster until we reach 16kB. + } + bytes = Arrays.copyOf(bytes, capacity); + bytes[length++] = (byte) nextByte; + } + } + return ByteBuffer.wrap(bytes, 0, length); + } + catch (IOException e) { + throw new UncheckedIOException(e); + } + } + + /** + * Same as readHeader(), but returns a VersionInfo rather than a compact int. + */ + public static VersionInfo readHeaderAndDataVersion(ByteBuffer bytes, + int dataFormat, + Authenticate authenticate) + throws IOException { + return getVersionInfoFromCompactInt(readHeader(bytes, dataFormat, authenticate)); + } + + private static final byte BIG_ENDIAN_ = 1; + public static final byte[] readHeader(InputStream inputStream, + byte dataFormatIDExpected[], + Authenticate authenticate) + throws IOException + { + DataInputStream input = new DataInputStream(inputStream); + char headersize = input.readChar(); + int readcount = 2; + //reading the header format + byte magic1 = input.readByte(); + readcount ++; + byte magic2 = input.readByte(); + readcount ++; + if (magic1 != MAGIC1 || magic2 != MAGIC2) { + throw new IOException(MAGIC_NUMBER_AUTHENTICATION_FAILED_); + } + + input.readChar(); // reading size + readcount += 2; + input.readChar(); // reading reserved word + readcount += 2; + byte bigendian = input.readByte(); + readcount ++; + byte charset = input.readByte(); + readcount ++; + byte charsize = input.readByte(); + readcount ++; + input.readByte(); // reading reserved byte + readcount ++; + + byte dataFormatID[] = new byte[4]; + input.readFully(dataFormatID); + readcount += 4; + byte dataVersion[] = new byte[4]; + input.readFully(dataVersion); + readcount += 4; + byte unicodeVersion[] = new byte[4]; + input.readFully(unicodeVersion); + readcount += 4; + if (headersize < readcount) { + throw new IOException("Internal Error: Header size error"); + } + input.skipBytes(headersize - readcount); + + if (bigendian != BIG_ENDIAN_ || charset != CHAR_SET_ + || charsize != CHAR_SIZE_ + || !Arrays.equals(dataFormatIDExpected, dataFormatID) + || (authenticate != null + && !authenticate.isDataVersionAcceptable(dataVersion))) { + throw new IOException(HEADER_AUTHENTICATION_FAILED_); + } + return unicodeVersion; + } + + /** + * Reads an ICU data header, checks the data format, and returns the data version. + * + *

Assumes that the ByteBuffer position is 0 on input. + * The buffer byte order is set according to the data. + * The buffer position is advanced past the header (including UDataInfo and comment). + * + *

See C++ ucmndata.h and unicode/udata.h. + * + * @return dataVersion + * @throws IOException if this is not a valid ICU data item of the expected dataFormat + */ + public static int readHeader(ByteBuffer bytes, int dataFormat, Authenticate authenticate) + throws IOException { + assert bytes.position() == 0; + byte magic1 = bytes.get(2); + byte magic2 = bytes.get(3); + if (magic1 != MAGIC1 || magic2 != MAGIC2) { + throw new IOException(MAGIC_NUMBER_AUTHENTICATION_FAILED_); + } + + byte isBigEndian = bytes.get(8); + byte charsetFamily = bytes.get(9); + byte sizeofUChar = bytes.get(10); + if (isBigEndian < 0 || 1 < isBigEndian || + charsetFamily != CHAR_SET_ || sizeofUChar != CHAR_SIZE_) { + throw new IOException(HEADER_AUTHENTICATION_FAILED_); + } + bytes.order(isBigEndian != 0 ? ByteOrder.BIG_ENDIAN : ByteOrder.LITTLE_ENDIAN); + + int headerSize = bytes.getChar(0); + int sizeofUDataInfo = bytes.getChar(4); + if (sizeofUDataInfo < 20 || headerSize < (sizeofUDataInfo + 4)) { + throw new IOException("Internal Error: Header size error"); + } + // TODO: Change Authenticate to take int major, int minor, int milli, int micro + // to avoid array allocation. + byte[] formatVersion = new byte[] { + bytes.get(16), bytes.get(17), bytes.get(18), bytes.get(19) + }; + if (bytes.get(12) != (byte)(dataFormat >> 24) || + bytes.get(13) != (byte)(dataFormat >> 16) || + bytes.get(14) != (byte)(dataFormat >> 8) || + bytes.get(15) != (byte)dataFormat || + (authenticate != null && !authenticate.isDataVersionAcceptable(formatVersion))) { + throw new IOException(HEADER_AUTHENTICATION_FAILED_ + + String.format("; data format %02x%02x%02x%02x, format version %d.%d.%d.%d", + bytes.get(12), bytes.get(13), bytes.get(14), bytes.get(15), + formatVersion[0] & 0xff, formatVersion[1] & 0xff, + formatVersion[2] & 0xff, formatVersion[3] & 0xff)); + } + + bytes.position(headerSize); + return // dataVersion + ((int)bytes.get(20) << 24) | + ((bytes.get(21) & 0xff) << 16) | + ((bytes.get(22) & 0xff) << 8) | + (bytes.get(23) & 0xff); + } + + public static void skipBytes(ByteBuffer bytes, int skipLength) { + if (skipLength > 0) { + bytes.position(bytes.position() + skipLength); + } + } + + public static byte[] getBytes(ByteBuffer bytes, int length, int additionalSkipLength) { + byte[] dest = new byte[length]; + bytes.get(dest); + if (additionalSkipLength > 0) { + skipBytes(bytes, additionalSkipLength); + } + return dest; + } + + public static String getString(ByteBuffer bytes, int length, int additionalSkipLength) { + CharSequence cs = bytes.asCharBuffer(); + String s = cs.subSequence(0, length).toString(); + skipBytes(bytes, length * 2 + additionalSkipLength); + return s; + } + + public static char[] getChars(ByteBuffer bytes, int length, int additionalSkipLength) { + char[] dest = new char[length]; + bytes.asCharBuffer().get(dest); + skipBytes(bytes, length * 2 + additionalSkipLength); + return dest; + } + + public static int[] getInts(ByteBuffer bytes, int length, int additionalSkipLength) { + int[] dest = new int[length]; + bytes.asIntBuffer().get(dest); + skipBytes(bytes, length * 4 + additionalSkipLength); + return dest; + } + + /** + * Returns a VersionInfo for the bytes in the compact version integer. + */ + public static VersionInfo getVersionInfoFromCompactInt(int version) { + return VersionInfo.getInstance( + version >>> 24, (version >> 16) & 0xff, (version >> 8) & 0xff, version & 0xff); + } + + // private variables ------------------------------------------------- + + /** + * Magic numbers to authenticate the data file + */ + private static final byte MAGIC1 = (byte)0xda; + private static final byte MAGIC2 = (byte)0x27; + + /** + * File format authentication values + */ + private static final byte CHAR_SET_ = 0; + private static final byte CHAR_SIZE_ = 2; + + /** + * Error messages + */ + private static final String MAGIC_NUMBER_AUTHENTICATION_FAILED_ = + "ICUBinary data file error: Magic number authentication failed"; + private static final String HEADER_AUTHENTICATION_FAILED_ = + "ICUBinary data file error: Header authentication failed"; +} --- old/src/java.base/share/classes/sun/text/normalizer/Norm2AllModes.java 2020-01-10 15:57:29.000000000 -0800 +++ /dev/null 2020-01-10 15:57:29.000000000 -0800 @@ -1,287 +0,0 @@ -/* - * Copyright (c) 2015, 2019, Oracle and/or its affiliates. All rights reserved. - * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. - * - * This code is free software; you can redistribute it and/or modify it - * under the terms of the GNU General Public License version 2 only, as - * published by the Free Software Foundation. Oracle designates this - * particular file as subject to the "Classpath" exception as provided - * by Oracle in the LICENSE file that accompanied this code. - * - * This code is distributed in the hope that it will be useful, but WITHOUT - * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or - * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License - * version 2 for more details (a copy is included in the LICENSE file that - * accompanied this code). - * - * You should have received a copy of the GNU General Public License version - * 2 along with this work; if not, write to the Free Software Foundation, - * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. - * - * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA - * or visit www.oracle.com if you need additional information or have any - * questions. - */ - -/* - ******************************************************************************* - * Copyright (C) 2009-2014, International Business Machines - * Corporation and others. All Rights Reserved. - ******************************************************************************* - */ - -package sun.text.normalizer; - -import java.io.IOException; - -final class Norm2AllModes { - // Public API dispatch via Normalizer2 subclasses -------------------------- *** - - // Normalizer2 implementation for the old UNORM_NONE. - public static final class NoopNormalizer2 extends Normalizer2 { - @Override - public StringBuilder normalize(CharSequence src, StringBuilder dest) { - if(dest!=src) { - dest.setLength(0); - return dest.append(src); - } else { - throw new IllegalArgumentException(); - } - } - - @Override - public Appendable normalize(CharSequence src, Appendable dest) { - if(dest!=src) { - try { - return dest.append(src); - } catch(IOException e) { - throw new InternalError(e.toString(), e); - } - } else { - throw new IllegalArgumentException(); - } - } - - @Override - public StringBuilder normalizeSecondAndAppend(StringBuilder first, CharSequence second) { - if(first!=second) { - return first.append(second); - } else { - throw new IllegalArgumentException(); - } - } - - @Override - public StringBuilder append(StringBuilder first, CharSequence second) { - if(first!=second) { - return first.append(second); - } else { - throw new IllegalArgumentException(); - } - } - - @Override - public String getDecomposition(int c) { - return null; - } - - // No need to override the default getRawDecomposition(). - @Override - public boolean isNormalized(CharSequence s) { return true; } - - @Override - public int spanQuickCheckYes(CharSequence s) { return s.length(); } - - @Override - public boolean hasBoundaryBefore(int c) { return true; } - } - - // Intermediate class: - // Has NormalizerImpl and does boilerplate argument checking and setup. - public abstract static class Normalizer2WithImpl extends Normalizer2 { - public Normalizer2WithImpl(NormalizerImpl ni) { - impl=ni; - } - - // normalize - @Override - public StringBuilder normalize(CharSequence src, StringBuilder dest) { - if(dest==src) { - throw new IllegalArgumentException(); - } - dest.setLength(0); - normalize(src, new NormalizerImpl.ReorderingBuffer(impl, dest, src.length())); - return dest; - } - - @Override - public Appendable normalize(CharSequence src, Appendable dest) { - if(dest==src) { - throw new IllegalArgumentException(); - } - NormalizerImpl.ReorderingBuffer buffer= - new NormalizerImpl.ReorderingBuffer(impl, dest, src.length()); - normalize(src, buffer); - buffer.flush(); - return dest; - } - - protected abstract void normalize(CharSequence src, NormalizerImpl.ReorderingBuffer buffer); - - // normalize and append - @Override - public StringBuilder normalizeSecondAndAppend(StringBuilder first, CharSequence second) { - return normalizeSecondAndAppend(first, second, true); - } - - @Override - public StringBuilder append(StringBuilder first, CharSequence second) { - return normalizeSecondAndAppend(first, second, false); - } - - public StringBuilder normalizeSecondAndAppend( - StringBuilder first, CharSequence second, boolean doNormalize) { - if(first==second) { - throw new IllegalArgumentException(); - } - normalizeAndAppend( - second, doNormalize, - new NormalizerImpl.ReorderingBuffer(impl, first, first.length()+second.length())); - return first; - } - - protected abstract void normalizeAndAppend( - CharSequence src, boolean doNormalize, NormalizerImpl.ReorderingBuffer buffer); - - @Override - public String getDecomposition(int c) { - return impl.getDecomposition(c); - } - - @Override - public int getCombiningClass(int c) { - return impl.getCC(impl.getNorm16(c)); - } - - // quick checks - @Override - public boolean isNormalized(CharSequence s) { - return s.length()==spanQuickCheckYes(s); - } - - public final NormalizerImpl impl; - } - - public static final class DecomposeNormalizer2 extends Normalizer2WithImpl { - public DecomposeNormalizer2(NormalizerImpl ni) { - super(ni); - } - - @Override - protected void normalize(CharSequence src, NormalizerImpl.ReorderingBuffer buffer) { - impl.decompose(src, 0, src.length(), buffer); - } - - @Override - protected void normalizeAndAppend( - CharSequence src, boolean doNormalize, NormalizerImpl.ReorderingBuffer buffer) { - impl.decomposeAndAppend(src, doNormalize, buffer); - } - - @Override - public int spanQuickCheckYes(CharSequence s) { - return impl.decompose(s, 0, s.length(), null); - } - - @Override - public boolean hasBoundaryBefore(int c) { return impl.hasDecompBoundaryBefore(c); } - } - - public static final class ComposeNormalizer2 extends Normalizer2WithImpl { - public ComposeNormalizer2(NormalizerImpl ni, boolean fcc) { - super(ni); - onlyContiguous=fcc; - } - - @Override - protected void normalize(CharSequence src, NormalizerImpl.ReorderingBuffer buffer) { - impl.compose(src, 0, src.length(), onlyContiguous, true, buffer); - } - - @Override - protected void normalizeAndAppend( - CharSequence src, boolean doNormalize, NormalizerImpl.ReorderingBuffer buffer) { - impl.composeAndAppend(src, doNormalize, onlyContiguous, buffer); - } - - @Override - public boolean isNormalized(CharSequence s) { - // 5: small destCapacity for substring normalization - return impl.compose(s, 0, s.length(), - onlyContiguous, false, - new NormalizerImpl.ReorderingBuffer(impl, new StringBuilder(), 5)); - } - - @Override - public int spanQuickCheckYes(CharSequence s) { - return impl.composeQuickCheck(s, 0, s.length(), onlyContiguous, true)>>>1; - } - - @Override - public boolean hasBoundaryBefore(int c) { return impl.hasCompBoundaryBefore(c); } - - private final boolean onlyContiguous; - } - - // instance cache ---------------------------------------------------------- *** - - private Norm2AllModes(NormalizerImpl ni) { - impl=ni; - comp=new ComposeNormalizer2(ni, false); - decomp=new DecomposeNormalizer2(ni); - } - - public final NormalizerImpl impl; - public final ComposeNormalizer2 comp; - public final DecomposeNormalizer2 decomp; - - private static Norm2AllModes getInstanceFromSingleton(Norm2AllModesSingleton singleton) { - if(singleton.exception!=null) { - throw singleton.exception; - } - return singleton.allModes; - } - - public static Norm2AllModes getNFCInstance() { - return getInstanceFromSingleton(NFCSingleton.INSTANCE); - } - - public static Norm2AllModes getNFKCInstance() { - return getInstanceFromSingleton(NFKCSingleton.INSTANCE); - } - - public static final NoopNormalizer2 NOOP_NORMALIZER2=new NoopNormalizer2(); - - private static final class Norm2AllModesSingleton { - private Norm2AllModesSingleton(String name) { - try { - String DATA_FILE_NAME = "/sun/text/resources/" + name + ".nrm"; - NormalizerImpl impl=new NormalizerImpl().load(DATA_FILE_NAME); - allModes=new Norm2AllModes(impl); - } catch (RuntimeException e) { - exception=e; - } - } - - private Norm2AllModes allModes; - private RuntimeException exception; - } - - private static final class NFCSingleton { - private static final Norm2AllModesSingleton INSTANCE=new Norm2AllModesSingleton("nfc"); - } - - private static final class NFKCSingleton { - private static final Norm2AllModesSingleton INSTANCE=new Norm2AllModesSingleton("nfkc"); - } -} --- /dev/null 2020-01-10 15:57:29.000000000 -0800 +++ new/src/java.base/share/classes/jdk/internal/icu/impl/Norm2AllModes.java 2020-01-10 15:57:29.000000000 -0800 @@ -0,0 +1,292 @@ +/* + * Copyright (c) 2015, 2020, Oracle and/or its affiliates. All rights reserved. + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This code is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License version 2 only, as + * published by the Free Software Foundation. Oracle designates this + * particular file as subject to the "Classpath" exception as provided + * by Oracle in the LICENSE file that accompanied this code. + * + * This code is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License + * version 2 for more details (a copy is included in the LICENSE file that + * accompanied this code). + * + * You should have received a copy of the GNU General Public License version + * 2 along with this work; if not, write to the Free Software Foundation, + * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. + * + * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA + * or visit www.oracle.com if you need additional information or have any + * questions. + */ + +/* + ******************************************************************************* + * Copyright (C) 2009-2014, International Business Machines + * Corporation and others. All Rights Reserved. + ******************************************************************************* + */ + +package jdk.internal.icu.impl; + +import java.io.IOException; + +import jdk.internal.icu.text.Normalizer2; +import jdk.internal.icu.util.VersionInfo; + +public final class Norm2AllModes { + // Public API dispatch via Normalizer2 subclasses -------------------------- *** + + // Normalizer2 implementation for the old UNORM_NONE. + public static final class NoopNormalizer2 extends Normalizer2 { + @Override + public StringBuilder normalize(CharSequence src, StringBuilder dest) { + if(dest!=src) { + dest.setLength(0); + return dest.append(src); + } else { + throw new IllegalArgumentException(); + } + } + + @Override + public Appendable normalize(CharSequence src, Appendable dest) { + if(dest!=src) { + try { + return dest.append(src); + } catch(IOException e) { + throw new InternalError(e.toString(), e); + } + } else { + throw new IllegalArgumentException(); + } + } + + @Override + public StringBuilder normalizeSecondAndAppend(StringBuilder first, CharSequence second) { + if(first!=second) { + return first.append(second); + } else { + throw new IllegalArgumentException(); + } + } + + @Override + public StringBuilder append(StringBuilder first, CharSequence second) { + if(first!=second) { + return first.append(second); + } else { + throw new IllegalArgumentException(); + } + } + + @Override + public String getDecomposition(int c) { + return null; + } + + // No need to override the default getRawDecomposition(). + @Override + public boolean isNormalized(CharSequence s) { return true; } + + @Override + public int spanQuickCheckYes(CharSequence s) { return s.length(); } + + @Override + public boolean hasBoundaryBefore(int c) { return true; } + } + + // Intermediate class: + // Has NormalizerImpl and does boilerplate argument checking and setup. + public abstract static class Normalizer2WithImpl extends Normalizer2 { + public Normalizer2WithImpl(NormalizerImpl ni) { + impl=ni; + } + + // normalize + @Override + public StringBuilder normalize(CharSequence src, StringBuilder dest) { + if(dest==src) { + throw new IllegalArgumentException(); + } + dest.setLength(0); + normalize(src, new NormalizerImpl.ReorderingBuffer(impl, dest, src.length())); + return dest; + } + + @Override + public Appendable normalize(CharSequence src, Appendable dest) { + if(dest==src) { + throw new IllegalArgumentException(); + } + NormalizerImpl.ReorderingBuffer buffer= + new NormalizerImpl.ReorderingBuffer(impl, dest, src.length()); + normalize(src, buffer); + buffer.flush(); + return dest; + } + + protected abstract void normalize(CharSequence src, NormalizerImpl.ReorderingBuffer buffer); + + // normalize and append + @Override + public StringBuilder normalizeSecondAndAppend(StringBuilder first, CharSequence second) { + return normalizeSecondAndAppend(first, second, true); + } + + @Override + public StringBuilder append(StringBuilder first, CharSequence second) { + return normalizeSecondAndAppend(first, second, false); + } + + public StringBuilder normalizeSecondAndAppend( + StringBuilder first, CharSequence second, boolean doNormalize) { + if(first==second) { + throw new IllegalArgumentException(); + } + normalizeAndAppend( + second, doNormalize, + new NormalizerImpl.ReorderingBuffer(impl, first, first.length()+second.length())); + return first; + } + + protected abstract void normalizeAndAppend( + CharSequence src, boolean doNormalize, NormalizerImpl.ReorderingBuffer buffer); + + @Override + public String getDecomposition(int c) { + return impl.getDecomposition(c); + } + + @Override + public int getCombiningClass(int c) { + return impl.getCC(impl.getNorm16(c)); + } + + // quick checks + @Override + public boolean isNormalized(CharSequence s) { + return s.length()==spanQuickCheckYes(s); + } + + public final NormalizerImpl impl; + } + + public static final class DecomposeNormalizer2 extends Normalizer2WithImpl { + public DecomposeNormalizer2(NormalizerImpl ni) { + super(ni); + } + + @Override + protected void normalize(CharSequence src, NormalizerImpl.ReorderingBuffer buffer) { + impl.decompose(src, 0, src.length(), buffer); + } + + @Override + protected void normalizeAndAppend( + CharSequence src, boolean doNormalize, NormalizerImpl.ReorderingBuffer buffer) { + impl.decomposeAndAppend(src, doNormalize, buffer); + } + + @Override + public int spanQuickCheckYes(CharSequence s) { + return impl.decompose(s, 0, s.length(), null); + } + + @Override + public boolean hasBoundaryBefore(int c) { return impl.hasDecompBoundaryBefore(c); } + } + + public static final class ComposeNormalizer2 extends Normalizer2WithImpl { + public ComposeNormalizer2(NormalizerImpl ni, boolean fcc) { + super(ni); + onlyContiguous=fcc; + } + + @Override + protected void normalize(CharSequence src, NormalizerImpl.ReorderingBuffer buffer) { + impl.compose(src, 0, src.length(), onlyContiguous, true, buffer); + } + + @Override + protected void normalizeAndAppend( + CharSequence src, boolean doNormalize, NormalizerImpl.ReorderingBuffer buffer) { + impl.composeAndAppend(src, doNormalize, onlyContiguous, buffer); + } + + @Override + public boolean isNormalized(CharSequence s) { + // 5: small destCapacity for substring normalization + return impl.compose(s, 0, s.length(), + onlyContiguous, false, + new NormalizerImpl.ReorderingBuffer(impl, new StringBuilder(), 5)); + } + + @Override + public int spanQuickCheckYes(CharSequence s) { + return impl.composeQuickCheck(s, 0, s.length(), onlyContiguous, true)>>>1; + } + + @Override + public boolean hasBoundaryBefore(int c) { return impl.hasCompBoundaryBefore(c); } + + private final boolean onlyContiguous; + } + + // instance cache ---------------------------------------------------------- *** + + private Norm2AllModes(NormalizerImpl ni) { + impl=ni; + comp=new ComposeNormalizer2(ni, false); + decomp=new DecomposeNormalizer2(ni); + } + + public final NormalizerImpl impl; + public final ComposeNormalizer2 comp; + public final DecomposeNormalizer2 decomp; + + private static Norm2AllModes getInstanceFromSingleton(Norm2AllModesSingleton singleton) { + if(singleton.exception!=null) { + throw singleton.exception; + } + return singleton.allModes; + } + + public static Norm2AllModes getNFCInstance() { + return getInstanceFromSingleton(NFCSingleton.INSTANCE); + } + + public static Norm2AllModes getNFKCInstance() { + return getInstanceFromSingleton(NFKCSingleton.INSTANCE); + } + + public static final NoopNormalizer2 NOOP_NORMALIZER2=new NoopNormalizer2(); + + private static final class Norm2AllModesSingleton { + private Norm2AllModesSingleton(String name) { + try { + @SuppressWarnings("deprecation") + String DATA_FILE_NAME = "/jdk/internal/icu/impl/data/icudt" + + VersionInfo.ICU_DATA_VERSION_PATH + "/" + name + ".nrm"; + NormalizerImpl impl=new NormalizerImpl().load(DATA_FILE_NAME); + allModes=new Norm2AllModes(impl); + } catch (RuntimeException e) { + exception=e; + } + } + + private Norm2AllModes allModes; + private RuntimeException exception; + } + + private static final class NFCSingleton { + private static final Norm2AllModesSingleton INSTANCE=new Norm2AllModesSingleton("nfc"); + } + + private static final class NFKCSingleton { + private static final Norm2AllModesSingleton INSTANCE=new Norm2AllModesSingleton("nfkc"); + } +} --- old/src/java.base/share/classes/sun/text/normalizer/NormalizerImpl.java 2020-01-10 15:57:31.000000000 -0800 +++ /dev/null 2020-01-10 15:57:31.000000000 -0800 @@ -1,2188 +0,0 @@ -/* - * Copyright (c) 2009, 2019, Oracle and/or its affiliates. All rights reserved. - * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. - * - * This code is free software; you can redistribute it and/or modify it - * under the terms of the GNU General Public License version 2 only, as - * published by the Free Software Foundation. Oracle designates this - * particular file as subject to the "Classpath" exception as provided - * by Oracle in the LICENSE file that accompanied this code. - * - * This code is distributed in the hope that it will be useful, but WITHOUT - * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or - * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License - * version 2 for more details (a copy is included in the LICENSE file that - * accompanied this code). - * - * You should have received a copy of the GNU General Public License version - * 2 along with this work; if not, write to the Free Software Foundation, - * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. - * - * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA - * or visit www.oracle.com if you need additional information or have any - * questions. - */ - -/* - ******************************************************************************* - * Copyright (C) 2009-2014, International Business Machines - * Corporation and others. All Rights Reserved. - ******************************************************************************* - */ -package sun.text.normalizer; - -import java.io.IOException; -import java.nio.ByteBuffer; -import java.text.Normalizer; - -// Original filename in ICU4J: Normalizer2Impl.java -public final class NormalizerImpl { - public static final class Hangul { - /* Korean Hangul and Jamo constants */ - public static final int JAMO_L_BASE=0x1100; /* "lead" jamo */ - public static final int JAMO_V_BASE=0x1161; /* "vowel" jamo */ - public static final int JAMO_T_BASE=0x11a7; /* "trail" jamo */ - - public static final int HANGUL_BASE=0xac00; - public static final int HANGUL_END=0xd7a3; - - public static final int JAMO_L_COUNT=19; - public static final int JAMO_V_COUNT=21; - public static final int JAMO_T_COUNT=28; - - public static final int HANGUL_COUNT=JAMO_L_COUNT*JAMO_V_COUNT*JAMO_T_COUNT; - public static final int HANGUL_LIMIT=HANGUL_BASE+HANGUL_COUNT; - - public static boolean isHangul(int c) { - return HANGUL_BASE<=c && c - * If dest is a StringBuilder, then the buffer writes directly to it. - * Otherwise, the buffer maintains a StringBuilder for intermediate text segments - * until no further changes are necessary and whole segments are appended. - * append() methods that take combining-class values always write to the StringBuilder. - * Other append() methods flush and append to the Appendable. - */ - public static final class ReorderingBuffer implements Appendable { - public ReorderingBuffer(NormalizerImpl ni, Appendable dest, int destCapacity) { - impl=ni; - app=dest; - if (app instanceof StringBuilder) { - appIsStringBuilder=true; - str=(StringBuilder)dest; - // In Java, the constructor subsumes public void init(int destCapacity) - str.ensureCapacity(destCapacity); - reorderStart=0; - if(str.length()==0) { - lastCC=0; - } else { - setIterator(); - lastCC=previousCC(); - // Set reorderStart after the last code point with cc<=1 if there is one. - if(lastCC>1) { - while(previousCC()>1) {} - } - reorderStart=codePointLimit; - } - } else { - appIsStringBuilder=false; - str=new StringBuilder(); - reorderStart=0; - lastCC=0; - } - } - - public boolean isEmpty() { return str.length()==0; } - public int length() { return str.length(); } - public int getLastCC() { return lastCC; } - - public StringBuilder getStringBuilder() { return str; } - - public boolean equals(CharSequence s, int start, int limit) { - return UTF16Plus.equal(str, 0, str.length(), s, start, limit); - } - - public void append(int c, int cc) { - if(lastCC<=cc || cc==0) { - str.appendCodePoint(c); - lastCC=cc; - if(cc<=1) { - reorderStart=str.length(); - } - } else { - insert(c, cc); - } - } - public void append(CharSequence s, int start, int limit, boolean isNFD, - int leadCC, int trailCC) { - if(start==limit) { - return; - } - if(lastCC<=leadCC || leadCC==0) { - if(trailCC<=1) { - reorderStart=str.length()+(limit-start); - } else if(leadCC<=1) { - reorderStart=str.length()+1; // Ok if not a code point boundary. - } - str.append(s, start, limit); - lastCC=trailCC; - } else { - int c=Character.codePointAt(s, start); - start+=Character.charCount(c); - insert(c, leadCC); // insert first code point - while(startcc;) {} - // insert c at codePointLimit, after the character with prevCC<=cc - if(c<=0xffff) { - str.insert(codePointLimit, (char)c); - if(cc<=1) { - reorderStart=codePointLimit+1; - } - } else { - str.insert(codePointLimit, Character.toChars(c)); - if(cc<=1) { - reorderStart=codePointLimit+2; - } - } - } - - private final NormalizerImpl impl; - private final Appendable app; - private final StringBuilder str; - private final boolean appIsStringBuilder; - private int reorderStart; - private int lastCC; - - // private backward iterator - private void setIterator() { codePointStart=str.length(); } - private void skipPrevious() { // Requires 0=codePointStart) { - return 0; - } - int c=str.codePointBefore(codePointStart); - codePointStart-=Character.charCount(c); - return impl.getCCFromYesOrMaybeCP(c); - } - private int codePointStart, codePointLimit; - } - - // TODO: Propose as public API on the UTF16 class. - // TODO: Propose widening UTF16 methods that take char to take int. - // TODO: Propose widening UTF16 methods that take String to take CharSequence. - public static final class UTF16Plus { - /** - * Is this code point a lead surrogate (U+d800..U+dbff)? - * @param c code unit or code point - * @return true or false - */ - public static boolean isLeadSurrogate(int c) { return (c & 0xfffffc00) == 0xd800; } - /** - * Assuming c is a surrogate code point (UTF16.isSurrogate(c)), - * is it a lead surrogate? - * @param c code unit or code point - * @return true or false - */ - public static boolean isSurrogateLead(int c) { return (c&0x400)==0; } - - /** - * Compares two CharSequence subsequences for binary equality. - * @param s1 first sequence - * @param start1 start offset in first sequence - * @param limit1 limit offset in first sequence - * @param s2 second sequence - * @param start2 start offset in second sequence - * @param limit2 limit offset in second sequence - * @return true if s1.subSequence(start1, limit1) contains the same text - * as s2.subSequence(start2, limit2) - */ - public static boolean equal(CharSequence s1, int start1, int limit1, - CharSequence s2, int start2, int limit2) { - if((limit1-start1)!=(limit2-start2)) { - return false; - } - if(s1==s2 && start1==start2) { - return true; - } - while(start1>DELTA_SHIFT)-MAX_DELTA-1; - - // Read the normTrie. - int offset=inIndexes[IX_NORM_TRIE_OFFSET]; - int nextOffset=inIndexes[IX_EXTRA_DATA_OFFSET]; - int triePosition = bytes.position(); - normTrie = CodePointTrie.Fast16.fromBinary(bytes); - int trieLength = bytes.position() - triePosition; - if(trieLength>(nextOffset-offset)) { - throw new InternalError("Normalizer2 data: not enough bytes for normTrie"); - } - ICUBinary.skipBytes(bytes, (nextOffset-offset)-trieLength); // skip padding after trie bytes - - // Read the composition and mapping data. - offset=nextOffset; - nextOffset=inIndexes[IX_SMALL_FCD_OFFSET]; - int numChars=(nextOffset-offset)/2; - if(numChars!=0) { - maybeYesCompositions=ICUBinary.getString(bytes, numChars, 0); - extraData=maybeYesCompositions.substring((MIN_NORMAL_MAYBE_YES-minMaybeYes)>>OFFSET_SHIFT); - } - - // smallFCD: new in formatVersion 2 - offset=nextOffset; - smallFCD=new byte[0x100]; - bytes.get(smallFCD); - - return this; - } catch(IOException e) { - throw new InternalError(e); - } - } - public NormalizerImpl load(String name) { - return load(ICUBinary.getRequiredData(name)); - } - - // The trie stores values for lead surrogate code *units*. - // Surrogate code *points* are inert. - public int getNorm16(int c) { - return UTF16Plus.isLeadSurrogate(c) ? INERT : normTrie.get(c); - } - public int getRawNorm16(int c) { return normTrie.get(c); } - public boolean isAlgorithmicNoNo(int norm16) { return limitNoNo<=norm16 && norm16=MIN_NORMAL_MAYBE_YES) { - return getCCFromNormalYesOrMaybe(norm16); - } - if(norm16> OFFSET_SHIFT) & 0xff; - } - public static int getCCFromYesOrMaybe(int norm16) { - return norm16>=MIN_NORMAL_MAYBE_YES ? getCCFromNormalYesOrMaybe(norm16) : 0; - } - public int getCCFromYesOrMaybeCP(int c) { - if (c < minCompNoMaybeCP) { return 0; } - return getCCFromYesOrMaybe(getNorm16(c)); - } - - /** - * Returns the FCD data for code point c. - * @param c A Unicode code point. - * @return The lccc(c) in bits 15..8 and tccc(c) in bits 7..0. - */ - public int getFCD16(int c) { - if(c>8]; - if(bits==0) { return false; } - return ((bits>>((lead>>5)&7))&1)!=0; - } - - /** Gets the FCD value from the regular normalization data. */ - public int getFCD16FromNormData(int c) { - int norm16=getNorm16(c); - if (norm16 >= limitNoNo) { - if(norm16>=MIN_NORMAL_MAYBE_YES) { - // combining mark - norm16=getCCFromNormalYesOrMaybe(norm16); - return norm16|(norm16<<8); - } else if(norm16>=minMaybeYes) { - return 0; - } else { // isDecompNoAlgorithmic(norm16) - int deltaTrailCC = norm16 & DELTA_TCCC_MASK; - if (deltaTrailCC <= DELTA_TCCC_1) { - return deltaTrailCC >> OFFSET_SHIFT; - } - // Maps to an isCompYesAndZeroCC. - c=mapAlgorithmic(c, norm16); - norm16=getRawNorm16(c); - } - } - if(norm16<=minYesNo || isHangulLVT(norm16)) { - // no decomposition or Hangul syllable, all zeros - return 0; - } - // c decomposes, get everything from the variable-length extra data - int mapping=norm16>>OFFSET_SHIFT; - int firstUnit=extraData.charAt(mapping); - int fcd16=firstUnit>>8; // tccc - if((firstUnit&MAPPING_HAS_CCC_LCCC_WORD)!=0) { - fcd16|=extraData.charAt(mapping-1)&0xff00; // lccc - } - return fcd16; - } - - /** - * Gets the decomposition for one code point. - * @param c code point - * @return c's decomposition, if it has one; returns null if it does not have a decomposition - */ - public String getDecomposition(int c) { - int norm16; - if(c>OFFSET_SHIFT; - int length=extraData.charAt(mapping++)&MAPPING_LENGTH_MASK; - return extraData.substring(mapping, mapping+length); - } - - // Fixed norm16 values. - public static final int MIN_YES_YES_WITH_CC=0xfe02; - public static final int JAMO_VT=0xfe00; - public static final int MIN_NORMAL_MAYBE_YES=0xfc00; - public static final int JAMO_L=2; // offset=1 hasCompBoundaryAfter=FALSE - public static final int INERT=1; // offset=0 hasCompBoundaryAfter=TRUE - - // norm16 bit 0 is comp-boundary-after. - public static final int HAS_COMP_BOUNDARY_AFTER=1; - public static final int OFFSET_SHIFT=1; - - // For algorithmic one-way mappings, norm16 bits 2..1 indicate the - // tccc (0, 1, >1) for quick FCC boundary-after tests. - public static final int DELTA_TCCC_0=0; - public static final int DELTA_TCCC_1=2; - public static final int DELTA_TCCC_GT_1=4; - public static final int DELTA_TCCC_MASK=6; - public static final int DELTA_SHIFT=3; - - public static final int MAX_DELTA=0x40; - - // Byte offsets from the start of the data, after the generic header. - public static final int IX_NORM_TRIE_OFFSET=0; - public static final int IX_EXTRA_DATA_OFFSET=1; - public static final int IX_SMALL_FCD_OFFSET=2; - public static final int IX_RESERVED3_OFFSET=3; - public static final int IX_TOTAL_SIZE=7; - public static final int MIN_CCC_LCCC_CP=0x300; - // Code point thresholds for quick check codes. - public static final int IX_MIN_DECOMP_NO_CP=8; - public static final int IX_MIN_COMP_NO_MAYBE_CP=9; - - // Norm16 value thresholds for quick check combinations and types of extra data. - - /** Mappings & compositions in [minYesNo..minYesNoMappingsOnly[. */ - public static final int IX_MIN_YES_NO=10; - /** Mappings are comp-normalized. */ - public static final int IX_MIN_NO_NO=11; - public static final int IX_LIMIT_NO_NO=12; - public static final int IX_MIN_MAYBE_YES=13; - - /** Mappings only in [minYesNoMappingsOnly..minNoNo[. */ - public static final int IX_MIN_YES_NO_MAPPINGS_ONLY=14; - /** Mappings are not comp-normalized but have a comp boundary before. */ - public static final int IX_MIN_NO_NO_COMP_BOUNDARY_BEFORE=15; - /** Mappings do not have a comp boundary before. */ - public static final int IX_MIN_NO_NO_COMP_NO_MAYBE_CC=16; - /** Mappings to the empty string. */ - public static final int IX_MIN_NO_NO_EMPTY=17; - - public static final int IX_MIN_LCCC_CP=18; - public static final int IX_COUNT=20; - - public static final int MAPPING_HAS_CCC_LCCC_WORD=0x80; - public static final int MAPPING_HAS_RAW_MAPPING=0x40; - // unused bit 0x20; - public static final int MAPPING_LENGTH_MASK=0x1f; - - public static final int COMP_1_LAST_TUPLE=0x8000; - public static final int COMP_1_TRIPLE=1; - public static final int COMP_1_TRAIL_LIMIT=0x3400; - public static final int COMP_1_TRAIL_MASK=0x7ffe; - public static final int COMP_1_TRAIL_SHIFT=9; // 10-1 for the "triple" bit - public static final int COMP_2_TRAIL_SHIFT=6; - public static final int COMP_2_TRAIL_MASK=0xffc0; - - // higher-level functionality ------------------------------------------ *** - - /** - * Decomposes s[src, limit[ and writes the result to dest. - * limit can be NULL if src is NUL-terminated. - * destLengthEstimate is the initial dest buffer capacity and can be -1. - */ - public void decompose(CharSequence s, int src, int limit, StringBuilder dest, - int destLengthEstimate) { - if(destLengthEstimate<0) { - destLengthEstimate=limit-src; - } - dest.setLength(0); - ReorderingBuffer buffer=new ReorderingBuffer(this, dest, destLengthEstimate); - decompose(s, src, limit, buffer); - } - - // Dual functionality: - // buffer!=NULL: normalize - // buffer==NULL: isNormalized/quickCheck/spanQuickCheckYes - public int decompose(CharSequence s, int src, int limit, - ReorderingBuffer buffer) { - int minNoCP=minDecompNoCP; - - int prevSrc; - int c=0; - int norm16=0; - - // only for quick check - int prevBoundary=src; - int prevCC=0; - - for(;;) { - // count code units below the minimum or with irrelevant data for the quick check - for(prevSrc=src; src!=limit;) { - if( (c=s.charAt(src))=limit) { - break; - } - c=Character.codePointAt(s, src); - cc=getCC(getNorm16(c)); - }; - buffer.append(s, 0, src, false, firstCC, prevCC); - buffer.append(s, src, limit); - } - - // Very similar to composeQuickCheck(): Make the same changes in both places if relevant. - // doCompose: normalize - // !doCompose: isNormalized (buffer must be empty and initialized) - public boolean compose(CharSequence s, int src, int limit, - boolean onlyContiguous, - boolean doCompose, - ReorderingBuffer buffer) { - int prevBoundary=src; - int minNoMaybeCP=minCompNoMaybeCP; - - for (;;) { - // Fast path: Scan over a sequence of characters below the minimum "no or maybe" code point, - // or with (compYes && ccc==0) properties. - int prevSrc; - int c = 0; - int norm16 = 0; - for (;;) { - if (src == limit) { - if (prevBoundary != limit && doCompose) { - buffer.append(s, prevBoundary, limit); - } - return true; - } - if( (c=s.charAt(src))=minNoNo. - // The current character is either a "noNo" (has a mapping) - // or a "maybeYes" (combines backward) - // or a "yesYes" with ccc!=0. - // It is not a Hangul syllable or Jamo L because those have "yes" properties. - - // Medium-fast path: Handle cases that do not require full decomposition and recomposition. - if (!isMaybeOrNonZeroCC(norm16)) { // minNoNo <= norm16 < minMaybeYes - if (!doCompose) { - return false; - } - // Fast path for mapping a character that is immediately surrounded by boundaries. - // In this case, we need not decompose around the current character. - if (isDecompNoAlgorithmic(norm16)) { - // Maps to a single isCompYesAndZeroCC character - // which also implies hasCompBoundaryBefore. - if (norm16HasCompBoundaryAfter(norm16, onlyContiguous) || - hasCompBoundaryBefore(s, src, limit)) { - if (prevBoundary != prevSrc) { - buffer.append(s, prevBoundary, prevSrc); - } - buffer.append(mapAlgorithmic(c, norm16), 0); - prevBoundary = src; - continue; - } - } else if (norm16 < minNoNoCompBoundaryBefore) { - // The mapping is comp-normalized which also implies hasCompBoundaryBefore. - if (norm16HasCompBoundaryAfter(norm16, onlyContiguous) || - hasCompBoundaryBefore(s, src, limit)) { - if (prevBoundary != prevSrc) { - buffer.append(s, prevBoundary, prevSrc); - } - int mapping = norm16 >> OFFSET_SHIFT; - int length = extraData.charAt(mapping++) & MAPPING_LENGTH_MASK; - buffer.append(extraData, mapping, mapping + length); - prevBoundary = src; - continue; - } - } else if (norm16 >= minNoNoEmpty) { - // The current character maps to nothing. - // Simply omit it from the output if there is a boundary before _or_ after it. - // The character itself implies no boundaries. - if (hasCompBoundaryBefore(s, src, limit) || - hasCompBoundaryAfter(s, prevBoundary, prevSrc, onlyContiguous)) { - if (prevBoundary != prevSrc) { - buffer.append(s, prevBoundary, prevSrc); - } - prevBoundary = src; - continue; - } - } - // Other "noNo" type, or need to examine more text around this character: - // Fall through to the slow path. - } else if (isJamoVT(norm16) && prevBoundary != prevSrc) { - char prev=s.charAt(prevSrc-1); - if(c= 0) { - int syllable = Hangul.HANGUL_BASE + - (l*Hangul.JAMO_V_COUNT + (c-Hangul.JAMO_V_BASE)) * - Hangul.JAMO_T_COUNT + t; - --prevSrc; // Replace the Jamo L as well. - if (prevBoundary != prevSrc) { - buffer.append(s, prevBoundary, prevSrc); - } - buffer.append((char)syllable); - prevBoundary = src; - continue; - } - // If we see L+V+x where x!=T then we drop to the slow path, - // decompose and recompose. - // This is to deal with NFKC finding normal L and V but a - // compatibility variant of a T. - // We need to either fully compose that combination here - // (which would complicate the code and may not work with strange custom data) - // or use the slow path. - } - } else if (Hangul.isHangulLV(prev)) { - // The current character is a Jamo Trailing consonant, - // compose with previous Hangul LV that does not contain a Jamo T. - if (!doCompose) { - return false; - } - int syllable = prev + c - Hangul.JAMO_T_BASE; - --prevSrc; // Replace the Hangul LV as well. - if (prevBoundary != prevSrc) { - buffer.append(s, prevBoundary, prevSrc); - } - buffer.append((char)syllable); - prevBoundary = src; - continue; - } - // No matching context, or may need to decompose surrounding text first: - // Fall through to the slow path. - } else if (norm16 > JAMO_VT) { // norm16 >= MIN_YES_YES_WITH_CC - // One or more combining marks that do not combine-back: - // Check for canonical order, copy unchanged if ok and - // if followed by a character with a boundary-before. - int cc = getCCFromNormalYesOrMaybe(norm16); // cc!=0 - if (onlyContiguous /* FCC */ && getPreviousTrailCC(s, prevBoundary, prevSrc) > cc) { - // Fails FCD test, need to decompose and contiguously recompose. - if (!doCompose) { - return false; - } - } else { - // If !onlyContiguous (not FCC), then we ignore the tccc of - // the previous character which passed the quick check "yes && ccc==0" test. - int n16; - for (;;) { - if (src == limit) { - if (doCompose) { - buffer.append(s, prevBoundary, limit); - } - return true; - } - int prevCC = cc; - c = Character.codePointAt(s, src); - n16 = normTrie.get(c); - if (n16 >= MIN_YES_YES_WITH_CC) { - cc = getCCFromNormalYesOrMaybe(n16); - if (prevCC > cc) { - if (!doCompose) { - return false; - } - break; - } - } else { - break; - } - src += Character.charCount(c); - } - // p is after the last in-order combining mark. - // If there is a boundary here, then we continue with no change. - if (norm16HasCompBoundaryBefore(n16)) { - if (isCompYesAndZeroCC(n16)) { - src += Character.charCount(c); - } - continue; - } - // Use the slow path. There is no boundary in [prevSrc, src[. - } - } - - // Slow path: Find the nearest boundaries around the current character, - // decompose and recompose. - if (prevBoundary != prevSrc && !norm16HasCompBoundaryBefore(norm16)) { - c = Character.codePointBefore(s, prevSrc); - norm16 = normTrie.get(c); - if (!norm16HasCompBoundaryAfter(norm16, onlyContiguous)) { - prevSrc -= Character.charCount(c); - } - } - if (doCompose && prevBoundary != prevSrc) { - buffer.append(s, prevBoundary, prevSrc); - } - int recomposeStartIndex=buffer.length(); - // We know there is not a boundary here. - decomposeShort(s, prevSrc, src, false /* !stopAtCompBoundary */, onlyContiguous, - buffer); - // Decompose until the next boundary. - src = decomposeShort(s, src, limit, true /* stopAtCompBoundary */, onlyContiguous, - buffer); - recompose(buffer, recomposeStartIndex, onlyContiguous); - if(!doCompose) { - if(!buffer.equals(s, prevSrc, src)) { - return false; - } - buffer.remove(); - } - prevBoundary=src; - } - } - - /** - * Very similar to compose(): Make the same changes in both places if relevant. - * doSpan: spanQuickCheckYes (ignore bit 0 of the return value) - * !doSpan: quickCheck - * @return bits 31..1: spanQuickCheckYes (==s.length() if "yes") and - * bit 0: set if "maybe"; otherwise, if the span length<s.length() - * then the quick check result is "no" - */ - public int composeQuickCheck(CharSequence s, int src, int limit, - boolean onlyContiguous, boolean doSpan) { - int qcResult=0; - int prevBoundary=src; - int minNoMaybeCP=minCompNoMaybeCP; - - for(;;) { - // Fast path: Scan over a sequence of characters below the minimum "no or maybe" code point, - // or with (compYes && ccc==0) properties. - int prevSrc; - int c = 0; - int norm16 = 0; - for (;;) { - if(src==limit) { - return (src<<1)|qcResult; // "yes" or "maybe" - } - if( (c=s.charAt(src))=minNoNo. - // The current character is either a "noNo" (has a mapping) - // or a "maybeYes" (combines backward) - // or a "yesYes" with ccc!=0. - // It is not a Hangul syllable or Jamo L because those have "yes" properties. - - int prevNorm16 = INERT; - if (prevBoundary != prevSrc) { - prevBoundary = prevSrc; - if (!norm16HasCompBoundaryBefore(norm16)) { - c = Character.codePointBefore(s, prevSrc); - int n16 = getNorm16(c); - if (!norm16HasCompBoundaryAfter(n16, onlyContiguous)) { - prevBoundary -= Character.charCount(c); - prevNorm16 = n16; - } - } - } - - if(isMaybeOrNonZeroCC(norm16)) { - int cc=getCCFromYesOrMaybe(norm16); - if (onlyContiguous /* FCC */ && cc != 0 && - getTrailCCFromCompYesAndZeroCC(prevNorm16) > cc) { - // The [prevBoundary..prevSrc[ character - // passed the quick check "yes && ccc==0" test - // but is out of canonical order with the current combining mark. - } else { - // If !onlyContiguous (not FCC), then we ignore the tccc of - // the previous character which passed the quick check "yes && ccc==0" test. - for (;;) { - if (norm16 < MIN_YES_YES_WITH_CC) { - if (!doSpan) { - qcResult = 1; - } else { - return prevBoundary << 1; // spanYes does not care to know it's "maybe" - } - } - if (src == limit) { - return (src<<1) | qcResult; // "yes" or "maybe" - } - int prevCC = cc; - c = Character.codePointAt(s, src); - norm16 = getNorm16(c); - if (isMaybeOrNonZeroCC(norm16)) { - cc = getCCFromYesOrMaybe(norm16); - if (!(prevCC <= cc || cc == 0)) { - break; - } - } else { - break; - } - src += Character.charCount(c); - } - // src is after the last in-order combining mark. - if (isCompYesAndZeroCC(norm16)) { - prevBoundary = src; - src += Character.charCount(c); - continue; - } - } - } - return prevBoundary<<1; // "no" - } - } - public void composeAndAppend(CharSequence s, - boolean doCompose, - boolean onlyContiguous, - ReorderingBuffer buffer) { - int src=0, limit=s.length(); - if(!buffer.isEmpty()) { - int firstStarterInSrc=findNextCompBoundary(s, 0, limit, onlyContiguous); - if(0!=firstStarterInSrc) { - int lastStarterInDest=findPreviousCompBoundary(buffer.getStringBuilder(), - buffer.length(), onlyContiguous); - StringBuilder middle=new StringBuilder((buffer.length()-lastStarterInDest)+ - firstStarterInSrc+16); - middle.append(buffer.getStringBuilder(), lastStarterInDest, buffer.length()); - buffer.removeSuffix(buffer.length()-lastStarterInDest); - middle.append(s, 0, firstStarterInSrc); - compose(middle, 0, middle.length(), onlyContiguous, true, buffer); - src=firstStarterInSrc; - } - } - if(doCompose) { - compose(s, src, limit, onlyContiguous, true, buffer); - } else { - buffer.append(s, src, limit); - } - } - // Dual functionality: - // buffer!=NULL: normalize - // buffer==NULL: isNormalized/quickCheck/spanQuickCheckYes - public int makeFCD(CharSequence s, int src, int limit, ReorderingBuffer buffer) { - // Note: In this function we use buffer->appendZeroCC() because we track - // the lead and trail combining classes here, rather than leaving it to - // the ReorderingBuffer. - // The exception is the call to decomposeShort() which uses the buffer - // in the normal way. - - // Tracks the last FCD-safe boundary, before lccc=0 or after properly-ordered tccc<=1. - // Similar to the prevBoundary in the compose() implementation. - int prevBoundary=src; - int prevSrc; - int c=0; - int prevFCD16=0; - int fcd16=0; - - for(;;) { - // count code units with lccc==0 - for(prevSrc=src; src!=limit;) { - if((c=s.charAt(src))1) { - --prevBoundary; - } - } - } else { - int p=src-1; - if( Character.isLowSurrogate(s.charAt(p)) && prevSrc

1) { - prevBoundary=p; - } - } - if(buffer!=null) { - // The last lccc==0 character is excluded from the - // flush-and-append call in case it needs to be modified. - buffer.flushAndAppendZeroCC(s, prevSrc, prevBoundary); - buffer.append(s, prevBoundary, src); - } - // The start of the current character (c). - prevSrc=src; - } else if(src==limit) { - break; - } - - src+=Character.charCount(c); - // The current character (c) at [prevSrc..src[ has a non-zero lead combining class. - // Check for proper order, and decompose locally if necessary. - if((prevFCD16&0xff)<=(fcd16>>8)) { - // proper order: prev tccc <= current lccc - if((fcd16&0xff)<=1) { - prevBoundary=src; - } - if(buffer!=null) { - buffer.appendZeroCC(c); - } - prevFCD16=fcd16; - continue; - } else if(buffer==null) { - return prevBoundary; // quick check "no" - } else { - /* - * Back out the part of the source that we copied or appended - * already but is now going to be decomposed. - * prevSrc is set to after what was copied/appended. - */ - buffer.removeSuffix(prevSrc-prevBoundary); - /* - * Find the part of the source that needs to be decomposed, - * up to the next safe boundary. - */ - src=findNextFCDBoundary(s, src, limit); - /* - * The source text does not fulfill the conditions for FCD. - * Decompose and reorder a limited piece of the text. - */ - decomposeShort(s, prevBoundary, src, false, false, buffer); - prevBoundary=src; - prevFCD16=0; - } - } - return src; - } - - public boolean hasDecompBoundaryBefore(int c) { - return c < minLcccCP || (c <= 0xffff && !singleLeadMightHaveNonZeroFCD16(c)) || - norm16HasDecompBoundaryBefore(getNorm16(c)); - } - public boolean norm16HasDecompBoundaryBefore(int norm16) { - if (norm16 < minNoNoCompNoMaybeCC) { - return true; - } - if (norm16 >= limitNoNo) { - return norm16 <= MIN_NORMAL_MAYBE_YES || norm16 == JAMO_VT; - } - // c decomposes, get everything from the variable-length extra data - int mapping=norm16>>OFFSET_SHIFT; - int firstUnit=extraData.charAt(mapping); - // true if leadCC==0 (hasFCDBoundaryBefore()) - return (firstUnit&MAPPING_HAS_CCC_LCCC_WORD)==0 || (extraData.charAt(mapping-1)&0xff00)==0; - } - public boolean hasDecompBoundaryAfter(int c) { - if (c < minDecompNoCP) { - return true; - } - if (c <= 0xffff && !singleLeadMightHaveNonZeroFCD16(c)) { - return true; - } - return norm16HasDecompBoundaryAfter(getNorm16(c)); - } - public boolean norm16HasDecompBoundaryAfter(int norm16) { - if(norm16 <= minYesNo || isHangulLVT(norm16)) { - return true; - } - if (norm16 >= limitNoNo) { - if (isMaybeOrNonZeroCC(norm16)) { - return norm16 <= MIN_NORMAL_MAYBE_YES || norm16 == JAMO_VT; - } - // Maps to an isCompYesAndZeroCC. - return (norm16 & DELTA_TCCC_MASK) <= DELTA_TCCC_1; - } - // c decomposes, get everything from the variable-length extra data - int mapping=norm16>>OFFSET_SHIFT; - int firstUnit=extraData.charAt(mapping); - // decomp after-boundary: same as hasFCDBoundaryAfter(), - // fcd16<=1 || trailCC==0 - if(firstUnit>0x1ff) { - return false; // trailCC>1 - } - if(firstUnit<=0xff) { - return true; // trailCC==0 - } - // if(trailCC==1) test leadCC==0, same as checking for before-boundary - // true if leadCC==0 (hasFCDBoundaryBefore()) - return (firstUnit&MAPPING_HAS_CCC_LCCC_WORD)==0 || (extraData.charAt(mapping-1)&0xff00)==0; - } - public boolean isDecompInert(int c) { return isDecompYesAndZeroCC(getNorm16(c)); } - - public boolean hasCompBoundaryBefore(int c) { - return c=minMaybeYes; } - private static boolean isInert(int norm16) { return norm16==INERT; } - private static boolean isJamoVT(int norm16) { return norm16==JAMO_VT; } - private int hangulLVT() { return minYesNoMappingsOnly|HAS_COMP_BOUNDARY_AFTER; } - private boolean isHangulLV(int norm16) { return norm16==minYesNo; } - private boolean isHangulLVT(int norm16) { - return norm16==hangulLVT(); - } - private boolean isCompYesAndZeroCC(int norm16) { return norm16=MIN_YES_YES_WITH_CC || norm16=limitNoNo; } - - // For use with isCompYes(). - // Perhaps the compiler can combine the two tests for MIN_YES_YES_WITH_CC. - // static uint8_t getCCFromYes(uint16_t norm16) { - // return norm16>=MIN_YES_YES_WITH_CC ? getCCFromNormalYesOrMaybe(norm16) : 0; - // } - private int getCCFromNoNo(int norm16) { - int mapping=norm16>>OFFSET_SHIFT; - if((extraData.charAt(mapping)&MAPPING_HAS_CCC_LCCC_WORD)!=0) { - return extraData.charAt(mapping-1)&0xff; - } else { - return 0; - } - } - int getTrailCCFromCompYesAndZeroCC(int norm16) { - if(norm16<=minYesNo) { - return 0; // yesYes and Hangul LV have ccc=tccc=0 - } else { - // For Hangul LVT we harmlessly fetch a firstUnit with tccc=0 here. - return extraData.charAt(norm16>>OFFSET_SHIFT)>>8; // tccc from yesNo - } - } - - // Requires algorithmic-NoNo. - private int mapAlgorithmic(int c, int norm16) { - return c+(norm16>>DELTA_SHIFT)-centerNoNoDelta; - } - - // Requires minYesNo>OFFSET_SHIFT); } - - /** - * @return index into maybeYesCompositions, or -1 - */ - private int getCompositionsListForDecompYes(int norm16) { - if(norm16>OFFSET_SHIFT; - } - } - /** - * @return index into maybeYesCompositions - */ - private int getCompositionsListForComposite(int norm16) { - // A composite has both mapping & compositions list. - int list=((MIN_NORMAL_MAYBE_YES-minMaybeYes)+norm16)>>OFFSET_SHIFT; - int firstUnit=maybeYesCompositions.charAt(list); - return list+ // mapping in maybeYesCompositions - 1+ // +1 to skip the first unit with the mapping length - (firstUnit&MAPPING_LENGTH_MASK); // + mapping length - } - - // Decompose a short piece of text which is likely to contain characters that - // fail the quick check loop and/or where the quick check loop's overhead - // is unlikely to be amortized. - // Called by the compose() and makeFCD() implementations. - // Public in Java for collation implementation code. - private int decomposeShort( - CharSequence s, int src, int limit, - boolean stopAtCompBoundary, boolean onlyContiguous, - ReorderingBuffer buffer) { - while(src= limitNoNo) { - if (isMaybeOrNonZeroCC(norm16)) { - buffer.append(c, getCCFromYesOrMaybe(norm16)); - return; - } - // Maps to an isCompYesAndZeroCC. - c=mapAlgorithmic(c, norm16); - norm16=getRawNorm16(c); - } - if (norm16 < minYesNo) { - // c does not decompose - buffer.append(c, 0); - } else if(isHangulLV(norm16) || isHangulLVT(norm16)) { - // Hangul syllable: decompose algorithmically - Hangul.decompose(c, buffer); - } else { - // c decomposes, get everything from the variable-length extra data - int mapping=norm16>>OFFSET_SHIFT; - int firstUnit=extraData.charAt(mapping); - int length=firstUnit&MAPPING_LENGTH_MASK; - int leadCC, trailCC; - trailCC=firstUnit>>8; - if((firstUnit&MAPPING_HAS_CCC_LCCC_WORD)!=0) { - leadCC=extraData.charAt(mapping-1)>>8; - } else { - leadCC=0; - } - ++mapping; // skip over the firstUnit - buffer.append(extraData, mapping, mapping+length, true, leadCC, trailCC); - } - } - - /** - * Finds the recomposition result for - * a forward-combining "lead" character, - * specified with a pointer to its compositions list, - * and a backward-combining "trail" character. - * - *

If the lead and trail characters combine, then this function returns - * the following "compositeAndFwd" value: - * 

-     * Bits 21..1  composite character
-     * Bit      0  set if the composite is a forward-combining starter
-     *
- * otherwise it returns -1. - * - *

The compositions list has (trail, compositeAndFwd) pair entries, - * encoded as either pairs or triples of 16-bit units. - * The last entry has the high bit of its first unit set. - * - *

The list is sorted by ascending trail characters (there are no duplicates). - * A linear search is used. - * - *

See normalizer2impl.h for a more detailed description - * of the compositions list format. - */ - private static int combine(String compositions, int list, int trail) { - int key1, firstUnit; - if(trail(firstUnit=compositions.charAt(list))) { - list+=2+(firstUnit&COMP_1_TRIPLE); - } - if(key1==(firstUnit&COMP_1_TRAIL_MASK)) { - if((firstUnit&COMP_1_TRIPLE)!=0) { - return (compositions.charAt(list+1)<<16)|compositions.charAt(list+2); - } else { - return compositions.charAt(list+1); - } - } - } else { - // trail character is 3400..10FFFF - // result entry has 3 units - key1=COMP_1_TRAIL_LIMIT+(((trail>>COMP_1_TRAIL_SHIFT))&~COMP_1_TRIPLE); - int key2=(trail<(firstUnit=compositions.charAt(list))) { - list+=2+(firstUnit&COMP_1_TRIPLE); - } else if(key1==(firstUnit&COMP_1_TRAIL_MASK)) { - if(key2>(secondUnit=compositions.charAt(list+1))) { - if((firstUnit&COMP_1_LAST_TUPLE)!=0) { - break; - } else { - list+=3; - } - } else if(key2==(secondUnit&COMP_2_TRAIL_MASK)) { - return ((secondUnit&~COMP_2_TRAIL_MASK)<<16)|compositions.charAt(list+2); - } else { - break; - } - } else { - break; - } - } - } - return -1; - } - - /* - * Recomposes the buffer text starting at recomposeStartIndex - * (which is in NFD - decomposed and canonically ordered), - * and truncates the buffer contents. - * - * Note that recomposition never lengthens the text: - * Any character consists of either one or two code units; - * a composition may contain at most one more code unit than the original starter, - * while the combining mark that is removed has at least one code unit. - */ - private void recompose(ReorderingBuffer buffer, int recomposeStartIndex, - boolean onlyContiguous) { - StringBuilder sb=buffer.getStringBuilder(); - int p=recomposeStartIndex; - if(p==sb.length()) { - return; - } - - int starter, pRemove; - int compositionsList; - int c, compositeAndFwd; - int norm16; - int cc, prevCC; - boolean starterIsSupplementary; - - // Some of the following variables are not used until we have a forward-combining starter - // and are only initialized now to avoid compiler warnings. - compositionsList=-1; // used as indicator for whether we have a forward-combining starter - starter=-1; - starterIsSupplementary=false; - prevCC=0; - - for(;;) { - c=sb.codePointAt(p); - p+=Character.charCount(c); - norm16=getNorm16(c); - cc=getCCFromYesOrMaybe(norm16); - if( // this character combines backward and - isMaybe(norm16) && - // we have seen a starter that combines forward and - compositionsList>=0 && - // the backward-combining character is not blocked - (prevCC=0) { - // The starter and the combining mark (c) do combine. - int composite=compositeAndFwd>>1; - - // Remove the combining mark. - pRemove=p-Character.charCount(c); // pRemove & p: start & limit of the combining mark - sb.delete(pRemove, p); - p=pRemove; - // Replace the starter with the composite. - if(starterIsSupplementary) { - if(composite>0xffff) { - // both are supplementary - sb.setCharAt(starter, UTF16.getLeadSurrogate(composite)); - sb.setCharAt(starter+1, UTF16.getTrailSurrogate(composite)); - } else { - sb.setCharAt(starter, (char)c); - sb.deleteCharAt(starter+1); - // The composite is shorter than the starter, - // move the intermediate characters forward one. - starterIsSupplementary=false; - --p; - } - } else if(composite>0xffff) { - // The composite is longer than the starter, - // move the intermediate characters back one. - starterIsSupplementary=true; - sb.setCharAt(starter, UTF16.getLeadSurrogate(composite)); - sb.insert(starter+1, UTF16.getTrailSurrogate(composite)); - ++p; - } else { - // both are on the BMP - sb.setCharAt(starter, (char)composite); - } - - // Keep prevCC because we removed the combining mark. - - if(p==sb.length()) { - break; - } - // Is the composite a starter that combines forward? - if((compositeAndFwd&1)!=0) { - compositionsList= - getCompositionsListForComposite(getRawNorm16(composite)); - } else { - compositionsList=-1; - } - - // We combined; continue with looking for compositions. - continue; - } - } - - // no combination this time - prevCC=cc; - if(p==sb.length()) { - break; - } - - // If c did not combine, then check if it is a starter. - if(cc==0) { - // Found a new starter. - if((compositionsList=getCompositionsListForDecompYes(norm16))>=0) { - // It may combine with something, prepare for it. - if(c<=0xffff) { - starterIsSupplementary=false; - starter=p-1; - } else { - starterIsSupplementary=true; - starter=p-2; - } - } - } else if(onlyContiguous) { - // FCC: no discontiguous compositions; any intervening character blocks. - compositionsList=-1; - } - } - buffer.flush(); - } - - /** - * Does c have a composition boundary before it? - * True if its decomposition begins with a character that has - * ccc=0 && NFC_QC=Yes (isCompYesAndZeroCC()). - * As a shortcut, this is true if c itself has ccc=0 && NFC_QC=Yes - * (isCompYesAndZeroCC()) so we need not decompose. - */ - private boolean hasCompBoundaryBefore(int c, int norm16) { - return c> OFFSET_SHIFT) <= 0x1ff); - } - - private int findPreviousCompBoundary(CharSequence s, int p, boolean onlyContiguous) { - while(p>0) { - int c=Character.codePointBefore(s, p); - int norm16 = getNorm16(c); - if (norm16HasCompBoundaryAfter(norm16, onlyContiguous)) { - break; - } - p-=Character.charCount(c); - if(hasCompBoundaryBefore(c, norm16)) { - break; - } - } - return p; - } - private int findNextCompBoundary(CharSequence s, int p, int limit, boolean onlyContiguous) { - while(p= 0x0009 && c <= 0x000D) || - (c >= 0x0020 && c <= 0x002F) || - (c >= 0x003A && c <= 0x0040) || - (c >= 0x005B && c <= 0x0060) || - (c >= 0x007B && c <= 0x007E); - } - - public static String canonicalDecomposeWithSingleQuotation(String string) { - Normalizer2 impl = Normalizer2.getNFDInstance(); - char[] src = string.toCharArray(); - int srcIndex = 0; - int srcLimit = src.length; - char[] dest = new char[src.length * 3]; //MAX_BUF_SIZE_DECOMPOSE = 3 - int destIndex = 0; - int destLimit = dest.length; - - int prevSrc; - String norm; - int reorderStartIndex, length; - char c1, c2; - int cp; - int minNoMaybe = 0x00c0; - int cc, prevCC, trailCC; - char[] p; - int pStart; - - // initialize - reorderStartIndex = 0; - prevCC = 0; - norm = null; - cp = 0; - pStart = 0; - - cc = trailCC = -1; // initialize to bogus value - c1 = 0; - for (;;) { - prevSrc=srcIndex; - //quick check (1)less than minNoMaybe (2)no decomp (3)hangual - while (srcIndex != srcLimit && - ((c1 = src[srcIndex]) < minNoMaybe || - (norm = impl.getDecomposition(cp = string.codePointAt(srcIndex))) == null || - (c1 >= '\uac00' && c1 <= '\ud7a3'))) { // Hangul Syllables - prevCC = 0; - srcIndex += (cp < 0x10000) ? 1 : 2; - } - - // copy these code units all at once - if (srcIndex != prevSrc) { - length = srcIndex - prevSrc; - if ((destIndex + length) <= destLimit) { - System.arraycopy(src,prevSrc,dest,destIndex,length); - } - - destIndex += length; - reorderStartIndex = destIndex; - } - - // end of source reached? - if (srcIndex == srcLimit) { - break; - } - - // cp already contains *src and norm32 is set for it, increment src - srcIndex += (cp < 0x10000) ? 1 : 2; - - if (cp < Character.MIN_SUPPLEMENTARY_CODE_POINT) { - c2 = 0; - length = 1; - - if (Character.isHighSurrogate(c1) - || Character.isLowSurrogate(c1)) { - norm = null; - } - } else { - length = 2; - c2 = src[srcIndex-1]; - } - - // get the decomposition and the lead and trail cc's - if (norm == null) { - // cp does not decompose - cc = trailCC = UCharacter.getCombiningClass(cp); - p = null; - pStart = -1; - } else { - - pStart = 0; - p = norm.toCharArray(); - length = p.length; - int cpNum = norm.codePointCount(0, length); - cc= UCharacter.getCombiningClass(norm.codePointAt(0)); - trailCC= UCharacter.getCombiningClass(norm.codePointAt(cpNum-1)); - if (length == 1) { - // fastpath a single code unit from decomposition - c1 = p[pStart]; - c2 = 0; - p = null; - pStart = -1; - } - } - - if((destIndex + length * 3) >= destLimit) { // 2 SingleQuotations - // buffer overflow - char[] tmpBuf = new char[destLimit * 2]; - System.arraycopy(dest, 0, tmpBuf, 0, destIndex); - dest = tmpBuf; - destLimit = dest.length; - } - - // append the decomposition to the destination buffer, assume length>0 - { - int reorderSplit = destIndex; - if (p == null) { - // fastpath: single code point - if (needSingleQuotation(c1)) { - //if we need single quotation, no need to consider "prevCC" - //and it must NOT be a supplementary pair - dest[destIndex++] = '\''; - dest[destIndex++] = c1; - dest[destIndex++] = '\''; - trailCC = 0; - } else if(cc != 0 && cc < prevCC) { - // (c1, c2) is out of order with respect to the preceding - // text - destIndex += length; - trailCC = insertOrdered(dest, reorderStartIndex, - reorderSplit, destIndex, c1, c2, cc); - } else { - // just append (c1, c2) - dest[destIndex++] = c1; - if(c2 != 0) { - dest[destIndex++] = c2; - } - } - } else { - // general: multiple code points (ordered by themselves) - // from decomposition - if (needSingleQuotation(p[pStart])) { - dest[destIndex++] = '\''; - dest[destIndex++] = p[pStart++]; - dest[destIndex++] = '\''; - length--; - do { - dest[destIndex++] = p[pStart++]; - } while(--length > 0); - } else if (cc != 0 && cc < prevCC) { - destIndex += length; - trailCC = mergeOrdered(dest, reorderStartIndex, - reorderSplit, p, pStart, - pStart+length); - } else { - // just append the decomposition - do { - dest[destIndex++] = p[pStart++]; - } while (--length > 0); - } - } - } - prevCC = trailCC; - if(prevCC == 0) { - reorderStartIndex = destIndex; - } - } - - return new String(dest, 0, destIndex); - } - - /** - * simpler, single-character version of mergeOrdered() - - * bubble-insert one single code point into the preceding string - * which is already canonically ordered - * (c, c2) may or may not yet have been inserted at src[current]..src[p] - * - * it must be p=current+lengthof(c, c2) i.e. p=current+(c2==0 ? 1 : 2) - * - * before: src[start]..src[current] is already ordered, and - * src[current]..src[p] may or may not hold (c, c2) but - * must be exactly the same length as (c, c2) - * after: src[start]..src[p] is ordered - * - * @return the trailing combining class - */ - private static int/*unsigned byte*/ insertOrdered(char[] source, - int start, - int current, int p, - char c1, char c2, - int/*unsigned byte*/ cc) { - int back, preBack; - int r; - int prevCC, trailCC=cc; - - if (start=prevCC - preBack=back=current; - - PrevArgs prevArgs = new PrevArgs(); - prevArgs.current = current; - prevArgs.start = start; - prevArgs.src = source; - prevArgs.c1 = c1; - prevArgs.c2 = c2; - - // get the prevCC - prevCC=getPrevCC(prevArgs); - preBack = prevArgs.current; - - if(cc=prevCC) { - break; - } - back=preBack; - } - - // this is where we are right now with all these indicies: - // [start]..[pPreBack] 0..? code points that we can ignore - // [pPreBack]..[pBack] 0..1 code points with prevCC<=cc - // [pBack]..[current] 0..n code points with >cc, move up to insert (c, c2) - // [current]..[p] 1 code point (c, c2) with cc - - // move the code units in between up - r=p; - do { - source[--r]=source[--current]; - } while (back!=current); - } - } - - // insert (c1, c2) - source[current] = c1; - if (c2!=0) { - source[(current+1)] = c2; - } - - // we know the cc of the last code point - return trailCC; - } - /** - * merge two UTF-16 string parts together - * to canonically order (order by combining classes) their concatenation - * - * the two strings may already be adjacent, so that the merging is done - * in-place if the two strings are not adjacent, then the buffer holding the - * first one must be large enough - * the second string may or may not be ordered in itself - * - * before: [start]..[current] is already ordered, and - * [next]..[limit] may be ordered in itself, but - * is not in relation to [start..current[ - * after: [start..current+(limit-next)[ is ordered - * - * the algorithm is a simple bubble-sort that takes the characters from - * src[next++] and inserts them in correct combining class order into the - * preceding part of the string - * - * since this function is called much less often than the single-code point - * insertOrdered(), it just uses that for easier maintenance - * - * @return the trailing combining class - */ - private static int /*unsigned byte*/ mergeOrdered(char[] source, - int start, - int current, - char[] data, - int next, - int limit) { - int r; - int /*unsigned byte*/ cc, trailCC=0; - boolean adjacent; - - adjacent= current==next; - NextCCArgs ncArgs = new NextCCArgs(); - ncArgs.source = data; - ncArgs.next = next; - ncArgs.limit = limit; - - if(start!=current) { - - while(ncArgs.next + * If dest is a StringBuilder, then the buffer writes directly to it. + * Otherwise, the buffer maintains a StringBuilder for intermediate text segments + * until no further changes are necessary and whole segments are appended. + * append() methods that take combining-class values always write to the StringBuilder. + * Other append() methods flush and append to the Appendable. + */ + public static final class ReorderingBuffer implements Appendable { + public ReorderingBuffer(NormalizerImpl ni, Appendable dest, int destCapacity) { + impl=ni; + app=dest; + if (app instanceof StringBuilder) { + appIsStringBuilder=true; + str=(StringBuilder)dest; + // In Java, the constructor subsumes public void init(int destCapacity) + str.ensureCapacity(destCapacity); + reorderStart=0; + if(str.length()==0) { + lastCC=0; + } else { + setIterator(); + lastCC=previousCC(); + // Set reorderStart after the last code point with cc<=1 if there is one. + if(lastCC>1) { + while(previousCC()>1) {} + } + reorderStart=codePointLimit; + } + } else { + appIsStringBuilder=false; + str=new StringBuilder(); + reorderStart=0; + lastCC=0; + } + } + + public boolean isEmpty() { return str.length()==0; } + public int length() { return str.length(); } + public int getLastCC() { return lastCC; } + + public StringBuilder getStringBuilder() { return str; } + + public boolean equals(CharSequence s, int start, int limit) { + return UTF16Plus.equal(str, 0, str.length(), s, start, limit); + } + + public void append(int c, int cc) { + if(lastCC<=cc || cc==0) { + str.appendCodePoint(c); + lastCC=cc; + if(cc<=1) { + reorderStart=str.length(); + } + } else { + insert(c, cc); + } + } + public void append(CharSequence s, int start, int limit, boolean isNFD, + int leadCC, int trailCC) { + if(start==limit) { + return; + } + if(lastCC<=leadCC || leadCC==0) { + if(trailCC<=1) { + reorderStart=str.length()+(limit-start); + } else if(leadCC<=1) { + reorderStart=str.length()+1; // Ok if not a code point boundary. + } + str.append(s, start, limit); + lastCC=trailCC; + } else { + int c=Character.codePointAt(s, start); + start+=Character.charCount(c); + insert(c, leadCC); // insert first code point + while(startcc;) {} + // insert c at codePointLimit, after the character with prevCC<=cc + if(c<=0xffff) { + str.insert(codePointLimit, (char)c); + if(cc<=1) { + reorderStart=codePointLimit+1; + } + } else { + str.insert(codePointLimit, Character.toChars(c)); + if(cc<=1) { + reorderStart=codePointLimit+2; + } + } + } + + private final NormalizerImpl impl; + private final Appendable app; + private final StringBuilder str; + private final boolean appIsStringBuilder; + private int reorderStart; + private int lastCC; + + // private backward iterator + private void setIterator() { codePointStart=str.length(); } + private void skipPrevious() { // Requires 0=codePointStart) { + return 0; + } + int c=str.codePointBefore(codePointStart); + codePointStart-=Character.charCount(c); + return impl.getCCFromYesOrMaybeCP(c); + } + private int codePointStart, codePointLimit; + } + + // TODO: Propose as public API on the UTF16 class. + // TODO: Propose widening UTF16 methods that take char to take int. + // TODO: Propose widening UTF16 methods that take String to take CharSequence. + public static final class UTF16Plus { + /** + * Is this code point a lead surrogate (U+d800..U+dbff)? + * @param c code unit or code point + * @return true or false + */ + public static boolean isLeadSurrogate(int c) { return (c & 0xfffffc00) == 0xd800; } + /** + * Assuming c is a surrogate code point (UTF16.isSurrogate(c)), + * is it a lead surrogate? + * @param c code unit or code point + * @return true or false + */ + public static boolean isSurrogateLead(int c) { return (c&0x400)==0; } + + /** + * Compares two CharSequence subsequences for binary equality. + * @param s1 first sequence + * @param start1 start offset in first sequence + * @param limit1 limit offset in first sequence + * @param s2 second sequence + * @param start2 start offset in second sequence + * @param limit2 limit offset in second sequence + * @return true if s1.subSequence(start1, limit1) contains the same text + * as s2.subSequence(start2, limit2) + */ + public static boolean equal(CharSequence s1, int start1, int limit1, + CharSequence s2, int start2, int limit2) { + if((limit1-start1)!=(limit2-start2)) { + return false; + } + if(s1==s2 && start1==start2) { + return true; + } + while(start1>DELTA_SHIFT)-MAX_DELTA-1; + + // Read the normTrie. + int offset=inIndexes[IX_NORM_TRIE_OFFSET]; + int nextOffset=inIndexes[IX_EXTRA_DATA_OFFSET]; + int triePosition = bytes.position(); + normTrie = CodePointTrie.Fast16.fromBinary(bytes); + int trieLength = bytes.position() - triePosition; + if(trieLength>(nextOffset-offset)) { + throw new InternalError("Normalizer2 data: not enough bytes for normTrie"); + } + ICUBinary.skipBytes(bytes, (nextOffset-offset)-trieLength); // skip padding after trie bytes + + // Read the composition and mapping data. + offset=nextOffset; + nextOffset=inIndexes[IX_SMALL_FCD_OFFSET]; + int numChars=(nextOffset-offset)/2; + if(numChars!=0) { + maybeYesCompositions=ICUBinary.getString(bytes, numChars, 0); + extraData=maybeYesCompositions.substring((MIN_NORMAL_MAYBE_YES-minMaybeYes)>>OFFSET_SHIFT); + } + + // smallFCD: new in formatVersion 2 + offset=nextOffset; + smallFCD=new byte[0x100]; + bytes.get(smallFCD); + + return this; + } catch(IOException e) { + throw new InternalError(e); + } + } + public NormalizerImpl load(String name) { + return load(ICUBinary.getRequiredData(name)); + } + + // The trie stores values for lead surrogate code *units*. + // Surrogate code *points* are inert. + public int getNorm16(int c) { + return UTF16Plus.isLeadSurrogate(c) ? INERT : normTrie.get(c); + } + public int getRawNorm16(int c) { return normTrie.get(c); } + public boolean isAlgorithmicNoNo(int norm16) { return limitNoNo<=norm16 && norm16=MIN_NORMAL_MAYBE_YES) { + return getCCFromNormalYesOrMaybe(norm16); + } + if(norm16> OFFSET_SHIFT) & 0xff; + } + public static int getCCFromYesOrMaybe(int norm16) { + return norm16>=MIN_NORMAL_MAYBE_YES ? getCCFromNormalYesOrMaybe(norm16) : 0; + } + public int getCCFromYesOrMaybeCP(int c) { + if (c < minCompNoMaybeCP) { return 0; } + return getCCFromYesOrMaybe(getNorm16(c)); + } + + /** + * Returns the FCD data for code point c. + * @param c A Unicode code point. + * @return The lccc(c) in bits 15..8 and tccc(c) in bits 7..0. + */ + public int getFCD16(int c) { + if(c>8]; + if(bits==0) { return false; } + return ((bits>>((lead>>5)&7))&1)!=0; + } + + /** Gets the FCD value from the regular normalization data. */ + public int getFCD16FromNormData(int c) { + int norm16=getNorm16(c); + if (norm16 >= limitNoNo) { + if(norm16>=MIN_NORMAL_MAYBE_YES) { + // combining mark + norm16=getCCFromNormalYesOrMaybe(norm16); + return norm16|(norm16<<8); + } else if(norm16>=minMaybeYes) { + return 0; + } else { // isDecompNoAlgorithmic(norm16) + int deltaTrailCC = norm16 & DELTA_TCCC_MASK; + if (deltaTrailCC <= DELTA_TCCC_1) { + return deltaTrailCC >> OFFSET_SHIFT; + } + // Maps to an isCompYesAndZeroCC. + c=mapAlgorithmic(c, norm16); + norm16=getRawNorm16(c); + } + } + if(norm16<=minYesNo || isHangulLVT(norm16)) { + // no decomposition or Hangul syllable, all zeros + return 0; + } + // c decomposes, get everything from the variable-length extra data + int mapping=norm16>>OFFSET_SHIFT; + int firstUnit=extraData.charAt(mapping); + int fcd16=firstUnit>>8; // tccc + if((firstUnit&MAPPING_HAS_CCC_LCCC_WORD)!=0) { + fcd16|=extraData.charAt(mapping-1)&0xff00; // lccc + } + return fcd16; + } + + /** + * Gets the decomposition for one code point. + * @param c code point + * @return c's decomposition, if it has one; returns null if it does not have a decomposition + */ + public String getDecomposition(int c) { + int norm16; + if(c>OFFSET_SHIFT; + int length=extraData.charAt(mapping++)&MAPPING_LENGTH_MASK; + return extraData.substring(mapping, mapping+length); + } + + // Fixed norm16 values. + public static final int MIN_YES_YES_WITH_CC=0xfe02; + public static final int JAMO_VT=0xfe00; + public static final int MIN_NORMAL_MAYBE_YES=0xfc00; + public static final int JAMO_L=2; // offset=1 hasCompBoundaryAfter=FALSE + public static final int INERT=1; // offset=0 hasCompBoundaryAfter=TRUE + + // norm16 bit 0 is comp-boundary-after. + public static final int HAS_COMP_BOUNDARY_AFTER=1; + public static final int OFFSET_SHIFT=1; + + // For algorithmic one-way mappings, norm16 bits 2..1 indicate the + // tccc (0, 1, >1) for quick FCC boundary-after tests. + public static final int DELTA_TCCC_0=0; + public static final int DELTA_TCCC_1=2; + public static final int DELTA_TCCC_GT_1=4; + public static final int DELTA_TCCC_MASK=6; + public static final int DELTA_SHIFT=3; + + public static final int MAX_DELTA=0x40; + + // Byte offsets from the start of the data, after the generic header. + public static final int IX_NORM_TRIE_OFFSET=0; + public static final int IX_EXTRA_DATA_OFFSET=1; + public static final int IX_SMALL_FCD_OFFSET=2; + public static final int IX_RESERVED3_OFFSET=3; + public static final int IX_TOTAL_SIZE=7; + public static final int MIN_CCC_LCCC_CP=0x300; + // Code point thresholds for quick check codes. + public static final int IX_MIN_DECOMP_NO_CP=8; + public static final int IX_MIN_COMP_NO_MAYBE_CP=9; + + // Norm16 value thresholds for quick check combinations and types of extra data. + + /** Mappings & compositions in [minYesNo..minYesNoMappingsOnly[. */ + public static final int IX_MIN_YES_NO=10; + /** Mappings are comp-normalized. */ + public static final int IX_MIN_NO_NO=11; + public static final int IX_LIMIT_NO_NO=12; + public static final int IX_MIN_MAYBE_YES=13; + + /** Mappings only in [minYesNoMappingsOnly..minNoNo[. */ + public static final int IX_MIN_YES_NO_MAPPINGS_ONLY=14; + /** Mappings are not comp-normalized but have a comp boundary before. */ + public static final int IX_MIN_NO_NO_COMP_BOUNDARY_BEFORE=15; + /** Mappings do not have a comp boundary before. */ + public static final int IX_MIN_NO_NO_COMP_NO_MAYBE_CC=16; + /** Mappings to the empty string. */ + public static final int IX_MIN_NO_NO_EMPTY=17; + + public static final int IX_MIN_LCCC_CP=18; + public static final int IX_COUNT=20; + + public static final int MAPPING_HAS_CCC_LCCC_WORD=0x80; + public static final int MAPPING_HAS_RAW_MAPPING=0x40; + // unused bit 0x20; + public static final int MAPPING_LENGTH_MASK=0x1f; + + public static final int COMP_1_LAST_TUPLE=0x8000; + public static final int COMP_1_TRIPLE=1; + public static final int COMP_1_TRAIL_LIMIT=0x3400; + public static final int COMP_1_TRAIL_MASK=0x7ffe; + public static final int COMP_1_TRAIL_SHIFT=9; // 10-1 for the "triple" bit + public static final int COMP_2_TRAIL_SHIFT=6; + public static final int COMP_2_TRAIL_MASK=0xffc0; + + // higher-level functionality ------------------------------------------ *** + + /** + * Decomposes s[src, limit[ and writes the result to dest. + * limit can be NULL if src is NUL-terminated. + * destLengthEstimate is the initial dest buffer capacity and can be -1. + */ + public void decompose(CharSequence s, int src, int limit, StringBuilder dest, + int destLengthEstimate) { + if(destLengthEstimate<0) { + destLengthEstimate=limit-src; + } + dest.setLength(0); + ReorderingBuffer buffer=new ReorderingBuffer(this, dest, destLengthEstimate); + decompose(s, src, limit, buffer); + } + + // Dual functionality: + // buffer!=NULL: normalize + // buffer==NULL: isNormalized/quickCheck/spanQuickCheckYes + public int decompose(CharSequence s, int src, int limit, + ReorderingBuffer buffer) { + int minNoCP=minDecompNoCP; + + int prevSrc; + int c=0; + int norm16=0; + + // only for quick check + int prevBoundary=src; + int prevCC=0; + + for(;;) { + // count code units below the minimum or with irrelevant data for the quick check + for(prevSrc=src; src!=limit;) { + if( (c=s.charAt(src))=limit) { + break; + } + c=Character.codePointAt(s, src); + cc=getCC(getNorm16(c)); + }; + buffer.append(s, 0, src, false, firstCC, prevCC); + buffer.append(s, src, limit); + } + + // Very similar to composeQuickCheck(): Make the same changes in both places if relevant. + // doCompose: normalize + // !doCompose: isNormalized (buffer must be empty and initialized) + public boolean compose(CharSequence s, int src, int limit, + boolean onlyContiguous, + boolean doCompose, + ReorderingBuffer buffer) { + int prevBoundary=src; + int minNoMaybeCP=minCompNoMaybeCP; + + for (;;) { + // Fast path: Scan over a sequence of characters below the minimum "no or maybe" code point, + // or with (compYes && ccc==0) properties. + int prevSrc; + int c = 0; + int norm16 = 0; + for (;;) { + if (src == limit) { + if (prevBoundary != limit && doCompose) { + buffer.append(s, prevBoundary, limit); + } + return true; + } + if( (c=s.charAt(src))=minNoNo. + // The current character is either a "noNo" (has a mapping) + // or a "maybeYes" (combines backward) + // or a "yesYes" with ccc!=0. + // It is not a Hangul syllable or Jamo L because those have "yes" properties. + + // Medium-fast path: Handle cases that do not require full decomposition and recomposition. + if (!isMaybeOrNonZeroCC(norm16)) { // minNoNo <= norm16 < minMaybeYes + if (!doCompose) { + return false; + } + // Fast path for mapping a character that is immediately surrounded by boundaries. + // In this case, we need not decompose around the current character. + if (isDecompNoAlgorithmic(norm16)) { + // Maps to a single isCompYesAndZeroCC character + // which also implies hasCompBoundaryBefore. + if (norm16HasCompBoundaryAfter(norm16, onlyContiguous) || + hasCompBoundaryBefore(s, src, limit)) { + if (prevBoundary != prevSrc) { + buffer.append(s, prevBoundary, prevSrc); + } + buffer.append(mapAlgorithmic(c, norm16), 0); + prevBoundary = src; + continue; + } + } else if (norm16 < minNoNoCompBoundaryBefore) { + // The mapping is comp-normalized which also implies hasCompBoundaryBefore. + if (norm16HasCompBoundaryAfter(norm16, onlyContiguous) || + hasCompBoundaryBefore(s, src, limit)) { + if (prevBoundary != prevSrc) { + buffer.append(s, prevBoundary, prevSrc); + } + int mapping = norm16 >> OFFSET_SHIFT; + int length = extraData.charAt(mapping++) & MAPPING_LENGTH_MASK; + buffer.append(extraData, mapping, mapping + length); + prevBoundary = src; + continue; + } + } else if (norm16 >= minNoNoEmpty) { + // The current character maps to nothing. + // Simply omit it from the output if there is a boundary before _or_ after it. + // The character itself implies no boundaries. + if (hasCompBoundaryBefore(s, src, limit) || + hasCompBoundaryAfter(s, prevBoundary, prevSrc, onlyContiguous)) { + if (prevBoundary != prevSrc) { + buffer.append(s, prevBoundary, prevSrc); + } + prevBoundary = src; + continue; + } + } + // Other "noNo" type, or need to examine more text around this character: + // Fall through to the slow path. + } else if (isJamoVT(norm16) && prevBoundary != prevSrc) { + char prev=s.charAt(prevSrc-1); + if(c= 0) { + int syllable = Hangul.HANGUL_BASE + + (l*Hangul.JAMO_V_COUNT + (c-Hangul.JAMO_V_BASE)) * + Hangul.JAMO_T_COUNT + t; + --prevSrc; // Replace the Jamo L as well. + if (prevBoundary != prevSrc) { + buffer.append(s, prevBoundary, prevSrc); + } + buffer.append((char)syllable); + prevBoundary = src; + continue; + } + // If we see L+V+x where x!=T then we drop to the slow path, + // decompose and recompose. + // This is to deal with NFKC finding normal L and V but a + // compatibility variant of a T. + // We need to either fully compose that combination here + // (which would complicate the code and may not work with strange custom data) + // or use the slow path. + } + } else if (Hangul.isHangulLV(prev)) { + // The current character is a Jamo Trailing consonant, + // compose with previous Hangul LV that does not contain a Jamo T. + if (!doCompose) { + return false; + } + int syllable = prev + c - Hangul.JAMO_T_BASE; + --prevSrc; // Replace the Hangul LV as well. + if (prevBoundary != prevSrc) { + buffer.append(s, prevBoundary, prevSrc); + } + buffer.append((char)syllable); + prevBoundary = src; + continue; + } + // No matching context, or may need to decompose surrounding text first: + // Fall through to the slow path. + } else if (norm16 > JAMO_VT) { // norm16 >= MIN_YES_YES_WITH_CC + // One or more combining marks that do not combine-back: + // Check for canonical order, copy unchanged if ok and + // if followed by a character with a boundary-before. + int cc = getCCFromNormalYesOrMaybe(norm16); // cc!=0 + if (onlyContiguous /* FCC */ && getPreviousTrailCC(s, prevBoundary, prevSrc) > cc) { + // Fails FCD test, need to decompose and contiguously recompose. + if (!doCompose) { + return false; + } + } else { + // If !onlyContiguous (not FCC), then we ignore the tccc of + // the previous character which passed the quick check "yes && ccc==0" test. + int n16; + for (;;) { + if (src == limit) { + if (doCompose) { + buffer.append(s, prevBoundary, limit); + } + return true; + } + int prevCC = cc; + c = Character.codePointAt(s, src); + n16 = normTrie.get(c); + if (n16 >= MIN_YES_YES_WITH_CC) { + cc = getCCFromNormalYesOrMaybe(n16); + if (prevCC > cc) { + if (!doCompose) { + return false; + } + break; + } + } else { + break; + } + src += Character.charCount(c); + } + // p is after the last in-order combining mark. + // If there is a boundary here, then we continue with no change. + if (norm16HasCompBoundaryBefore(n16)) { + if (isCompYesAndZeroCC(n16)) { + src += Character.charCount(c); + } + continue; + } + // Use the slow path. There is no boundary in [prevSrc, src[. + } + } + + // Slow path: Find the nearest boundaries around the current character, + // decompose and recompose. + if (prevBoundary != prevSrc && !norm16HasCompBoundaryBefore(norm16)) { + c = Character.codePointBefore(s, prevSrc); + norm16 = normTrie.get(c); + if (!norm16HasCompBoundaryAfter(norm16, onlyContiguous)) { + prevSrc -= Character.charCount(c); + } + } + if (doCompose && prevBoundary != prevSrc) { + buffer.append(s, prevBoundary, prevSrc); + } + int recomposeStartIndex=buffer.length(); + // We know there is not a boundary here. + decomposeShort(s, prevSrc, src, false /* !stopAtCompBoundary */, onlyContiguous, + buffer); + // Decompose until the next boundary. + src = decomposeShort(s, src, limit, true /* stopAtCompBoundary */, onlyContiguous, + buffer); + recompose(buffer, recomposeStartIndex, onlyContiguous); + if(!doCompose) { + if(!buffer.equals(s, prevSrc, src)) { + return false; + } + buffer.remove(); + } + prevBoundary=src; + } + } + + /** + * Very similar to compose(): Make the same changes in both places if relevant. + * doSpan: spanQuickCheckYes (ignore bit 0 of the return value) + * !doSpan: quickCheck + * @return bits 31..1: spanQuickCheckYes (==s.length() if "yes") and + * bit 0: set if "maybe"; otherwise, if the span length<s.length() + * then the quick check result is "no" + */ + public int composeQuickCheck(CharSequence s, int src, int limit, + boolean onlyContiguous, boolean doSpan) { + int qcResult=0; + int prevBoundary=src; + int minNoMaybeCP=minCompNoMaybeCP; + + for(;;) { + // Fast path: Scan over a sequence of characters below the minimum "no or maybe" code point, + // or with (compYes && ccc==0) properties. + int prevSrc; + int c = 0; + int norm16 = 0; + for (;;) { + if(src==limit) { + return (src<<1)|qcResult; // "yes" or "maybe" + } + if( (c=s.charAt(src))=minNoNo. + // The current character is either a "noNo" (has a mapping) + // or a "maybeYes" (combines backward) + // or a "yesYes" with ccc!=0. + // It is not a Hangul syllable or Jamo L because those have "yes" properties. + + int prevNorm16 = INERT; + if (prevBoundary != prevSrc) { + prevBoundary = prevSrc; + if (!norm16HasCompBoundaryBefore(norm16)) { + c = Character.codePointBefore(s, prevSrc); + int n16 = getNorm16(c); + if (!norm16HasCompBoundaryAfter(n16, onlyContiguous)) { + prevBoundary -= Character.charCount(c); + prevNorm16 = n16; + } + } + } + + if(isMaybeOrNonZeroCC(norm16)) { + int cc=getCCFromYesOrMaybe(norm16); + if (onlyContiguous /* FCC */ && cc != 0 && + getTrailCCFromCompYesAndZeroCC(prevNorm16) > cc) { + // The [prevBoundary..prevSrc[ character + // passed the quick check "yes && ccc==0" test + // but is out of canonical order with the current combining mark. + } else { + // If !onlyContiguous (not FCC), then we ignore the tccc of + // the previous character which passed the quick check "yes && ccc==0" test. + for (;;) { + if (norm16 < MIN_YES_YES_WITH_CC) { + if (!doSpan) { + qcResult = 1; + } else { + return prevBoundary << 1; // spanYes does not care to know it's "maybe" + } + } + if (src == limit) { + return (src<<1) | qcResult; // "yes" or "maybe" + } + int prevCC = cc; + c = Character.codePointAt(s, src); + norm16 = getNorm16(c); + if (isMaybeOrNonZeroCC(norm16)) { + cc = getCCFromYesOrMaybe(norm16); + if (!(prevCC <= cc || cc == 0)) { + break; + } + } else { + break; + } + src += Character.charCount(c); + } + // src is after the last in-order combining mark. + if (isCompYesAndZeroCC(norm16)) { + prevBoundary = src; + src += Character.charCount(c); + continue; + } + } + } + return prevBoundary<<1; // "no" + } + } + public void composeAndAppend(CharSequence s, + boolean doCompose, + boolean onlyContiguous, + ReorderingBuffer buffer) { + int src=0, limit=s.length(); + if(!buffer.isEmpty()) { + int firstStarterInSrc=findNextCompBoundary(s, 0, limit, onlyContiguous); + if(0!=firstStarterInSrc) { + int lastStarterInDest=findPreviousCompBoundary(buffer.getStringBuilder(), + buffer.length(), onlyContiguous); + StringBuilder middle=new StringBuilder((buffer.length()-lastStarterInDest)+ + firstStarterInSrc+16); + middle.append(buffer.getStringBuilder(), lastStarterInDest, buffer.length()); + buffer.removeSuffix(buffer.length()-lastStarterInDest); + middle.append(s, 0, firstStarterInSrc); + compose(middle, 0, middle.length(), onlyContiguous, true, buffer); + src=firstStarterInSrc; + } + } + if(doCompose) { + compose(s, src, limit, onlyContiguous, true, buffer); + } else { + buffer.append(s, src, limit); + } + } + // Dual functionality: + // buffer!=NULL: normalize + // buffer==NULL: isNormalized/quickCheck/spanQuickCheckYes + public int makeFCD(CharSequence s, int src, int limit, ReorderingBuffer buffer) { + // Note: In this function we use buffer->appendZeroCC() because we track + // the lead and trail combining classes here, rather than leaving it to + // the ReorderingBuffer. + // The exception is the call to decomposeShort() which uses the buffer + // in the normal way. + + // Tracks the last FCD-safe boundary, before lccc=0 or after properly-ordered tccc<=1. + // Similar to the prevBoundary in the compose() implementation. + int prevBoundary=src; + int prevSrc; + int c=0; + int prevFCD16=0; + int fcd16=0; + + for(;;) { + // count code units with lccc==0 + for(prevSrc=src; src!=limit;) { + if((c=s.charAt(src))1) { + --prevBoundary; + } + } + } else { + int p=src-1; + if( Character.isLowSurrogate(s.charAt(p)) && prevSrc

1) { + prevBoundary=p; + } + } + if(buffer!=null) { + // The last lccc==0 character is excluded from the + // flush-and-append call in case it needs to be modified. + buffer.flushAndAppendZeroCC(s, prevSrc, prevBoundary); + buffer.append(s, prevBoundary, src); + } + // The start of the current character (c). + prevSrc=src; + } else if(src==limit) { + break; + } + + src+=Character.charCount(c); + // The current character (c) at [prevSrc..src[ has a non-zero lead combining class. + // Check for proper order, and decompose locally if necessary. + if((prevFCD16&0xff)<=(fcd16>>8)) { + // proper order: prev tccc <= current lccc + if((fcd16&0xff)<=1) { + prevBoundary=src; + } + if(buffer!=null) { + buffer.appendZeroCC(c); + } + prevFCD16=fcd16; + continue; + } else if(buffer==null) { + return prevBoundary; // quick check "no" + } else { + /* + * Back out the part of the source that we copied or appended + * already but is now going to be decomposed. + * prevSrc is set to after what was copied/appended. + */ + buffer.removeSuffix(prevSrc-prevBoundary); + /* + * Find the part of the source that needs to be decomposed, + * up to the next safe boundary. + */ + src=findNextFCDBoundary(s, src, limit); + /* + * The source text does not fulfill the conditions for FCD. + * Decompose and reorder a limited piece of the text. + */ + decomposeShort(s, prevBoundary, src, false, false, buffer); + prevBoundary=src; + prevFCD16=0; + } + } + return src; + } + + public boolean hasDecompBoundaryBefore(int c) { + return c < minLcccCP || (c <= 0xffff && !singleLeadMightHaveNonZeroFCD16(c)) || + norm16HasDecompBoundaryBefore(getNorm16(c)); + } + public boolean norm16HasDecompBoundaryBefore(int norm16) { + if (norm16 < minNoNoCompNoMaybeCC) { + return true; + } + if (norm16 >= limitNoNo) { + return norm16 <= MIN_NORMAL_MAYBE_YES || norm16 == JAMO_VT; + } + // c decomposes, get everything from the variable-length extra data + int mapping=norm16>>OFFSET_SHIFT; + int firstUnit=extraData.charAt(mapping); + // true if leadCC==0 (hasFCDBoundaryBefore()) + return (firstUnit&MAPPING_HAS_CCC_LCCC_WORD)==0 || (extraData.charAt(mapping-1)&0xff00)==0; + } + public boolean hasDecompBoundaryAfter(int c) { + if (c < minDecompNoCP) { + return true; + } + if (c <= 0xffff && !singleLeadMightHaveNonZeroFCD16(c)) { + return true; + } + return norm16HasDecompBoundaryAfter(getNorm16(c)); + } + public boolean norm16HasDecompBoundaryAfter(int norm16) { + if(norm16 <= minYesNo || isHangulLVT(norm16)) { + return true; + } + if (norm16 >= limitNoNo) { + if (isMaybeOrNonZeroCC(norm16)) { + return norm16 <= MIN_NORMAL_MAYBE_YES || norm16 == JAMO_VT; + } + // Maps to an isCompYesAndZeroCC. + return (norm16 & DELTA_TCCC_MASK) <= DELTA_TCCC_1; + } + // c decomposes, get everything from the variable-length extra data + int mapping=norm16>>OFFSET_SHIFT; + int firstUnit=extraData.charAt(mapping); + // decomp after-boundary: same as hasFCDBoundaryAfter(), + // fcd16<=1 || trailCC==0 + if(firstUnit>0x1ff) { + return false; // trailCC>1 + } + if(firstUnit<=0xff) { + return true; // trailCC==0 + } + // if(trailCC==1) test leadCC==0, same as checking for before-boundary + // true if leadCC==0 (hasFCDBoundaryBefore()) + return (firstUnit&MAPPING_HAS_CCC_LCCC_WORD)==0 || (extraData.charAt(mapping-1)&0xff00)==0; + } + public boolean isDecompInert(int c) { return isDecompYesAndZeroCC(getNorm16(c)); } + + public boolean hasCompBoundaryBefore(int c) { + return c=minMaybeYes; } + private static boolean isInert(int norm16) { return norm16==INERT; } + private static boolean isJamoVT(int norm16) { return norm16==JAMO_VT; } + private int hangulLVT() { return minYesNoMappingsOnly|HAS_COMP_BOUNDARY_AFTER; } + private boolean isHangulLV(int norm16) { return norm16==minYesNo; } + private boolean isHangulLVT(int norm16) { + return norm16==hangulLVT(); + } + private boolean isCompYesAndZeroCC(int norm16) { return norm16=MIN_YES_YES_WITH_CC || norm16=limitNoNo; } + + // For use with isCompYes(). + // Perhaps the compiler can combine the two tests for MIN_YES_YES_WITH_CC. + // static uint8_t getCCFromYes(uint16_t norm16) { + // return norm16>=MIN_YES_YES_WITH_CC ? getCCFromNormalYesOrMaybe(norm16) : 0; + // } + private int getCCFromNoNo(int norm16) { + int mapping=norm16>>OFFSET_SHIFT; + if((extraData.charAt(mapping)&MAPPING_HAS_CCC_LCCC_WORD)!=0) { + return extraData.charAt(mapping-1)&0xff; + } else { + return 0; + } + } + int getTrailCCFromCompYesAndZeroCC(int norm16) { + if(norm16<=minYesNo) { + return 0; // yesYes and Hangul LV have ccc=tccc=0 + } else { + // For Hangul LVT we harmlessly fetch a firstUnit with tccc=0 here. + return extraData.charAt(norm16>>OFFSET_SHIFT)>>8; // tccc from yesNo + } + } + + // Requires algorithmic-NoNo. + private int mapAlgorithmic(int c, int norm16) { + return c+(norm16>>DELTA_SHIFT)-centerNoNoDelta; + } + + // Requires minYesNo>OFFSET_SHIFT); } + + /** + * @return index into maybeYesCompositions, or -1 + */ + private int getCompositionsListForDecompYes(int norm16) { + if(norm16>OFFSET_SHIFT; + } + } + /** + * @return index into maybeYesCompositions + */ + private int getCompositionsListForComposite(int norm16) { + // A composite has both mapping & compositions list. + int list=((MIN_NORMAL_MAYBE_YES-minMaybeYes)+norm16)>>OFFSET_SHIFT; + int firstUnit=maybeYesCompositions.charAt(list); + return list+ // mapping in maybeYesCompositions + 1+ // +1 to skip the first unit with the mapping length + (firstUnit&MAPPING_LENGTH_MASK); // + mapping length + } + + // Decompose a short piece of text which is likely to contain characters that + // fail the quick check loop and/or where the quick check loop's overhead + // is unlikely to be amortized. + // Called by the compose() and makeFCD() implementations. + // Public in Java for collation implementation code. + private int decomposeShort( + CharSequence s, int src, int limit, + boolean stopAtCompBoundary, boolean onlyContiguous, + ReorderingBuffer buffer) { + while(src= limitNoNo) { + if (isMaybeOrNonZeroCC(norm16)) { + buffer.append(c, getCCFromYesOrMaybe(norm16)); + return; + } + // Maps to an isCompYesAndZeroCC. + c=mapAlgorithmic(c, norm16); + norm16=getRawNorm16(c); + } + if (norm16 < minYesNo) { + // c does not decompose + buffer.append(c, 0); + } else if(isHangulLV(norm16) || isHangulLVT(norm16)) { + // Hangul syllable: decompose algorithmically + Hangul.decompose(c, buffer); + } else { + // c decomposes, get everything from the variable-length extra data + int mapping=norm16>>OFFSET_SHIFT; + int firstUnit=extraData.charAt(mapping); + int length=firstUnit&MAPPING_LENGTH_MASK; + int leadCC, trailCC; + trailCC=firstUnit>>8; + if((firstUnit&MAPPING_HAS_CCC_LCCC_WORD)!=0) { + leadCC=extraData.charAt(mapping-1)>>8; + } else { + leadCC=0; + } + ++mapping; // skip over the firstUnit + buffer.append(extraData, mapping, mapping+length, true, leadCC, trailCC); + } + } + + /** + * Finds the recomposition result for + * a forward-combining "lead" character, + * specified with a pointer to its compositions list, + * and a backward-combining "trail" character. + * + *

If the lead and trail characters combine, then this function returns + * the following "compositeAndFwd" value: + * 

+     * Bits 21..1  composite character
+     * Bit      0  set if the composite is a forward-combining starter
+     *
+ * otherwise it returns -1. + * + *

The compositions list has (trail, compositeAndFwd) pair entries, + * encoded as either pairs or triples of 16-bit units. + * The last entry has the high bit of its first unit set. + * + *

The list is sorted by ascending trail characters (there are no duplicates). + * A linear search is used. + * + *

See normalizer2impl.h for a more detailed description + * of the compositions list format. + */ + private static int combine(String compositions, int list, int trail) { + int key1, firstUnit; + if(trail(firstUnit=compositions.charAt(list))) { + list+=2+(firstUnit&COMP_1_TRIPLE); + } + if(key1==(firstUnit&COMP_1_TRAIL_MASK)) { + if((firstUnit&COMP_1_TRIPLE)!=0) { + return (compositions.charAt(list+1)<<16)|compositions.charAt(list+2); + } else { + return compositions.charAt(list+1); + } + } + } else { + // trail character is 3400..10FFFF + // result entry has 3 units + key1=COMP_1_TRAIL_LIMIT+(((trail>>COMP_1_TRAIL_SHIFT))&~COMP_1_TRIPLE); + int key2=(trail<(firstUnit=compositions.charAt(list))) { + list+=2+(firstUnit&COMP_1_TRIPLE); + } else if(key1==(firstUnit&COMP_1_TRAIL_MASK)) { + if(key2>(secondUnit=compositions.charAt(list+1))) { + if((firstUnit&COMP_1_LAST_TUPLE)!=0) { + break; + } else { + list+=3; + } + } else if(key2==(secondUnit&COMP_2_TRAIL_MASK)) { + return ((secondUnit&~COMP_2_TRAIL_MASK)<<16)|compositions.charAt(list+2); + } else { + break; + } + } else { + break; + } + } + } + return -1; + } + + /* + * Recomposes the buffer text starting at recomposeStartIndex + * (which is in NFD - decomposed and canonically ordered), + * and truncates the buffer contents. + * + * Note that recomposition never lengthens the text: + * Any character consists of either one or two code units; + * a composition may contain at most one more code unit than the original starter, + * while the combining mark that is removed has at least one code unit. + */ + private void recompose(ReorderingBuffer buffer, int recomposeStartIndex, + boolean onlyContiguous) { + StringBuilder sb=buffer.getStringBuilder(); + int p=recomposeStartIndex; + if(p==sb.length()) { + return; + } + + int starter, pRemove; + int compositionsList; + int c, compositeAndFwd; + int norm16; + int cc, prevCC; + boolean starterIsSupplementary; + + // Some of the following variables are not used until we have a forward-combining starter + // and are only initialized now to avoid compiler warnings. + compositionsList=-1; // used as indicator for whether we have a forward-combining starter + starter=-1; + starterIsSupplementary=false; + prevCC=0; + + for(;;) { + c=sb.codePointAt(p); + p+=Character.charCount(c); + norm16=getNorm16(c); + cc=getCCFromYesOrMaybe(norm16); + if( // this character combines backward and + isMaybe(norm16) && + // we have seen a starter that combines forward and + compositionsList>=0 && + // the backward-combining character is not blocked + (prevCC=0) { + // The starter and the combining mark (c) do combine. + int composite=compositeAndFwd>>1; + + // Remove the combining mark. + pRemove=p-Character.charCount(c); // pRemove & p: start & limit of the combining mark + sb.delete(pRemove, p); + p=pRemove; + // Replace the starter with the composite. + if(starterIsSupplementary) { + if(composite>0xffff) { + // both are supplementary + sb.setCharAt(starter, UTF16.getLeadSurrogate(composite)); + sb.setCharAt(starter+1, UTF16.getTrailSurrogate(composite)); + } else { + sb.setCharAt(starter, (char)c); + sb.deleteCharAt(starter+1); + // The composite is shorter than the starter, + // move the intermediate characters forward one. + starterIsSupplementary=false; + --p; + } + } else if(composite>0xffff) { + // The composite is longer than the starter, + // move the intermediate characters back one. + starterIsSupplementary=true; + sb.setCharAt(starter, UTF16.getLeadSurrogate(composite)); + sb.insert(starter+1, UTF16.getTrailSurrogate(composite)); + ++p; + } else { + // both are on the BMP + sb.setCharAt(starter, (char)composite); + } + + // Keep prevCC because we removed the combining mark. + + if(p==sb.length()) { + break; + } + // Is the composite a starter that combines forward? + if((compositeAndFwd&1)!=0) { + compositionsList= + getCompositionsListForComposite(getRawNorm16(composite)); + } else { + compositionsList=-1; + } + + // We combined; continue with looking for compositions. + continue; + } + } + + // no combination this time + prevCC=cc; + if(p==sb.length()) { + break; + } + + // If c did not combine, then check if it is a starter. + if(cc==0) { + // Found a new starter. + if((compositionsList=getCompositionsListForDecompYes(norm16))>=0) { + // It may combine with something, prepare for it. + if(c<=0xffff) { + starterIsSupplementary=false; + starter=p-1; + } else { + starterIsSupplementary=true; + starter=p-2; + } + } + } else if(onlyContiguous) { + // FCC: no discontiguous compositions; any intervening character blocks. + compositionsList=-1; + } + } + buffer.flush(); + } + + /** + * Does c have a composition boundary before it? + * True if its decomposition begins with a character that has + * ccc=0 && NFC_QC=Yes (isCompYesAndZeroCC()). + * As a shortcut, this is true if c itself has ccc=0 && NFC_QC=Yes + * (isCompYesAndZeroCC()) so we need not decompose. + */ + private boolean hasCompBoundaryBefore(int c, int norm16) { + return c> OFFSET_SHIFT) <= 0x1ff); + } + + private int findPreviousCompBoundary(CharSequence s, int p, boolean onlyContiguous) { + while(p>0) { + int c=Character.codePointBefore(s, p); + int norm16 = getNorm16(c); + if (norm16HasCompBoundaryAfter(norm16, onlyContiguous)) { + break; + } + p-=Character.charCount(c); + if(hasCompBoundaryBefore(c, norm16)) { + break; + } + } + return p; + } + private int findNextCompBoundary(CharSequence s, int p, int limit, boolean onlyContiguous) { + while(p= 0x0009 && c <= 0x000D) || + (c >= 0x0020 && c <= 0x002F) || + (c >= 0x003A && c <= 0x0040) || + (c >= 0x005B && c <= 0x0060) || + (c >= 0x007B && c <= 0x007E); + } + + public static String canonicalDecomposeWithSingleQuotation(String string) { + Normalizer2 impl = Normalizer2.getNFDInstance(); + char[] src = string.toCharArray(); + int srcIndex = 0; + int srcLimit = src.length; + char[] dest = new char[src.length * 3]; //MAX_BUF_SIZE_DECOMPOSE = 3 + int destIndex = 0; + int destLimit = dest.length; + + int prevSrc; + String norm; + int reorderStartIndex, length; + char c1, c2; + int cp; + int minNoMaybe = 0x00c0; + int cc, prevCC, trailCC; + char[] p; + int pStart; + + // initialize + reorderStartIndex = 0; + prevCC = 0; + norm = null; + cp = 0; + pStart = 0; + + cc = trailCC = -1; // initialize to bogus value + c1 = 0; + for (;;) { + prevSrc=srcIndex; + //quick check (1)less than minNoMaybe (2)no decomp (3)hangual + while (srcIndex != srcLimit && + ((c1 = src[srcIndex]) < minNoMaybe || + (norm = impl.getDecomposition(cp = string.codePointAt(srcIndex))) == null || + (c1 >= '\uac00' && c1 <= '\ud7a3'))) { // Hangul Syllables + prevCC = 0; + srcIndex += (cp < 0x10000) ? 1 : 2; + } + + // copy these code units all at once + if (srcIndex != prevSrc) { + length = srcIndex - prevSrc; + if ((destIndex + length) <= destLimit) { + System.arraycopy(src,prevSrc,dest,destIndex,length); + } + + destIndex += length; + reorderStartIndex = destIndex; + } + + // end of source reached? + if (srcIndex == srcLimit) { + break; + } + + // cp already contains *src and norm32 is set for it, increment src + srcIndex += (cp < 0x10000) ? 1 : 2; + + if (cp < Character.MIN_SUPPLEMENTARY_CODE_POINT) { + c2 = 0; + length = 1; + + if (Character.isHighSurrogate(c1) + || Character.isLowSurrogate(c1)) { + norm = null; + } + } else { + length = 2; + c2 = src[srcIndex-1]; + } + + // get the decomposition and the lead and trail cc's + if (norm == null) { + // cp does not decompose + cc = trailCC = UCharacter.getCombiningClass(cp); + p = null; + pStart = -1; + } else { + + pStart = 0; + p = norm.toCharArray(); + length = p.length; + int cpNum = norm.codePointCount(0, length); + cc= UCharacter.getCombiningClass(norm.codePointAt(0)); + trailCC= UCharacter.getCombiningClass(norm.codePointAt(cpNum-1)); + if (length == 1) { + // fastpath a single code unit from decomposition + c1 = p[pStart]; + c2 = 0; + p = null; + pStart = -1; + } + } + + if((destIndex + length * 3) >= destLimit) { // 2 SingleQuotations + // buffer overflow + char[] tmpBuf = new char[destLimit * 2]; + System.arraycopy(dest, 0, tmpBuf, 0, destIndex); + dest = tmpBuf; + destLimit = dest.length; + } + + // append the decomposition to the destination buffer, assume length>0 + { + int reorderSplit = destIndex; + if (p == null) { + // fastpath: single code point + if (needSingleQuotation(c1)) { + //if we need single quotation, no need to consider "prevCC" + //and it must NOT be a supplementary pair + dest[destIndex++] = '\''; + dest[destIndex++] = c1; + dest[destIndex++] = '\''; + trailCC = 0; + } else if(cc != 0 && cc < prevCC) { + // (c1, c2) is out of order with respect to the preceding + // text + destIndex += length; + trailCC = insertOrdered(dest, reorderStartIndex, + reorderSplit, destIndex, c1, c2, cc); + } else { + // just append (c1, c2) + dest[destIndex++] = c1; + if(c2 != 0) { + dest[destIndex++] = c2; + } + } + } else { + // general: multiple code points (ordered by themselves) + // from decomposition + if (needSingleQuotation(p[pStart])) { + dest[destIndex++] = '\''; + dest[destIndex++] = p[pStart++]; + dest[destIndex++] = '\''; + length--; + do { + dest[destIndex++] = p[pStart++]; + } while(--length > 0); + } else if (cc != 0 && cc < prevCC) { + destIndex += length; + trailCC = mergeOrdered(dest, reorderStartIndex, + reorderSplit, p, pStart, + pStart+length); + } else { + // just append the decomposition + do { + dest[destIndex++] = p[pStart++]; + } while (--length > 0); + } + } + } + prevCC = trailCC; + if(prevCC == 0) { + reorderStartIndex = destIndex; + } + } + + return new String(dest, 0, destIndex); + } + + /** + * simpler, single-character version of mergeOrdered() - + * bubble-insert one single code point into the preceding string + * which is already canonically ordered + * (c, c2) may or may not yet have been inserted at src[current]..src[p] + * + * it must be p=current+lengthof(c, c2) i.e. p=current+(c2==0 ? 1 : 2) + * + * before: src[start]..src[current] is already ordered, and + * src[current]..src[p] may or may not hold (c, c2) but + * must be exactly the same length as (c, c2) + * after: src[start]..src[p] is ordered + * + * @return the trailing combining class + */ + private static int/*unsigned byte*/ insertOrdered(char[] source, + int start, + int current, int p, + char c1, char c2, + int/*unsigned byte*/ cc) { + int back, preBack; + int r; + int prevCC, trailCC=cc; + + if (start=prevCC + preBack=back=current; + + PrevArgs prevArgs = new PrevArgs(); + prevArgs.current = current; + prevArgs.start = start; + prevArgs.src = source; + prevArgs.c1 = c1; + prevArgs.c2 = c2; + + // get the prevCC + prevCC=getPrevCC(prevArgs); + preBack = prevArgs.current; + + if(cc=prevCC) { + break; + } + back=preBack; + } + + // this is where we are right now with all these indicies: + // [start]..[pPreBack] 0..? code points that we can ignore + // [pPreBack]..[pBack] 0..1 code points with prevCC<=cc + // [pBack]..[current] 0..n code points with >cc, move up to insert (c, c2) + // [current]..[p] 1 code point (c, c2) with cc + + // move the code units in between up + r=p; + do { + source[--r]=source[--current]; + } while (back!=current); + } + } + + // insert (c1, c2) + source[current] = c1; + if (c2!=0) { + source[(current+1)] = c2; + } + + // we know the cc of the last code point + return trailCC; + } + /** + * merge two UTF-16 string parts together + * to canonically order (order by combining classes) their concatenation + * + * the two strings may already be adjacent, so that the merging is done + * in-place if the two strings are not adjacent, then the buffer holding the + * first one must be large enough + * the second string may or may not be ordered in itself + * + * before: [start]..[current] is already ordered, and + * [next]..[limit] may be ordered in itself, but + * is not in relation to [start..current[ + * after: [start..current+(limit-next)[ is ordered + * + * the algorithm is a simple bubble-sort that takes the characters from + * src[next++] and inserts them in correct combining class order into the + * preceding part of the string + * + * since this function is called much less often than the single-code point + * insertOrdered(), it just uses that for easier maintenance + * + * @return the trailing combining class + */ + private static int /*unsigned byte*/ mergeOrdered(char[] source, + int start, + int current, + char[] data, + int next, + int limit) { + int r; + int /*unsigned byte*/ cc, trailCC=0; + boolean adjacent; + + adjacent= current==next; + NextCCArgs ncArgs = new NextCCArgs(); + ncArgs.source = data; + ncArgs.next = next; + ncArgs.limit = limit; + + if(start!=current) { + + while(ncArgs.next((BASE-TMIN)*TMAX)/2; count+=BASE) { - delta/=(BASE-TMIN); - } - - return count+(((BASE-TMIN+1)*delta)/(delta+SKEW)); - } - - /** - * basicToDigit[] contains the numeric value of a basic code - * point (for use in representing integers) in the range 0 to - * BASE-1, or -1 if b is does not represent a value. - */ - static final int[] basicToDigit= new int[]{ - -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, - -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, - - -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, - 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, -1, -1, -1, -1, -1, -1, - - -1, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, - 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, -1, -1, -1, -1, -1, - - -1, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, - 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, -1, -1, -1, -1, -1, - - -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, - -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, - - -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, - -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, - - -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, - -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, - - -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, - -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1 - }; - - private static char asciiCaseMap(char b, boolean uppercase) { - if(uppercase) { - if(SMALL_A<=b && b<=SMALL_Z) { - b-=(SMALL_A-CAPITAL_A); - } - } else { - if(CAPITAL_A<=b && b<=CAPITAL_Z) { - b+=(SMALL_A-CAPITAL_A); - } - } - return b; - } - - /** - * digitToBasic() returns the basic code point whose value - * (when used for representing integers) is d, which must be in the - * range 0 to BASE-1. The lowercase form is used unless the uppercase flag is - * nonzero, in which case the uppercase form is used. - */ - private static char digitToBasic(int digit, boolean uppercase) { - /* 0..25 map to ASCII a..z or A..Z */ - /* 26..35 map to ASCII 0..9 */ - if(digit<26) { - if(uppercase) { - return (char)(CAPITAL_A+digit); - } else { - return (char)(SMALL_A+digit); - } - } else { - return (char)((ZERO-26)+digit); - } - } - /** - * Converts Unicode to Punycode. - * The input string must not contain single, unpaired surrogates. - * The output will be represented as an array of ASCII code points. - * - * @param src - * @param caseFlags - * @return - * @throws ParseException - */ - public static StringBuffer encode(StringBuffer src, boolean[] caseFlags) throws ParseException{ - - int[] cpBuffer = new int[MAX_CP_COUNT]; - int n, delta, handledCPCount, basicLength, destLength, bias, j, m, q, k, t, srcCPCount; - char c, c2; - int srcLength = src.length(); - int destCapacity = MAX_CP_COUNT; - char[] dest = new char[destCapacity]; - StringBuffer result = new StringBuffer(); - /* - * Handle the basic code points and - * convert extended ones to UTF-32 in cpBuffer (caseFlag in sign bit): - */ - srcCPCount=destLength=0; - - for(j=0; j0) { - if(destLength state to , but guard against overflow: - */ - if(m-n>(0x7fffffff-MAX_CP_COUNT-delta)/(handledCPCount+1)) { - throw new RuntimeException("Internal program error"); - } - delta+=(m-n)*(handledCPCount+1); - n=m; - - /* Encode a sequence of same code points n */ - for(j=0; jTMAX) { - t=TMAX; - } - */ - - t=k-bias; - if(t=(bias+TMAX)) { - t=TMAX; - } - - if(q0;) { - if(src.charAt(--j)==DELIMITER) { - break; - } - } - destLength=basicLength=destCPCount=j; - - while(j>0) { - b=src.charAt(--j); - if(!isBasic(b)) { - throw new ParseException("Illegal char found", -1); - } - - if(j0 ? basicLength+1 : 0; in=srcLength) { - throw new ParseException("Illegal char found", -1); - } - - digit=basicToDigit[(byte)src.charAt(in++)]; - if(digit<0) { - throw new ParseException("Invalid char found", -1); - } - if(digit>(0x7fffffff-i)/w) { - /* integer overflow */ - throw new ParseException("Illegal char found", -1); - } - - i+=digit*w; - t=k-bias; - if(t=(bias+TMAX)) { - t=TMAX; - } - if(digit0x7fffffff/(BASE-t)) { - /* integer overflow */ - throw new ParseException("Illegal char found", -1); - } - w*=BASE-t; - } - - /* - * Modification from sample code: - * Increments destCPCount here, - * where needed instead of in for() loop tail. - */ - ++destCPCount; - bias=adaptBias(i-oldi, destCPCount, (oldi==0)); - - /* - * i was supposed to wrap around from (incremented) destCPCount to 0, - * incrementing n each time, so we'll fix that now: - */ - if(i/destCPCount>(0x7fffffff-n)) { - /* integer overflow */ - throw new ParseException("Illegal char found", -1); - } - - n+=i/destCPCount; - i%=destCPCount; - /* not needed for Punycode: */ - /* if (decode_digit(n) <= BASE) return punycode_invalid_input; */ - - if(n>0x10ffff || isSurrogate(n)) { - /* Unicode code point overflow */ - throw new ParseException("Illegal char found", -1); - } - - /* Insert n at position i of the output: */ - cpLength=UTF16.getCharCount(n); - if((destLength+cpLength)1) { - firstSupplementaryIndex=codeUnitIndex; - } else { - ++firstSupplementaryIndex; - } - } else { - codeUnitIndex=firstSupplementaryIndex; - codeUnitIndex=UTF16.moveCodePointOffset(dest, 0, destLength, codeUnitIndex, i-codeUnitIndex); - } - - /* use the UChar index codeUnitIndex instead of the code point index i */ - if(codeUnitIndex((BASE-TMIN)*TMAX)/2; count+=BASE) { + delta/=(BASE-TMIN); + } + + return count+(((BASE-TMIN+1)*delta)/(delta+SKEW)); + } + + /** + * basicToDigit[] contains the numeric value of a basic code + * point (for use in representing integers) in the range 0 to + * BASE-1, or -1 if b is does not represent a value. + */ + static final int[] basicToDigit= new int[]{ + -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, + + -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, + 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, -1, -1, -1, -1, -1, -1, + + -1, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, + 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, -1, -1, -1, -1, -1, + + -1, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, + 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, -1, -1, -1, -1, -1, + + -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, + + -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, + + -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, + + -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, + -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1 + }; + + private static char asciiCaseMap(char b, boolean uppercase) { + if(uppercase) { + if(SMALL_A<=b && b<=SMALL_Z) { + b-=(SMALL_A-CAPITAL_A); + } + } else { + if(CAPITAL_A<=b && b<=CAPITAL_Z) { + b+=(SMALL_A-CAPITAL_A); + } + } + return b; + } + + /** + * digitToBasic() returns the basic code point whose value + * (when used for representing integers) is d, which must be in the + * range 0 to BASE-1. The lowercase form is used unless the uppercase flag is + * nonzero, in which case the uppercase form is used. + */ + private static char digitToBasic(int digit, boolean uppercase) { + /* 0..25 map to ASCII a..z or A..Z */ + /* 26..35 map to ASCII 0..9 */ + if(digit<26) { + if(uppercase) { + return (char)(CAPITAL_A+digit); + } else { + return (char)(SMALL_A+digit); + } + } else { + return (char)((ZERO-26)+digit); + } + } + /** + * Converts Unicode to Punycode. + * The input string must not contain single, unpaired surrogates. + * The output will be represented as an array of ASCII code points. + * + * @param src + * @param caseFlags + * @return + * @throws ParseException + */ + public static StringBuffer encode(StringBuffer src, boolean[] caseFlags) throws ParseException{ + + int[] cpBuffer = new int[MAX_CP_COUNT]; + int n, delta, handledCPCount, basicLength, destLength, bias, j, m, q, k, t, srcCPCount; + char c, c2; + int srcLength = src.length(); + int destCapacity = MAX_CP_COUNT; + char[] dest = new char[destCapacity]; + StringBuffer result = new StringBuffer(); + /* + * Handle the basic code points and + * convert extended ones to UTF-32 in cpBuffer (caseFlag in sign bit): + */ + srcCPCount=destLength=0; + + for(j=0; j0) { + if(destLength state to , but guard against overflow: + */ + if(m-n>(0x7fffffff-MAX_CP_COUNT-delta)/(handledCPCount+1)) { + throw new RuntimeException("Internal program error"); + } + delta+=(m-n)*(handledCPCount+1); + n=m; + + /* Encode a sequence of same code points n */ + for(j=0; jTMAX) { + t=TMAX; + } + */ + + t=k-bias; + if(t=(bias+TMAX)) { + t=TMAX; + } + + if(q0;) { + if(src.charAt(--j)==DELIMITER) { + break; + } + } + destLength=basicLength=destCPCount=j; + + while(j>0) { + b=src.charAt(--j); + if(!isBasic(b)) { + throw new ParseException("Illegal char found", -1); + } + + if(j0 ? basicLength+1 : 0; in=srcLength) { + throw new ParseException("Illegal char found", -1); + } + + digit=basicToDigit[(byte)src.charAt(in++)]; + if(digit<0) { + throw new ParseException("Invalid char found", -1); + } + if(digit>(0x7fffffff-i)/w) { + /* integer overflow */ + throw new ParseException("Illegal char found", -1); + } + + i+=digit*w; + t=k-bias; + if(t=(bias+TMAX)) { + t=TMAX; + } + if(digit0x7fffffff/(BASE-t)) { + /* integer overflow */ + throw new ParseException("Illegal char found", -1); + } + w*=BASE-t; + } + + /* + * Modification from sample code: + * Increments destCPCount here, + * where needed instead of in for() loop tail. + */ + ++destCPCount; + bias=adaptBias(i-oldi, destCPCount, (oldi==0)); + + /* + * i was supposed to wrap around from (incremented) destCPCount to 0, + * incrementing n each time, so we'll fix that now: + */ + if(i/destCPCount>(0x7fffffff-n)) { + /* integer overflow */ + throw new ParseException("Illegal char found", -1); + } + + n+=i/destCPCount; + i%=destCPCount; + /* not needed for Punycode: */ + /* if (decode_digit(n) <= BASE) return punycode_invalid_input; */ + + if(n>0x10ffff || isSurrogate(n)) { + /* Unicode code point overflow */ + throw new ParseException("Illegal char found", -1); + } + + /* Insert n at position i of the output: */ + cpLength=UTF16.getCharCount(n); + if((destLength+cpLength)1) { + firstSupplementaryIndex=codeUnitIndex; + } else { + ++firstSupplementaryIndex; + } + } else { + codeUnitIndex=firstSupplementaryIndex; + codeUnitIndex=UTF16.moveCodePointOffset(dest, 0, destLength, codeUnitIndex, i-codeUnitIndex); + } + + /* use the UChar index codeUnitIndex instead of the code point index i */ + if(codeUnitIndexReplaceableobject - * @return copy of this iterator - */ - public Object clone(){ - try { - return super.clone(); - } catch (CloneNotSupportedException e) { - return null; // never invoked - } - } - - /** - * Returns the current UTF16 character. - * @return current UTF16 character - */ - public int current(){ - if (currentIndex < replaceable.length()) { - return replaceable.charAt(currentIndex); - } - return DONE; - } - - /** - * Returns the length of the text - * @return length of the text - */ - public int getLength(){ - return replaceable.length(); - } - - /** - * Gets the current currentIndex in text. - * @return current currentIndex in text. - */ - public int getIndex(){ - return currentIndex; - } - - /** - * Returns next UTF16 character and increments the iterator's currentIndex by 1. - * If the resulting currentIndex is greater or equal to the text length, the - * currentIndex is reset to the text length and a value of DONECODEPOINT is - * returned. - * @return next UTF16 character in text or DONE if the new currentIndex is off the - * end of the text range. - */ - public int next(){ - if (currentIndex < replaceable.length()) { - return replaceable.charAt(currentIndex++); - } - return DONE; - } - - - /** - * Returns previous UTF16 character and decrements the iterator's currentIndex by - * 1. - * If the resulting currentIndex is less than 0, the currentIndex is reset to 0 and a - * value of DONECODEPOINT is returned. - * @return next UTF16 character in text or DONE if the new currentIndex is off the - * start of the text range. - */ - public int previous(){ - if (currentIndex > 0) { - return replaceable.charAt(--currentIndex); - } - return DONE; - } - - /** - * Sets the currentIndex to the specified currentIndex in the text and returns that - * single UTF16 character at currentIndex. - * This assumes the text is stored as 16-bit code units. - * @param currentIndex the currentIndex within the text. - * @exception IllegalArgumentException is thrown if an invalid currentIndex is - * supplied. i.e. currentIndex is out of bounds. - */ - public void setIndex(int currentIndex) { - if (currentIndex < 0 || currentIndex > replaceable.length()) { - throw new IllegalArgumentException(); - } - this.currentIndex = currentIndex; - } - - public int getText(char[] fillIn, int offset){ - int length = replaceable.length(); - if(offset < 0 || offset + length > fillIn.length){ - throw new IndexOutOfBoundsException(Integer.toString(length)); - } - replaceable.getChars(0,length,fillIn,offset); - return length; - } - - // private data members ---------------------------------------------------- - - /** - * Replaceable object - */ - private Replaceable replaceable; - /** - * Current currentIndex - */ - private int currentIndex; - -} --- /dev/null 2020-01-10 15:57:33.000000000 -0800 +++ new/src/java.base/share/classes/jdk/internal/icu/impl/ReplaceableUCharacterIterator.java 2020-01-10 15:57:33.000000000 -0800 @@ -0,0 +1,190 @@ +/* + * Copyright (c) 2005, 2020, Oracle and/or its affiliates. All rights reserved. + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This code is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License version 2 only, as + * published by the Free Software Foundation. Oracle designates this + * particular file as subject to the "Classpath" exception as provided + * by Oracle in the LICENSE file that accompanied this code. + * + * This code is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License + * version 2 for more details (a copy is included in the LICENSE file that + * accompanied this code). + * + * You should have received a copy of the GNU General Public License version + * 2 along with this work; if not, write to the Free Software Foundation, + * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. + * + * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA + * or visit www.oracle.com if you need additional information or have any + * questions. + */ + +/* + ******************************************************************************* + * (C) Copyright IBM Corp. 1996-2005 - All Rights Reserved * + * * + * The original version of this source code and documentation is copyrighted * + * and owned by IBM, These materials are provided under terms of a License * + * Agreement between IBM and Sun. This technology is protected by multiple * + * US and International patents. This notice and attribution to IBM may not * + * to removed. * + ******************************************************************************* + */ + +package jdk.internal.icu.impl; + +import jdk.internal.icu.text.Replaceable; +import jdk.internal.icu.text.ReplaceableString; +import jdk.internal.icu.text.UCharacterIterator; + +/** + * DLF docs must define behavior when Replaceable is mutated underneath + * the iterator. + * + * This and ICUCharacterIterator share some code, maybe they should share + * an implementation, or the common state and implementation should be + * moved up into UCharacterIterator. + * + * What are first, last, and getBeginIndex doing here?!?!?! + */ +public class ReplaceableUCharacterIterator extends UCharacterIterator { + + // public constructor ------------------------------------------------------ + + /** + * Public constructor + * @param str text which the iterator will be based on + */ + public ReplaceableUCharacterIterator(String str){ + if(str==null){ + throw new IllegalArgumentException(); + } + this.replaceable = new ReplaceableString(str); + this.currentIndex = 0; + } + + /** + * Public constructor + * @param buf buffer of text on which the iterator will be based + */ + public ReplaceableUCharacterIterator(StringBuffer buf){ + if(buf==null){ + throw new IllegalArgumentException(); + } + this.replaceable = new ReplaceableString(buf); + this.currentIndex = 0; + } + + // public methods ---------------------------------------------------------- + + /** + * Creates a copy of this iterator, does not clone the underlying + * Replaceableobject + * @return copy of this iterator + */ + public Object clone(){ + try { + return super.clone(); + } catch (CloneNotSupportedException e) { + return null; // never invoked + } + } + + /** + * Returns the current UTF16 character. + * @return current UTF16 character + */ + public int current(){ + if (currentIndex < replaceable.length()) { + return replaceable.charAt(currentIndex); + } + return DONE; + } + + /** + * Returns the length of the text + * @return length of the text + */ + public int getLength(){ + return replaceable.length(); + } + + /** + * Gets the current currentIndex in text. + * @return current currentIndex in text. + */ + public int getIndex(){ + return currentIndex; + } + + /** + * Returns next UTF16 character and increments the iterator's currentIndex by 1. + * If the resulting currentIndex is greater or equal to the text length, the + * currentIndex is reset to the text length and a value of DONECODEPOINT is + * returned. + * @return next UTF16 character in text or DONE if the new currentIndex is off the + * end of the text range. + */ + public int next(){ + if (currentIndex < replaceable.length()) { + return replaceable.charAt(currentIndex++); + } + return DONE; + } + + + /** + * Returns previous UTF16 character and decrements the iterator's currentIndex by + * 1. + * If the resulting currentIndex is less than 0, the currentIndex is reset to 0 and a + * value of DONECODEPOINT is returned. + * @return next UTF16 character in text or DONE if the new currentIndex is off the + * start of the text range. + */ + public int previous(){ + if (currentIndex > 0) { + return replaceable.charAt(--currentIndex); + } + return DONE; + } + + /** + * Sets the currentIndex to the specified currentIndex in the text and returns that + * single UTF16 character at currentIndex. + * This assumes the text is stored as 16-bit code units. + * @param currentIndex the currentIndex within the text. + * @exception IllegalArgumentException is thrown if an invalid currentIndex is + * supplied. i.e. currentIndex is out of bounds. + */ + public void setIndex(int currentIndex) { + if (currentIndex < 0 || currentIndex > replaceable.length()) { + throw new IllegalArgumentException(); + } + this.currentIndex = currentIndex; + } + + public int getText(char[] fillIn, int offset){ + int length = replaceable.length(); + if(offset < 0 || offset + length > fillIn.length){ + throw new IndexOutOfBoundsException(Integer.toString(length)); + } + replaceable.getChars(0,length,fillIn,offset); + return length; + } + + // private data members ---------------------------------------------------- + + /** + * Replaceable object + */ + private Replaceable replaceable; + /** + * Current currentIndex + */ + private int currentIndex; + +} --- old/src/java.base/share/classes/sun/net/idn/StringPrepDataReader.java 2020-01-10 15:57:35.000000000 -0800 +++ /dev/null 2020-01-10 15:57:35.000000000 -0800 @@ -1,127 +0,0 @@ -/* - * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. - * - * This code is free software; you can redistribute it and/or modify it - * under the terms of the GNU General Public License version 2 only, as - * published by the Free Software Foundation. Oracle designates this - * particular file as subject to the "Classpath" exception as provided - * by Oracle in the LICENSE file that accompanied this code. - * - * This code is distributed in the hope that it will be useful, but WITHOUT - * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or - * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License - * version 2 for more details (a copy is included in the LICENSE file that - * accompanied this code). - * - * You should have received a copy of the GNU General Public License version - * 2 along with this work; if not, write to the Free Software Foundation, - * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. - * - * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA - * or visit www.oracle.com if you need additional information or have any - * questions. - */ -/* -/* - ****************************************************************************** - * Copyright (C) 2003, International Business Machines Corporation and * - * others. All Rights Reserved. * - ****************************************************************************** - * - * Created on May 2, 2003 - * - * To change the template for this generated file go to - * Window>Preferences>Java>Code Generation>Code and Comments - */ -// CHANGELOG -// 2005-05-19 Edward Wang -// - copy this file from icu4jsrc_3_2/src/com/ibm/icu/impl/StringPrepDataReader.java -// - move from package com.ibm.icu.impl to package sun.net.idn -// -package sun.net.idn; - -import java.io.DataInputStream; -import java.io.IOException; -import java.io.InputStream; - -import sun.text.normalizer.ICUBinary; - - -/** - * @author ram - * - * To change the template for this generated type comment go to - * Window>Preferences>Java>Code Generation>Code and Comments - */ -final class StringPrepDataReader implements ICUBinary.Authenticate { - - /** - *

private constructor.

- * @param inputStream ICU uprop.dat file input stream - * @exception IOException throw if data file fails authentication - * @draft 2.1 - */ - public StringPrepDataReader(InputStream inputStream) - throws IOException{ - - unicodeVersion = ICUBinary.readHeader(inputStream, DATA_FORMAT_ID, this); - - - dataInputStream = new DataInputStream(inputStream); - - } - - public void read(byte[] idnaBytes, - char[] mappingTable) - throws IOException{ - - //Read the bytes that make up the idnaTrie - dataInputStream.read(idnaBytes); - - //Read the extra data - for(int i=0;iPreferences>Java>Code Generation>Code and Comments + */ +// CHANGELOG +// 2005-05-19 Edward Wang +// - copy this file from icu4jsrc_3_2/src/com/ibm/icu/impl/StringPrepDataReader.java +// - move from package com.ibm.icu.impl to package sun.net.idn +// +package jdk.internal.icu.impl; + +import java.io.DataInputStream; +import java.io.IOException; +import java.io.InputStream; + +import jdk.internal.icu.impl.ICUBinary; + + +/** + * @author ram + * + * To change the template for this generated type comment go to + * Window>Preferences>Java>Code Generation>Code and Comments + */ +public final class StringPrepDataReader implements ICUBinary.Authenticate { + + /** + *

private constructor.

+ * @param inputStream ICU uprop.dat file input stream + * @exception IOException throw if data file fails authentication + * @draft 2.1 + */ + public StringPrepDataReader(InputStream inputStream) + throws IOException{ + + unicodeVersion = ICUBinary.readHeader(inputStream, DATA_FORMAT_ID, this); + + + dataInputStream = new DataInputStream(inputStream); + + } + + public void read(byte[] idnaBytes, + char[] mappingTable) + throws IOException{ + + //Read the bytes that make up the idnaTrie + dataInputStream.read(idnaBytes); + + //Read the extra data + for(int i=0;iA trie is a kind of compressed, serializable table of values - * associated with Unicode code points (0..0x10ffff).

- *

This class defines the basic structure of a trie and provides methods - * to retrieve the offsets to the actual data.

- *

Data will be the form of an array of basic types, char or int.

- *

The actual data format will have to be specified by the user in the - * inner static interface com.ibm.icu.impl.Trie.DataManipulate.

- *

This trie implementation is optimized for getting offset while walking - * forward through a UTF-16 string. - * Therefore, the simplest and fastest access macros are the - * fromLead() and fromOffsetTrail() methods. - * The fromBMP() method are a little more complicated; they get offsets even - * for lead surrogate codepoints, while the fromLead() method get special - * "folded" offsets for lead surrogate code units if there is relevant data - * associated with them. - * From such a folded offsets, an offset needs to be extracted to supply - * to the fromOffsetTrail() methods. - * To handle such supplementary codepoints, some offset information are kept - * in the data.

- *

Methods in com.ibm.icu.impl.Trie.DataManipulate are called to retrieve - * that offset from the folded value for the lead surrogate unit.

- *

For examples of use, see com.ibm.icu.impl.CharTrie or - * com.ibm.icu.impl.IntTrie.

- * @author synwee - * @see com.ibm.icu.impl.CharTrie - * @see com.ibm.icu.impl.IntTrie - * @since release 2.1, Jan 01 2002 - */ -public abstract class Trie -{ - // public class declaration ---------------------------------------- - - /** - * Character data in com.ibm.impl.Trie have different user-specified format - * for different purposes. - * This interface specifies methods to be implemented in order for - * com.ibm.impl.Trie, to surrogate offset information encapsulated within - * the data. - */ - public static interface DataManipulate - { - /** - * Called by com.ibm.icu.impl.Trie to extract from a lead surrogate's - * data - * the index array offset of the indexes for that lead surrogate. - * @param value data value for a surrogate from the trie, including the - * folding offset - * @return data offset or 0 if there is no data for the lead surrogate - */ - public int getFoldingOffset(int value); - } - - // default implementation - private static class DefaultGetFoldingOffset implements DataManipulate { - public int getFoldingOffset(int value) { - return value; - } - } - - // protected constructor ------------------------------------------- - - /** - * Trie constructor for CharTrie use. - * @param inputStream ICU data file input stream which contains the - * trie - * @param dataManipulate object containing the information to parse the - * trie data - * @throws IOException thrown when input stream does not have the - * right header. - */ - protected Trie(InputStream inputStream, - DataManipulate dataManipulate) throws IOException - { - DataInputStream input = new DataInputStream(inputStream); - // Magic number to authenticate the data. - int signature = input.readInt(); - m_options_ = input.readInt(); - - if (!checkHeader(signature)) { - throw new IllegalArgumentException("ICU data file error: Trie header authentication failed, please check if you have the most updated ICU data file"); - } - - if(dataManipulate != null) { - m_dataManipulate_ = dataManipulate; - } else { - m_dataManipulate_ = new DefaultGetFoldingOffset(); - } - m_isLatin1Linear_ = (m_options_ & - HEADER_OPTIONS_LATIN1_IS_LINEAR_MASK_) != 0; - m_dataOffset_ = input.readInt(); - m_dataLength_ = input.readInt(); - unserialize(inputStream); - } - - // protected data members ------------------------------------------ - - /** - * Lead surrogate code points' index displacement in the index array. - * 
{@code
-     * 0x10000-0xd800=0x2800
-     * 0x2800 >> INDEX_STAGE_1_SHIFT_
-     * }
- */ - protected static final int LEAD_INDEX_OFFSET_ = 0x2800 >> 5; - /** - * Shift size for shifting right the input index. 1..9 - */ - protected static final int INDEX_STAGE_1_SHIFT_ = 5; - /** - * Shift size for shifting left the index array values. - * Increases possible data size with 16-bit index values at the cost - * of compactability. - * This requires blocks of stage 2 data to be aligned by - * DATA_GRANULARITY. - * 0..INDEX_STAGE_1_SHIFT - */ - protected static final int INDEX_STAGE_2_SHIFT_ = 2; - /** - * Number of data values in a stage 2 (data array) block. - */ - protected static final int DATA_BLOCK_LENGTH=1< - * getRawOffset(0, ch); - *

- * will do. Otherwise if it is a supplementary character formed by - * surrogates lead and trail. Then we would have to call getRawOffset() - * with getFoldingIndexOffset(). See getSurrogateOffset(). - * @param offset index offset which ch is to start from - * @param ch index to be used after offset - * @return offset to the data - */ - protected final int getRawOffset(int offset, char ch) - { - return (m_index_[offset + (ch >> INDEX_STAGE_1_SHIFT_)] - << INDEX_STAGE_2_SHIFT_) - + (ch & INDEX_STAGE_3_MASK_); - } - - /** - * Gets the offset to data which the BMP character points to - * Treats a lead surrogate as a normal code point. - * @param ch BMP character - * @return offset to data - */ - protected final int getBMPOffset(char ch) - { - return (ch >= UTF16.LEAD_SURROGATE_MIN_VALUE - && ch <= UTF16.LEAD_SURROGATE_MAX_VALUE) - ? getRawOffset(LEAD_INDEX_OFFSET_, ch) - : getRawOffset(0, ch); - // using a getRawOffset(ch) makes no diff - } - - /** - * Gets the offset to the data which this lead surrogate character points - * to. - * Data at the returned offset may contain folding offset information for - * the next trailing surrogate character. - * @param ch lead surrogate character - * @return offset to data - */ - protected final int getLeadOffset(char ch) - { - return getRawOffset(0, ch); - } - - /** - * Internal trie getter from a code point. - * Could be faster(?) but longer with - * {@code if((c32)<=0xd7ff) { (result)=_TRIE_GET_RAW(trie, data, 0, c32); }} - * Gets the offset to data which the codepoint points to - * @param ch codepoint - * @return offset to data - */ - protected final int getCodePointOffset(int ch) - { - // if ((ch >> 16) == 0) slower - if (ch < 0) { - return -1; - } else if (ch < UTF16.LEAD_SURROGATE_MIN_VALUE) { - // fastpath for the part of the BMP below surrogates (D800) where getRawOffset() works - return getRawOffset(0, (char)ch); - } else if (ch < UTF16.SUPPLEMENTARY_MIN_VALUE) { - // BMP codepoint - return getBMPOffset((char)ch); - } else if (ch <= UCharacter.MAX_VALUE) { - // look at the construction of supplementary characters - // trail forms the ends of it. - return getSurrogateOffset(UTF16.getLeadSurrogate(ch), - (char)(ch & SURROGATE_MASK_)); - } else { - // return -1 if there is an error, in this case we return - return -1; - } - } - - /** - *

Parses the inputstream and creates the trie index with it.

- *

This is overwritten by the child classes. - * @param inputStream input stream containing the trie information - * @exception IOException thrown when data reading fails. - */ - protected void unserialize(InputStream inputStream) throws IOException - { - //indexLength is a multiple of 1024 >> INDEX_STAGE_2_SHIFT_ - m_index_ = new char[m_dataOffset_]; - DataInputStream input = new DataInputStream(inputStream); - for (int i = 0; i < m_dataOffset_; i ++) { - m_index_[i] = input.readChar(); - } - } - - /** - * Determines if this is a 16 bit trie - * @return true if this is a 16 bit trie - */ - protected final boolean isCharTrie() - { - return (m_options_ & HEADER_OPTIONS_DATA_IS_32_BIT_) == 0; - } - - // private data members -------------------------------------------- - - /** - * Latin 1 option mask - */ - protected static final int HEADER_OPTIONS_LATIN1_IS_LINEAR_MASK_ = 0x200; - /** - * Constant number to authenticate the byte block - */ - protected static final int HEADER_SIGNATURE_ = 0x54726965; - /** - * Header option formatting - */ - private static final int HEADER_OPTIONS_SHIFT_MASK_ = 0xF; - protected static final int HEADER_OPTIONS_INDEX_SHIFT_ = 4; - protected static final int HEADER_OPTIONS_DATA_IS_32_BIT_ = 0x100; - - /** - * Flag indicator for Latin quick access data block - */ - private boolean m_isLatin1Linear_; - - /** - *

Trie options field.

- *

options bit field:
- * 9 1 = Latin-1 data is stored linearly at data + DATA_BLOCK_LENGTH
- * 8 0 = 16-bit data, 1=32-bit data
- * 7..4 INDEX_STAGE_1_SHIFT // 0..INDEX_STAGE_2_SHIFT
- * 3..0 INDEX_STAGE_2_SHIFT // 1..9
- */ - private int m_options_; - - // private methods --------------------------------------------------- - - /** - * Authenticates raw data header. - * Checking the header information, signature and options. - * @param signature This contains the options and type of a Trie - * @return true if the header is authenticated valid - */ - private final boolean checkHeader(int signature) - { - // check the signature - // Trie in big-endian US-ASCII (0x54726965). - // Magic number to authenticate the data. - if (signature != HEADER_SIGNATURE_) { - return false; - } - - if ((m_options_ & HEADER_OPTIONS_SHIFT_MASK_) != - INDEX_STAGE_1_SHIFT_ || - ((m_options_ >> HEADER_OPTIONS_INDEX_SHIFT_) & - HEADER_OPTIONS_SHIFT_MASK_) - != INDEX_STAGE_2_SHIFT_) { - return false; - } - return true; - } -} --- /dev/null 2020-01-10 15:57:36.000000000 -0800 +++ new/src/java.base/share/classes/jdk/internal/icu/impl/Trie.java 2020-01-10 15:57:36.000000000 -0800 @@ -0,0 +1,367 @@ +/* + * Copyright (c) 2005, 2020, Oracle and/or its affiliates. All rights reserved. + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This code is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License version 2 only, as + * published by the Free Software Foundation. Oracle designates this + * particular file as subject to the "Classpath" exception as provided + * by Oracle in the LICENSE file that accompanied this code. + * + * This code is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License + * version 2 for more details (a copy is included in the LICENSE file that + * accompanied this code). + * + * You should have received a copy of the GNU General Public License version + * 2 along with this work; if not, write to the Free Software Foundation, + * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. + * + * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA + * or visit www.oracle.com if you need additional information or have any + * questions. + */ + +/* + ****************************************************************************** + * Copyright (C) 1996-2014, International Business Machines Corporation and + * others. All Rights Reserved. + ****************************************************************************** + */ + +package jdk.internal.icu.impl; + +import jdk.internal.icu.lang.UCharacter; +import jdk.internal.icu.text.UTF16; + +import java.io.DataInputStream; +import java.io.InputStream; +import java.io.IOException; + +/** + *

A trie is a kind of compressed, serializable table of values + * associated with Unicode code points (0..0x10ffff).

+ *

This class defines the basic structure of a trie and provides methods + * to retrieve the offsets to the actual data.

+ *

Data will be the form of an array of basic types, char or int.

+ *

The actual data format will have to be specified by the user in the + * inner static interface com.ibm.icu.impl.Trie.DataManipulate.

+ *

This trie implementation is optimized for getting offset while walking + * forward through a UTF-16 string. + * Therefore, the simplest and fastest access macros are the + * fromLead() and fromOffsetTrail() methods. + * The fromBMP() method are a little more complicated; they get offsets even + * for lead surrogate codepoints, while the fromLead() method get special + * "folded" offsets for lead surrogate code units if there is relevant data + * associated with them. + * From such a folded offsets, an offset needs to be extracted to supply + * to the fromOffsetTrail() methods. + * To handle such supplementary codepoints, some offset information are kept + * in the data.

+ *

Methods in com.ibm.icu.impl.Trie.DataManipulate are called to retrieve + * that offset from the folded value for the lead surrogate unit.

+ *

For examples of use, see com.ibm.icu.impl.CharTrie or + * com.ibm.icu.impl.IntTrie.

+ * @author synwee + * @see com.ibm.icu.impl.CharTrie + * @see com.ibm.icu.impl.IntTrie + * @since release 2.1, Jan 01 2002 + */ +public abstract class Trie +{ + // public class declaration ---------------------------------------- + + /** + * Character data in com.ibm.impl.Trie have different user-specified format + * for different purposes. + * This interface specifies methods to be implemented in order for + * com.ibm.impl.Trie, to surrogate offset information encapsulated within + * the data. + */ + public static interface DataManipulate + { + /** + * Called by com.ibm.icu.impl.Trie to extract from a lead surrogate's + * data + * the index array offset of the indexes for that lead surrogate. + * @param value data value for a surrogate from the trie, including the + * folding offset + * @return data offset or 0 if there is no data for the lead surrogate + */ + public int getFoldingOffset(int value); + } + + // default implementation + private static class DefaultGetFoldingOffset implements DataManipulate { + public int getFoldingOffset(int value) { + return value; + } + } + + // protected constructor ------------------------------------------- + + /** + * Trie constructor for CharTrie use. + * @param inputStream ICU data file input stream which contains the + * trie + * @param dataManipulate object containing the information to parse the + * trie data + * @throws IOException thrown when input stream does not have the + * right header. + */ + protected Trie(InputStream inputStream, + DataManipulate dataManipulate) throws IOException + { + DataInputStream input = new DataInputStream(inputStream); + // Magic number to authenticate the data. + int signature = input.readInt(); + m_options_ = input.readInt(); + + if (!checkHeader(signature)) { + throw new IllegalArgumentException("ICU data file error: Trie header authentication failed, please check if you have the most updated ICU data file"); + } + + if(dataManipulate != null) { + m_dataManipulate_ = dataManipulate; + } else { + m_dataManipulate_ = new DefaultGetFoldingOffset(); + } + m_isLatin1Linear_ = (m_options_ & + HEADER_OPTIONS_LATIN1_IS_LINEAR_MASK_) != 0; + m_dataOffset_ = input.readInt(); + m_dataLength_ = input.readInt(); + unserialize(inputStream); + } + + // protected data members ------------------------------------------ + + /** + * Lead surrogate code points' index displacement in the index array. + * 
{@code
+     * 0x10000-0xd800=0x2800
+     * 0x2800 >> INDEX_STAGE_1_SHIFT_
+     * }
+ */ + protected static final int LEAD_INDEX_OFFSET_ = 0x2800 >> 5; + /** + * Shift size for shifting right the input index. 1..9 + */ + protected static final int INDEX_STAGE_1_SHIFT_ = 5; + /** + * Shift size for shifting left the index array values. + * Increases possible data size with 16-bit index values at the cost + * of compactability. + * This requires blocks of stage 2 data to be aligned by + * DATA_GRANULARITY. + * 0..INDEX_STAGE_1_SHIFT + */ + protected static final int INDEX_STAGE_2_SHIFT_ = 2; + /** + * Number of data values in a stage 2 (data array) block. + */ + protected static final int DATA_BLOCK_LENGTH=1< + * getRawOffset(0, ch); + *

+ * will do. Otherwise if it is a supplementary character formed by + * surrogates lead and trail. Then we would have to call getRawOffset() + * with getFoldingIndexOffset(). See getSurrogateOffset(). + * @param offset index offset which ch is to start from + * @param ch index to be used after offset + * @return offset to the data + */ + protected final int getRawOffset(int offset, char ch) + { + return (m_index_[offset + (ch >> INDEX_STAGE_1_SHIFT_)] + << INDEX_STAGE_2_SHIFT_) + + (ch & INDEX_STAGE_3_MASK_); + } + + /** + * Gets the offset to data which the BMP character points to + * Treats a lead surrogate as a normal code point. + * @param ch BMP character + * @return offset to data + */ + protected final int getBMPOffset(char ch) + { + return (ch >= UTF16.LEAD_SURROGATE_MIN_VALUE + && ch <= UTF16.LEAD_SURROGATE_MAX_VALUE) + ? getRawOffset(LEAD_INDEX_OFFSET_, ch) + : getRawOffset(0, ch); + // using a getRawOffset(ch) makes no diff + } + + /** + * Gets the offset to the data which this lead surrogate character points + * to. + * Data at the returned offset may contain folding offset information for + * the next trailing surrogate character. + * @param ch lead surrogate character + * @return offset to data + */ + protected final int getLeadOffset(char ch) + { + return getRawOffset(0, ch); + } + + /** + * Internal trie getter from a code point. + * Could be faster(?) but longer with + * {@code if((c32)<=0xd7ff) { (result)=_TRIE_GET_RAW(trie, data, 0, c32); }} + * Gets the offset to data which the codepoint points to + * @param ch codepoint + * @return offset to data + */ + protected final int getCodePointOffset(int ch) + { + // if ((ch >> 16) == 0) slower + if (ch < 0) { + return -1; + } else if (ch < UTF16.LEAD_SURROGATE_MIN_VALUE) { + // fastpath for the part of the BMP below surrogates (D800) where getRawOffset() works + return getRawOffset(0, (char)ch); + } else if (ch < UTF16.SUPPLEMENTARY_MIN_VALUE) { + // BMP codepoint + return getBMPOffset((char)ch); + } else if (ch <= UCharacter.MAX_VALUE) { + // look at the construction of supplementary characters + // trail forms the ends of it. + return getSurrogateOffset(UTF16.getLeadSurrogate(ch), + (char)(ch & SURROGATE_MASK_)); + } else { + // return -1 if there is an error, in this case we return + return -1; + } + } + + /** + *

Parses the inputstream and creates the trie index with it.

+ *

This is overwritten by the child classes. + * @param inputStream input stream containing the trie information + * @exception IOException thrown when data reading fails. + */ + protected void unserialize(InputStream inputStream) throws IOException + { + //indexLength is a multiple of 1024 >> INDEX_STAGE_2_SHIFT_ + m_index_ = new char[m_dataOffset_]; + DataInputStream input = new DataInputStream(inputStream); + for (int i = 0; i < m_dataOffset_; i ++) { + m_index_[i] = input.readChar(); + } + } + + /** + * Determines if this is a 16 bit trie + * @return true if this is a 16 bit trie + */ + protected final boolean isCharTrie() + { + return (m_options_ & HEADER_OPTIONS_DATA_IS_32_BIT_) == 0; + } + + // private data members -------------------------------------------- + + /** + * Latin 1 option mask + */ + protected static final int HEADER_OPTIONS_LATIN1_IS_LINEAR_MASK_ = 0x200; + /** + * Constant number to authenticate the byte block + */ + protected static final int HEADER_SIGNATURE_ = 0x54726965; + /** + * Header option formatting + */ + private static final int HEADER_OPTIONS_SHIFT_MASK_ = 0xF; + protected static final int HEADER_OPTIONS_INDEX_SHIFT_ = 4; + protected static final int HEADER_OPTIONS_DATA_IS_32_BIT_ = 0x100; + + /** + * Flag indicator for Latin quick access data block + */ + private boolean m_isLatin1Linear_; + + /** + *

Trie options field.

+ *

options bit field:
+ * 9 1 = Latin-1 data is stored linearly at data + DATA_BLOCK_LENGTH
+ * 8 0 = 16-bit data, 1=32-bit data
+ * 7..4 INDEX_STAGE_1_SHIFT // 0..INDEX_STAGE_2_SHIFT
+ * 3..0 INDEX_STAGE_2_SHIFT // 1..9
+ */ + private int m_options_; + + // private methods --------------------------------------------------- + + /** + * Authenticates raw data header. + * Checking the header information, signature and options. + * @param signature This contains the options and type of a Trie + * @return true if the header is authenticated valid + */ + private final boolean checkHeader(int signature) + { + // check the signature + // Trie in big-endian US-ASCII (0x54726965). + // Magic number to authenticate the data. + if (signature != HEADER_SIGNATURE_) { + return false; + } + + if ((m_options_ & HEADER_OPTIONS_SHIFT_MASK_) != + INDEX_STAGE_1_SHIFT_ || + ((m_options_ >> HEADER_OPTIONS_INDEX_SHIFT_) & + HEADER_OPTIONS_SHIFT_MASK_) + != INDEX_STAGE_2_SHIFT_) { + return false; + } + return true; + } +} --- old/src/java.base/share/classes/sun/text/normalizer/Trie2.java 2020-01-10 15:57:37.000000000 -0800 +++ /dev/null 2020-01-10 15:57:37.000000000 -0800 @@ -1,655 +0,0 @@ -/* - * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved. - * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. - * - * This code is free software; you can redistribute it and/or modify it - * under the terms of the GNU General Public License version 2 only, as - * published by the Free Software Foundation. Oracle designates this - * particular file as subject to the "Classpath" exception as provided - * by Oracle in the LICENSE file that accompanied this code. - * - * This code is distributed in the hope that it will be useful, but WITHOUT - * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or - * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License - * version 2 for more details (a copy is included in the LICENSE file that - * accompanied this code). - * - * You should have received a copy of the GNU General Public License version - * 2 along with this work; if not, write to the Free Software Foundation, - * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. - * - * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA - * or visit www.oracle.com if you need additional information or have any - * questions. - */ - -/* - ******************************************************************************* - * Copyright (C) 2009-2014, International Business Machines Corporation and - * others. All Rights Reserved. - ******************************************************************************* - */ - -package sun.text.normalizer; - -import java.io.IOException; -import java.nio.ByteBuffer; -import java.nio.ByteOrder; -import java.util.Iterator; -import java.util.NoSuchElementException; - - -/** - * This is the interface and common implementation of a Unicode Trie2. - * It is a kind of compressed table that maps from Unicode code points (0..0x10ffff) - * to 16- or 32-bit integer values. It works best when there are ranges of - * characters with the same value, which is generally the case with Unicode - * character properties. - * - * This is the second common version of a Unicode trie (hence the name Trie2). - * - */ -abstract class Trie2 implements Iterable { - - /** - * Create a Trie2 from its serialized form. Inverse of utrie2_serialize(). - * - * Reads from the current position and leaves the buffer after the end of the trie. - * - * The serialized format is identical between ICU4C and ICU4J, so this function - * will work with serialized Trie2s from either. - * - * The actual type of the returned Trie2 will be either Trie2_16 or Trie2_32, depending - * on the width of the data. - * - * To obtain the width of the Trie2, check the actual class type of the returned Trie2. - * Or use the createFromSerialized() function of Trie2_16 or Trie2_32, which will - * return only Tries of their specific type/size. - * - * The serialized Trie2 on the stream may be in either little or big endian byte order. - * This allows using serialized Tries from ICU4C without needing to consider the - * byte order of the system that created them. - * - * @param bytes a byte buffer to the serialized form of a UTrie2. - * @return An unserialized Trie2, ready for use. - * @throws IllegalArgumentException if the stream does not contain a serialized Trie2. - * @throws IOException if a read error occurs in the buffer. - * - */ - public static Trie2 createFromSerialized(ByteBuffer bytes) throws IOException { - // From ICU4C utrie2_impl.h - // * Trie2 data structure in serialized form: - // * - // * UTrie2Header header; - // * uint16_t index[header.index2Length]; - // * uint16_t data[header.shiftedDataLength<<2]; -- or uint32_t data[...] - // * @internal - // */ - // typedef struct UTrie2Header { - // /** "Tri2" in big-endian US-ASCII (0x54726932) */ - // uint32_t signature; - - // /** - // * options bit field: - // * 15.. 4 reserved (0) - // * 3.. 0 UTrie2ValueBits valueBits - // */ - // uint16_t options; - // - // /** UTRIE2_INDEX_1_OFFSET..UTRIE2_MAX_INDEX_LENGTH */ - // uint16_t indexLength; - // - // /** (UTRIE2_DATA_START_OFFSET..UTRIE2_MAX_DATA_LENGTH)>>UTRIE2_INDEX_SHIFT */ - // uint16_t shiftedDataLength; - // - // /** Null index and data blocks, not shifted. */ - // uint16_t index2NullOffset, dataNullOffset; - // - // /** - // * First code point of the single-value range ending with U+10ffff, - // * rounded up and then shifted right by UTRIE2_SHIFT_1. - // */ - // uint16_t shiftedHighStart; - // } UTrie2Header; - - ByteOrder outerByteOrder = bytes.order(); - try { - UTrie2Header header = new UTrie2Header(); - - /* check the signature */ - header.signature = bytes.getInt(); - switch (header.signature) { - case 0x54726932: - // The buffer is already set to the trie data byte order. - break; - case 0x32697254: - // Temporarily reverse the byte order. - boolean isBigEndian = outerByteOrder == ByteOrder.BIG_ENDIAN; - bytes.order(isBigEndian ? ByteOrder.LITTLE_ENDIAN : ByteOrder.BIG_ENDIAN); - header.signature = 0x54726932; - break; - default: - throw new IllegalArgumentException("Buffer does not contain a serialized UTrie2"); - } - - header.options = bytes.getChar(); - header.indexLength = bytes.getChar(); - header.shiftedDataLength = bytes.getChar(); - header.index2NullOffset = bytes.getChar(); - header.dataNullOffset = bytes.getChar(); - header.shiftedHighStart = bytes.getChar(); - - if ((header.options & UTRIE2_OPTIONS_VALUE_BITS_MASK) != 0) { - throw new IllegalArgumentException("UTrie2 serialized format error."); - } - - Trie2 This; - This = new Trie2_16(); - This.header = header; - - /* get the length values and offsets */ - This.indexLength = header.indexLength; - This.dataLength = header.shiftedDataLength << UTRIE2_INDEX_SHIFT; - This.index2NullOffset = header.index2NullOffset; - This.dataNullOffset = header.dataNullOffset; - This.highStart = header.shiftedHighStart << UTRIE2_SHIFT_1; - This.highValueIndex = This.dataLength - UTRIE2_DATA_GRANULARITY; - This.highValueIndex += This.indexLength; - - // Allocate the Trie2 index array. If the data width is 16 bits, the array also - // includes the space for the data. - - int indexArraySize = This.indexLength; - indexArraySize += This.dataLength; - This.index = new char[indexArraySize]; - - /* Read in the index */ - int i; - for (i=0; i iterator() { - return iterator(defaultValueMapper); - } - - private static ValueMapper defaultValueMapper = new ValueMapper() { - public int map(int in) { - return in; - } - }; - - /** - * Create an iterator over the value ranges from this Trie2. - * Values from the Trie2 are passed through a caller-supplied remapping function, - * and it is the remapped values that determine the ranges that - * will be produced by the iterator. - * - * - * @param mapper provides a function to remap values obtained from the Trie2. - * @return an Iterator - */ - public Iterator iterator(ValueMapper mapper) { - return new Trie2Iterator(mapper); - } - - /** - * When iterating over the contents of a Trie2, an instance of TrieValueMapper may - * be used to remap the values from the Trie2. The remapped values will be used - * both in determining the ranges of codepoints and as the value to be returned - * for each range. - * - * Example of use, with an anonymous subclass of TrieValueMapper: - * - * - * ValueMapper m = new ValueMapper() { - * int map(int in) {return in & 0x1f;}; - * } - * for (Iterator iter = trie.iterator(m); i.hasNext(); ) { - * Trie2EnumRange r = i.next(); - * ... // Do something with the range r. - * } - * - */ - public interface ValueMapper { - public int map(int originalVal); - } - - //-------------------------------------------------------------------------------- - // - // Below this point are internal implementation items. No further public API. - // - //-------------------------------------------------------------------------------- - - /** - * Trie2 data structure in serialized form: - * - * UTrie2Header header; - * uint16_t index[header.index2Length]; - * uint16_t data[header.shiftedDataLength<<2]; -- or uint32_t data[...] - * - * For Java, this is read from the stream into an instance of UTrie2Header. - * (The C version just places a struct over the raw serialized data.) - * - * @internal - */ - static class UTrie2Header { - /** "Tri2" in big-endian US-ASCII (0x54726932) */ - int signature; - - /** - * options bit field (uint16_t): - * 15.. 4 reserved (0) - * 3.. 0 UTrie2ValueBits valueBits - */ - int options; - - /** UTRIE2_INDEX_1_OFFSET..UTRIE2_MAX_INDEX_LENGTH (uint16_t) */ - int indexLength; - - /** (UTRIE2_DATA_START_OFFSET..UTRIE2_MAX_DATA_LENGTH)>>UTRIE2_INDEX_SHIFT (uint16_t) */ - int shiftedDataLength; - - /** Null index and data blocks, not shifted. (uint16_t) */ - int index2NullOffset, dataNullOffset; - - /** - * First code point of the single-value range ending with U+10ffff, - * rounded up and then shifted right by UTRIE2_SHIFT_1. (uint16_t) - */ - int shiftedHighStart; - } - - // - // Data members of UTrie2. - // - UTrie2Header header; - char index[]; // Index array. Includes data for 16 bit Tries. - int data16; // Offset to data portion of the index array, if 16 bit data. - // zero if 32 bit data. - int data32[]; // NULL if 16b data is used via index - - int indexLength; - int dataLength; - int index2NullOffset; // 0xffff if there is no dedicated index-2 null block - int initialValue; - - /** Value returned for out-of-range code points and illegal UTF-8. */ - int errorValue; - - /* Start of the last range which ends at U+10ffff, and its value. */ - int highStart; - int highValueIndex; - - int dataNullOffset; - - /** - * Trie2 constants, defining shift widths, index array lengths, etc. - * - * These are needed for the runtime macros but users can treat these as - * implementation details and skip to the actual public API further below. - */ - - static final int UTRIE2_OPTIONS_VALUE_BITS_MASK=0x000f; - - - /** Shift size for getting the index-1 table offset. */ - static final int UTRIE2_SHIFT_1=6+5; - - /** Shift size for getting the index-2 table offset. */ - static final int UTRIE2_SHIFT_2=5; - - /** - * Difference between the two shift sizes, - * for getting an index-1 offset from an index-2 offset. 6=11-5 - */ - static final int UTRIE2_SHIFT_1_2=UTRIE2_SHIFT_1-UTRIE2_SHIFT_2; - - /** - * Number of index-1 entries for the BMP. 32=0x20 - * This part of the index-1 table is omitted from the serialized form. - */ - static final int UTRIE2_OMITTED_BMP_INDEX_1_LENGTH=0x10000>>UTRIE2_SHIFT_1; - - /** Number of entries in an index-2 block. 64=0x40 */ - static final int UTRIE2_INDEX_2_BLOCK_LENGTH=1<>UTRIE2_SHIFT_2. (There are 1024=0x400 lead surrogates.) - */ - static final int UTRIE2_LSCP_INDEX_2_OFFSET=0x10000>>UTRIE2_SHIFT_2; - static final int UTRIE2_LSCP_INDEX_2_LENGTH=0x400>>UTRIE2_SHIFT_2; - - /** Count the lengths of both BMP pieces. 2080=0x820 */ - static final int UTRIE2_INDEX_2_BMP_LENGTH=UTRIE2_LSCP_INDEX_2_OFFSET+UTRIE2_LSCP_INDEX_2_LENGTH; - - /** - * The 2-byte UTF-8 version of the index-2 table follows at offset 2080=0x820. - * Length 32=0x20 for lead bytes C0..DF, regardless of UTRIE2_SHIFT_2. - */ - static final int UTRIE2_UTF8_2B_INDEX_2_OFFSET=UTRIE2_INDEX_2_BMP_LENGTH; - static final int UTRIE2_UTF8_2B_INDEX_2_LENGTH=0x800>>6; /* U+0800 is the first code point after 2-byte UTF-8 */ - - /** - * The index-1 table, only used for supplementary code points, at offset 2112=0x840. - * Variable length, for code points up to highStart, where the last single-value range starts. - * Maximum length 512=0x200=0x100000>>UTRIE2_SHIFT_1. - * (For 0x100000 supplementary code points U+10000..U+10ffff.) - * - * The part of the index-2 table for supplementary code points starts - * after this index-1 table. - * - * Both the index-1 table and the following part of the index-2 table - * are omitted completely if there is only BMP data. - */ - static final int UTRIE2_INDEX_1_OFFSET=UTRIE2_UTF8_2B_INDEX_2_OFFSET+UTRIE2_UTF8_2B_INDEX_2_LENGTH; - - /** - * The illegal-UTF-8 data block follows the ASCII block, at offset 128=0x80. - * Used with linear access for single bytes 0..0xbf for simple error handling. - * Length 64=0x40, not UTRIE2_DATA_BLOCK_LENGTH. - */ - static final int UTRIE2_BAD_UTF8_DATA_OFFSET=0x80; - - /** - * Implementation class for an iterator over a Trie2. - * - * Iteration over a Trie2 first returns all of the ranges that are indexed by code points, - * then returns the special alternate values for the lead surrogates - * - * @internal - */ - class Trie2Iterator implements Iterator { - - // The normal constructor that configures the iterator to cover the complete - // contents of the Trie2 - Trie2Iterator(ValueMapper vm) { - mapper = vm; - nextStart = 0; - limitCP = 0x110000; - doLeadSurrogates = true; - } - - /** - * The main next() function for Trie2 iterators - * - */ - public Range next() { - if (!hasNext()) { - throw new NoSuchElementException(); - } - if (nextStart >= limitCP) { - // Switch over from iterating normal code point values to - // doing the alternate lead-surrogate values. - doingCodePoints = false; - nextStart = 0xd800; - } - int endOfRange = 0; - int val = 0; - int mappedVal = 0; - - if (doingCodePoints) { - // Iteration over code point values. - val = get(nextStart); - mappedVal = mapper.map(val); - endOfRange = rangeEnd(nextStart, limitCP, val); - // Loop once for each range in the Trie2 with the same raw (unmapped) value. - // Loop continues so long as the mapped values are the same. - for (;;) { - if (endOfRange >= limitCP-1) { - break; - } - val = get(endOfRange+1); - if (mapper.map(val) != mappedVal) { - break; - } - endOfRange = rangeEnd(endOfRange+1, limitCP, val); - } - } else { - // Iteration over the alternate lead surrogate values. - val = getFromU16SingleLead((char)nextStart); - mappedVal = mapper.map(val); - endOfRange = rangeEndLS((char)nextStart); - // Loop once for each range in the Trie2 with the same raw (unmapped) value. - // Loop continues so long as the mapped values are the same. - for (;;) { - if (endOfRange >= 0xdbff) { - break; - } - val = getFromU16SingleLead((char)(endOfRange+1)); - if (mapper.map(val) != mappedVal) { - break; - } - endOfRange = rangeEndLS((char)(endOfRange+1)); - } - } - returnValue.startCodePoint = nextStart; - returnValue.endCodePoint = endOfRange; - returnValue.value = mappedVal; - returnValue.leadSurrogate = !doingCodePoints; - nextStart = endOfRange+1; - return returnValue; - } - - /** - * - */ - public boolean hasNext() { - return doingCodePoints && (doLeadSurrogates || nextStart < limitCP) || nextStart < 0xdc00; - } - - private int rangeEndLS(char startingLS) { - if (startingLS >= 0xdbff) { - return 0xdbff; - } - - int c; - int val = getFromU16SingleLead(startingLS); - for (c = startingLS+1; c <= 0x0dbff; c++) { - if (getFromU16SingleLead((char)c) != val) { - break; - } - } - return c-1; - } - - // - // Iteration State Variables - // - private ValueMapper mapper; - private Range returnValue = new Range(); - // The starting code point for the next range to be returned. - private int nextStart; - // The upper limit for the last normal range to be returned. Normally 0x110000, but - // may be lower when iterating over the code points for a single lead surrogate. - private int limitCP; - - // True while iterating over the Trie2 values for code points. - // False while iterating over the alternate values for lead surrogates. - private boolean doingCodePoints = true; - - // True if the iterator should iterate the special values for lead surrogates in - // addition to the normal values for code points. - private boolean doLeadSurrogates = true; - } - - /** - * Find the last character in a contiguous range of characters with the - * same Trie2 value as the input character. - * - * @param c The character to begin with. - * @return The last contiguous character with the same value. - */ - int rangeEnd(int start, int limitp, int val) { - int c; - int limit = Math.min(highStart, limitp); - - for (c = start+1; c < limit; c++) { - if (get(c) != val) { - break; - } - } - if (c >= highStart) { - c = limitp; - } - return c - 1; - } - - - // - // Hashing implementation functions. FNV hash. Respected public domain algorithm. - // - private static int initHash() { - return 0x811c9DC5; // unsigned 2166136261 - } - - private static int hashByte(int h, int b) { - h = h * 16777619; - h = h ^ b; - return h; - } - - private static int hashUChar32(int h, int c) { - h = Trie2.hashByte(h, c & 255); - h = Trie2.hashByte(h, (c>>8) & 255); - h = Trie2.hashByte(h, c>>16); - return h; - } - - private static int hashInt(int h, int i) { - h = Trie2.hashByte(h, i & 255); - h = Trie2.hashByte(h, (i>>8) & 255); - h = Trie2.hashByte(h, (i>>16) & 255); - h = Trie2.hashByte(h, (i>>24) & 255); - return h; - } - -} --- /dev/null 2020-01-10 15:57:37.000000000 -0800 +++ new/src/java.base/share/classes/jdk/internal/icu/impl/Trie2.java 2020-01-10 15:57:37.000000000 -0800 @@ -0,0 +1,655 @@ +/* + * Copyright (c) 2015, 2020, Oracle and/or its affiliates. All rights reserved. + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This code is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License version 2 only, as + * published by the Free Software Foundation. Oracle designates this + * particular file as subject to the "Classpath" exception as provided + * by Oracle in the LICENSE file that accompanied this code. + * + * This code is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License + * version 2 for more details (a copy is included in the LICENSE file that + * accompanied this code). + * + * You should have received a copy of the GNU General Public License version + * 2 along with this work; if not, write to the Free Software Foundation, + * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. + * + * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA + * or visit www.oracle.com if you need additional information or have any + * questions. + */ + +/* + ******************************************************************************* + * Copyright (C) 2009-2014, International Business Machines Corporation and + * others. All Rights Reserved. + ******************************************************************************* + */ + +package jdk.internal.icu.impl; + +import java.io.IOException; +import java.nio.ByteBuffer; +import java.nio.ByteOrder; +import java.util.Iterator; +import java.util.NoSuchElementException; + + +/** + * This is the interface and common implementation of a Unicode Trie2. + * It is a kind of compressed table that maps from Unicode code points (0..0x10ffff) + * to 16- or 32-bit integer values. It works best when there are ranges of + * characters with the same value, which is generally the case with Unicode + * character properties. + * + * This is the second common version of a Unicode trie (hence the name Trie2). + * + */ +abstract class Trie2 implements Iterable { + + /** + * Create a Trie2 from its serialized form. Inverse of utrie2_serialize(). + * + * Reads from the current position and leaves the buffer after the end of the trie. + * + * The serialized format is identical between ICU4C and ICU4J, so this function + * will work with serialized Trie2s from either. + * + * The actual type of the returned Trie2 will be either Trie2_16 or Trie2_32, depending + * on the width of the data. + * + * To obtain the width of the Trie2, check the actual class type of the returned Trie2. + * Or use the createFromSerialized() function of Trie2_16 or Trie2_32, which will + * return only Tries of their specific type/size. + * + * The serialized Trie2 on the stream may be in either little or big endian byte order. + * This allows using serialized Tries from ICU4C without needing to consider the + * byte order of the system that created them. + * + * @param bytes a byte buffer to the serialized form of a UTrie2. + * @return An unserialized Trie2, ready for use. + * @throws IllegalArgumentException if the stream does not contain a serialized Trie2. + * @throws IOException if a read error occurs in the buffer. + * + */ + public static Trie2 createFromSerialized(ByteBuffer bytes) throws IOException { + // From ICU4C utrie2_impl.h + // * Trie2 data structure in serialized form: + // * + // * UTrie2Header header; + // * uint16_t index[header.index2Length]; + // * uint16_t data[header.shiftedDataLength<<2]; -- or uint32_t data[...] + // * @internal + // */ + // typedef struct UTrie2Header { + // /** "Tri2" in big-endian US-ASCII (0x54726932) */ + // uint32_t signature; + + // /** + // * options bit field: + // * 15.. 4 reserved (0) + // * 3.. 0 UTrie2ValueBits valueBits + // */ + // uint16_t options; + // + // /** UTRIE2_INDEX_1_OFFSET..UTRIE2_MAX_INDEX_LENGTH */ + // uint16_t indexLength; + // + // /** (UTRIE2_DATA_START_OFFSET..UTRIE2_MAX_DATA_LENGTH)>>UTRIE2_INDEX_SHIFT */ + // uint16_t shiftedDataLength; + // + // /** Null index and data blocks, not shifted. */ + // uint16_t index2NullOffset, dataNullOffset; + // + // /** + // * First code point of the single-value range ending with U+10ffff, + // * rounded up and then shifted right by UTRIE2_SHIFT_1. + // */ + // uint16_t shiftedHighStart; + // } UTrie2Header; + + ByteOrder outerByteOrder = bytes.order(); + try { + UTrie2Header header = new UTrie2Header(); + + /* check the signature */ + header.signature = bytes.getInt(); + switch (header.signature) { + case 0x54726932: + // The buffer is already set to the trie data byte order. + break; + case 0x32697254: + // Temporarily reverse the byte order. + boolean isBigEndian = outerByteOrder == ByteOrder.BIG_ENDIAN; + bytes.order(isBigEndian ? ByteOrder.LITTLE_ENDIAN : ByteOrder.BIG_ENDIAN); + header.signature = 0x54726932; + break; + default: + throw new IllegalArgumentException("Buffer does not contain a serialized UTrie2"); + } + + header.options = bytes.getChar(); + header.indexLength = bytes.getChar(); + header.shiftedDataLength = bytes.getChar(); + header.index2NullOffset = bytes.getChar(); + header.dataNullOffset = bytes.getChar(); + header.shiftedHighStart = bytes.getChar(); + + if ((header.options & UTRIE2_OPTIONS_VALUE_BITS_MASK) != 0) { + throw new IllegalArgumentException("UTrie2 serialized format error."); + } + + Trie2 This; + This = new Trie2_16(); + This.header = header; + + /* get the length values and offsets */ + This.indexLength = header.indexLength; + This.dataLength = header.shiftedDataLength << UTRIE2_INDEX_SHIFT; + This.index2NullOffset = header.index2NullOffset; + This.dataNullOffset = header.dataNullOffset; + This.highStart = header.shiftedHighStart << UTRIE2_SHIFT_1; + This.highValueIndex = This.dataLength - UTRIE2_DATA_GRANULARITY; + This.highValueIndex += This.indexLength; + + // Allocate the Trie2 index array. If the data width is 16 bits, the array also + // includes the space for the data. + + int indexArraySize = This.indexLength; + indexArraySize += This.dataLength; + This.index = new char[indexArraySize]; + + /* Read in the index */ + int i; + for (i=0; i iterator() { + return iterator(defaultValueMapper); + } + + private static ValueMapper defaultValueMapper = new ValueMapper() { + public int map(int in) { + return in; + } + }; + + /** + * Create an iterator over the value ranges from this Trie2. + * Values from the Trie2 are passed through a caller-supplied remapping function, + * and it is the remapped values that determine the ranges that + * will be produced by the iterator. + * + * + * @param mapper provides a function to remap values obtained from the Trie2. + * @return an Iterator + */ + public Iterator iterator(ValueMapper mapper) { + return new Trie2Iterator(mapper); + } + + /** + * When iterating over the contents of a Trie2, an instance of TrieValueMapper may + * be used to remap the values from the Trie2. The remapped values will be used + * both in determining the ranges of codepoints and as the value to be returned + * for each range. + * + * Example of use, with an anonymous subclass of TrieValueMapper: + * + * + * ValueMapper m = new ValueMapper() { + * int map(int in) {return in & 0x1f;}; + * } + * for (Iterator iter = trie.iterator(m); i.hasNext(); ) { + * Trie2EnumRange r = i.next(); + * ... // Do something with the range r. + * } + * + */ + public interface ValueMapper { + public int map(int originalVal); + } + + //-------------------------------------------------------------------------------- + // + // Below this point are internal implementation items. No further public API. + // + //-------------------------------------------------------------------------------- + + /** + * Trie2 data structure in serialized form: + * + * UTrie2Header header; + * uint16_t index[header.index2Length]; + * uint16_t data[header.shiftedDataLength<<2]; -- or uint32_t data[...] + * + * For Java, this is read from the stream into an instance of UTrie2Header. + * (The C version just places a struct over the raw serialized data.) + * + * @internal + */ + static class UTrie2Header { + /** "Tri2" in big-endian US-ASCII (0x54726932) */ + int signature; + + /** + * options bit field (uint16_t): + * 15.. 4 reserved (0) + * 3.. 0 UTrie2ValueBits valueBits + */ + int options; + + /** UTRIE2_INDEX_1_OFFSET..UTRIE2_MAX_INDEX_LENGTH (uint16_t) */ + int indexLength; + + /** (UTRIE2_DATA_START_OFFSET..UTRIE2_MAX_DATA_LENGTH)>>UTRIE2_INDEX_SHIFT (uint16_t) */ + int shiftedDataLength; + + /** Null index and data blocks, not shifted. (uint16_t) */ + int index2NullOffset, dataNullOffset; + + /** + * First code point of the single-value range ending with U+10ffff, + * rounded up and then shifted right by UTRIE2_SHIFT_1. (uint16_t) + */ + int shiftedHighStart; + } + + // + // Data members of UTrie2. + // + UTrie2Header header; + char index[]; // Index array. Includes data for 16 bit Tries. + int data16; // Offset to data portion of the index array, if 16 bit data. + // zero if 32 bit data. + int data32[]; // NULL if 16b data is used via index + + int indexLength; + int dataLength; + int index2NullOffset; // 0xffff if there is no dedicated index-2 null block + int initialValue; + + /** Value returned for out-of-range code points and illegal UTF-8. */ + int errorValue; + + /* Start of the last range which ends at U+10ffff, and its value. */ + int highStart; + int highValueIndex; + + int dataNullOffset; + + /** + * Trie2 constants, defining shift widths, index array lengths, etc. + * + * These are needed for the runtime macros but users can treat these as + * implementation details and skip to the actual public API further below. + */ + + static final int UTRIE2_OPTIONS_VALUE_BITS_MASK=0x000f; + + + /** Shift size for getting the index-1 table offset. */ + static final int UTRIE2_SHIFT_1=6+5; + + /** Shift size for getting the index-2 table offset. */ + static final int UTRIE2_SHIFT_2=5; + + /** + * Difference between the two shift sizes, + * for getting an index-1 offset from an index-2 offset. 6=11-5 + */ + static final int UTRIE2_SHIFT_1_2=UTRIE2_SHIFT_1-UTRIE2_SHIFT_2; + + /** + * Number of index-1 entries for the BMP. 32=0x20 + * This part of the index-1 table is omitted from the serialized form. + */ + static final int UTRIE2_OMITTED_BMP_INDEX_1_LENGTH=0x10000>>UTRIE2_SHIFT_1; + + /** Number of entries in an index-2 block. 64=0x40 */ + static final int UTRIE2_INDEX_2_BLOCK_LENGTH=1<>UTRIE2_SHIFT_2. (There are 1024=0x400 lead surrogates.) + */ + static final int UTRIE2_LSCP_INDEX_2_OFFSET=0x10000>>UTRIE2_SHIFT_2; + static final int UTRIE2_LSCP_INDEX_2_LENGTH=0x400>>UTRIE2_SHIFT_2; + + /** Count the lengths of both BMP pieces. 2080=0x820 */ + static final int UTRIE2_INDEX_2_BMP_LENGTH=UTRIE2_LSCP_INDEX_2_OFFSET+UTRIE2_LSCP_INDEX_2_LENGTH; + + /** + * The 2-byte UTF-8 version of the index-2 table follows at offset 2080=0x820. + * Length 32=0x20 for lead bytes C0..DF, regardless of UTRIE2_SHIFT_2. + */ + static final int UTRIE2_UTF8_2B_INDEX_2_OFFSET=UTRIE2_INDEX_2_BMP_LENGTH; + static final int UTRIE2_UTF8_2B_INDEX_2_LENGTH=0x800>>6; /* U+0800 is the first code point after 2-byte UTF-8 */ + + /** + * The index-1 table, only used for supplementary code points, at offset 2112=0x840. + * Variable length, for code points up to highStart, where the last single-value range starts. + * Maximum length 512=0x200=0x100000>>UTRIE2_SHIFT_1. + * (For 0x100000 supplementary code points U+10000..U+10ffff.) + * + * The part of the index-2 table for supplementary code points starts + * after this index-1 table. + * + * Both the index-1 table and the following part of the index-2 table + * are omitted completely if there is only BMP data. + */ + static final int UTRIE2_INDEX_1_OFFSET=UTRIE2_UTF8_2B_INDEX_2_OFFSET+UTRIE2_UTF8_2B_INDEX_2_LENGTH; + + /** + * The illegal-UTF-8 data block follows the ASCII block, at offset 128=0x80. + * Used with linear access for single bytes 0..0xbf for simple error handling. + * Length 64=0x40, not UTRIE2_DATA_BLOCK_LENGTH. + */ + static final int UTRIE2_BAD_UTF8_DATA_OFFSET=0x80; + + /** + * Implementation class for an iterator over a Trie2. + * + * Iteration over a Trie2 first returns all of the ranges that are indexed by code points, + * then returns the special alternate values for the lead surrogates + * + * @internal + */ + class Trie2Iterator implements Iterator { + + // The normal constructor that configures the iterator to cover the complete + // contents of the Trie2 + Trie2Iterator(ValueMapper vm) { + mapper = vm; + nextStart = 0; + limitCP = 0x110000; + doLeadSurrogates = true; + } + + /** + * The main next() function for Trie2 iterators + * + */ + public Range next() { + if (!hasNext()) { + throw new NoSuchElementException(); + } + if (nextStart >= limitCP) { + // Switch over from iterating normal code point values to + // doing the alternate lead-surrogate values. + doingCodePoints = false; + nextStart = 0xd800; + } + int endOfRange = 0; + int val = 0; + int mappedVal = 0; + + if (doingCodePoints) { + // Iteration over code point values. + val = get(nextStart); + mappedVal = mapper.map(val); + endOfRange = rangeEnd(nextStart, limitCP, val); + // Loop once for each range in the Trie2 with the same raw (unmapped) value. + // Loop continues so long as the mapped values are the same. + for (;;) { + if (endOfRange >= limitCP-1) { + break; + } + val = get(endOfRange+1); + if (mapper.map(val) != mappedVal) { + break; + } + endOfRange = rangeEnd(endOfRange+1, limitCP, val); + } + } else { + // Iteration over the alternate lead surrogate values. + val = getFromU16SingleLead((char)nextStart); + mappedVal = mapper.map(val); + endOfRange = rangeEndLS((char)nextStart); + // Loop once for each range in the Trie2 with the same raw (unmapped) value. + // Loop continues so long as the mapped values are the same. + for (;;) { + if (endOfRange >= 0xdbff) { + break; + } + val = getFromU16SingleLead((char)(endOfRange+1)); + if (mapper.map(val) != mappedVal) { + break; + } + endOfRange = rangeEndLS((char)(endOfRange+1)); + } + } + returnValue.startCodePoint = nextStart; + returnValue.endCodePoint = endOfRange; + returnValue.value = mappedVal; + returnValue.leadSurrogate = !doingCodePoints; + nextStart = endOfRange+1; + return returnValue; + } + + /** + * + */ + public boolean hasNext() { + return doingCodePoints && (doLeadSurrogates || nextStart < limitCP) || nextStart < 0xdc00; + } + + private int rangeEndLS(char startingLS) { + if (startingLS >= 0xdbff) { + return 0xdbff; + } + + int c; + int val = getFromU16SingleLead(startingLS); + for (c = startingLS+1; c <= 0x0dbff; c++) { + if (getFromU16SingleLead((char)c) != val) { + break; + } + } + return c-1; + } + + // + // Iteration State Variables + // + private ValueMapper mapper; + private Range returnValue = new Range(); + // The starting code point for the next range to be returned. + private int nextStart; + // The upper limit for the last normal range to be returned. Normally 0x110000, but + // may be lower when iterating over the code points for a single lead surrogate. + private int limitCP; + + // True while iterating over the Trie2 values for code points. + // False while iterating over the alternate values for lead surrogates. + private boolean doingCodePoints = true; + + // True if the iterator should iterate the special values for lead surrogates in + // addition to the normal values for code points. + private boolean doLeadSurrogates = true; + } + + /** + * Find the last character in a contiguous range of characters with the + * same Trie2 value as the input character. + * + * @param c The character to begin with. + * @return The last contiguous character with the same value. + */ + int rangeEnd(int start, int limitp, int val) { + int c; + int limit = Math.min(highStart, limitp); + + for (c = start+1; c < limit; c++) { + if (get(c) != val) { + break; + } + } + if (c >= highStart) { + c = limitp; + } + return c - 1; + } + + + // + // Hashing implementation functions. FNV hash. Respected public domain algorithm. + // + private static int initHash() { + return 0x811c9DC5; // unsigned 2166136261 + } + + private static int hashByte(int h, int b) { + h = h * 16777619; + h = h ^ b; + return h; + } + + private static int hashUChar32(int h, int c) { + h = Trie2.hashByte(h, c & 255); + h = Trie2.hashByte(h, (c>>8) & 255); + h = Trie2.hashByte(h, c>>16); + return h; + } + + private static int hashInt(int h, int i) { + h = Trie2.hashByte(h, i & 255); + h = Trie2.hashByte(h, (i>>8) & 255); + h = Trie2.hashByte(h, (i>>16) & 255); + h = Trie2.hashByte(h, (i>>24) & 255); + return h; + } + +} --- old/src/java.base/share/classes/sun/text/normalizer/Trie2_16.java 2020-01-10 15:57:39.000000000 -0800 +++ /dev/null 2020-01-10 15:57:39.000000000 -0800 @@ -1,167 +0,0 @@ -/* - * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved. - * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. - * - * This code is free software; you can redistribute it and/or modify it - * under the terms of the GNU General Public License version 2 only, as - * published by the Free Software Foundation. Oracle designates this - * particular file as subject to the "Classpath" exception as provided - * by Oracle in the LICENSE file that accompanied this code. - * - * This code is distributed in the hope that it will be useful, but WITHOUT - * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or - * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License - * version 2 for more details (a copy is included in the LICENSE file that - * accompanied this code). - * - * You should have received a copy of the GNU General Public License version - * 2 along with this work; if not, write to the Free Software Foundation, - * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. - * - * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA - * or visit www.oracle.com if you need additional information or have any - * questions. - */ - -/* - ******************************************************************************* - * Copyright (C) 2009-2014, International Business Machines Corporation and - * others. All Rights Reserved. - ******************************************************************************* - */ - -package sun.text.normalizer; - -import java.io.IOException; -import java.nio.ByteBuffer; - - -/** - * @author aheninger - * - * A read-only Trie2, holding 16 bit data values. - * - * A Trie2 is a highly optimized data structure for mapping from Unicode - * code points (values ranging from 0 to 0x10ffff) to a 16 or 32 bit value. - * - * See class Trie2 for descriptions of the API for accessing the contents of a trie. - * - * The fundamental data access methods are declared final in this class, with - * the intent that applications might gain a little extra performance, when compared - * with calling the same methods via the abstract UTrie2 base class. - */ -public final class Trie2_16 extends Trie2 { - - /** - * Internal constructor, not for general use. - */ - Trie2_16() { - } - - - /** - * Create a Trie2 from its serialized form. Inverse of utrie2_serialize(). - * The serialized format is identical between ICU4C and ICU4J, so this function - * will work with serialized Trie2s from either. - * - * The serialized Trie2 in the bytes may be in either little or big endian byte order. - * This allows using serialized Tries from ICU4C without needing to consider the - * byte order of the system that created them. - * - * @param bytes a byte buffer to the serialized form of a UTrie2. - * @return An unserialized Trie2_16, ready for use. - * @throws IllegalArgumentException if the buffer does not contain a serialized Trie2. - * @throws IOException if a read error occurs in the buffer. - * @throws ClassCastException if the bytes contain a serialized Trie2_32 - */ - public static Trie2_16 createFromSerialized(ByteBuffer bytes) throws IOException { - return (Trie2_16) Trie2.createFromSerialized(bytes); - } - - /** - * Get the value for a code point as stored in the Trie2. - * - * @param codePoint the code point - * @return the value - */ - @Override - public final int get(int codePoint) { - int value; - int ix; - - if (codePoint >= 0) { - if (codePoint < 0x0d800 || (codePoint > 0x0dbff && codePoint <= 0x0ffff)) { - // Ordinary BMP code point, excluding leading surrogates. - // BMP uses a single level lookup. BMP index starts at offset 0 in the Trie2 index. - // 16 bit data is stored in the index array itself. - ix = index[codePoint >> UTRIE2_SHIFT_2]; - ix = (ix << UTRIE2_INDEX_SHIFT) + (codePoint & UTRIE2_DATA_MASK); - value = index[ix]; - return value; - } - if (codePoint <= 0xffff) { - // Lead Surrogate Code Point. A Separate index section is stored for - // lead surrogate code units and code points. - // The main index has the code unit data. - // For this function, we need the code point data. - // Note: this expression could be refactored for slightly improved efficiency, but - // surrogate code points will be so rare in practice that it's not worth it. - ix = index[UTRIE2_LSCP_INDEX_2_OFFSET + ((codePoint - 0xd800) >> UTRIE2_SHIFT_2)]; - ix = (ix << UTRIE2_INDEX_SHIFT) + (codePoint & UTRIE2_DATA_MASK); - value = index[ix]; - return value; - } - if (codePoint < highStart) { - // Supplemental code point, use two-level lookup. - ix = (UTRIE2_INDEX_1_OFFSET - UTRIE2_OMITTED_BMP_INDEX_1_LENGTH) + (codePoint >> UTRIE2_SHIFT_1); - ix = index[ix]; - ix += (codePoint >> UTRIE2_SHIFT_2) & UTRIE2_INDEX_2_MASK; - ix = index[ix]; - ix = (ix << UTRIE2_INDEX_SHIFT) + (codePoint & UTRIE2_DATA_MASK); - value = index[ix]; - return value; - } - if (codePoint <= 0x10ffff) { - value = index[highValueIndex]; - return value; - } - } - - // Fall through. The code point is outside of the legal range of 0..0x10ffff. - return errorValue; - } - - - /** - * Get a Trie2 value for a UTF-16 code unit. - * - * This function returns the same value as get() if the input - * character is outside of the lead surrogate range - * - * There are two values stored in a Trie2 for inputs in the lead - * surrogate range. This function returns the alternate value, - * while Trie2.get() returns the main value. - * - * @param codeUnit a 16 bit code unit or lead surrogate value. - * @return the value - */ - @Override - public int getFromU16SingleLead(char codeUnit) { - int value; - int ix; - - // Because the input is a 16 bit char, we can skip the tests for it being in - // the BMP range. It is. - ix = index[codeUnit >> UTRIE2_SHIFT_2]; - ix = (ix << UTRIE2_INDEX_SHIFT) + (codeUnit & UTRIE2_DATA_MASK); - value = index[ix]; - return value; - } - - /** - * @return the number of bytes of the serialized trie - */ - public int getSerializedLength() { - return 16+(header.indexLength+dataLength)*2; - } -} --- /dev/null 2020-01-10 15:57:39.000000000 -0800 +++ new/src/java.base/share/classes/jdk/internal/icu/impl/Trie2_16.java 2020-01-10 15:57:38.000000000 -0800 @@ -0,0 +1,167 @@ +/* + * Copyright (c) 2015, 2020, Oracle and/or its affiliates. All rights reserved. + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This code is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License version 2 only, as + * published by the Free Software Foundation. Oracle designates this + * particular file as subject to the "Classpath" exception as provided + * by Oracle in the LICENSE file that accompanied this code. + * + * This code is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License + * version 2 for more details (a copy is included in the LICENSE file that + * accompanied this code). + * + * You should have received a copy of the GNU General Public License version + * 2 along with this work; if not, write to the Free Software Foundation, + * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. + * + * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA + * or visit www.oracle.com if you need additional information or have any + * questions. + */ + +/* + ******************************************************************************* + * Copyright (C) 2009-2014, International Business Machines Corporation and + * others. All Rights Reserved. + ******************************************************************************* + */ + +package jdk.internal.icu.impl; + +import java.io.IOException; +import java.nio.ByteBuffer; + + +/** + * @author aheninger + * + * A read-only Trie2, holding 16 bit data values. + * + * A Trie2 is a highly optimized data structure for mapping from Unicode + * code points (values ranging from 0 to 0x10ffff) to a 16 or 32 bit value. + * + * See class Trie2 for descriptions of the API for accessing the contents of a trie. + * + * The fundamental data access methods are declared final in this class, with + * the intent that applications might gain a little extra performance, when compared + * with calling the same methods via the abstract UTrie2 base class. + */ +public final class Trie2_16 extends Trie2 { + + /** + * Internal constructor, not for general use. + */ + Trie2_16() { + } + + + /** + * Create a Trie2 from its serialized form. Inverse of utrie2_serialize(). + * The serialized format is identical between ICU4C and ICU4J, so this function + * will work with serialized Trie2s from either. + * + * The serialized Trie2 in the bytes may be in either little or big endian byte order. + * This allows using serialized Tries from ICU4C without needing to consider the + * byte order of the system that created them. + * + * @param bytes a byte buffer to the serialized form of a UTrie2. + * @return An unserialized Trie2_16, ready for use. + * @throws IllegalArgumentException if the buffer does not contain a serialized Trie2. + * @throws IOException if a read error occurs in the buffer. + * @throws ClassCastException if the bytes contain a serialized Trie2_32 + */ + public static Trie2_16 createFromSerialized(ByteBuffer bytes) throws IOException { + return (Trie2_16) Trie2.createFromSerialized(bytes); + } + + /** + * Get the value for a code point as stored in the Trie2. + * + * @param codePoint the code point + * @return the value + */ + @Override + public final int get(int codePoint) { + int value; + int ix; + + if (codePoint >= 0) { + if (codePoint < 0x0d800 || (codePoint > 0x0dbff && codePoint <= 0x0ffff)) { + // Ordinary BMP code point, excluding leading surrogates. + // BMP uses a single level lookup. BMP index starts at offset 0 in the Trie2 index. + // 16 bit data is stored in the index array itself. + ix = index[codePoint >> UTRIE2_SHIFT_2]; + ix = (ix << UTRIE2_INDEX_SHIFT) + (codePoint & UTRIE2_DATA_MASK); + value = index[ix]; + return value; + } + if (codePoint <= 0xffff) { + // Lead Surrogate Code Point. A Separate index section is stored for + // lead surrogate code units and code points. + // The main index has the code unit data. + // For this function, we need the code point data. + // Note: this expression could be refactored for slightly improved efficiency, but + // surrogate code points will be so rare in practice that it's not worth it. + ix = index[UTRIE2_LSCP_INDEX_2_OFFSET + ((codePoint - 0xd800) >> UTRIE2_SHIFT_2)]; + ix = (ix << UTRIE2_INDEX_SHIFT) + (codePoint & UTRIE2_DATA_MASK); + value = index[ix]; + return value; + } + if (codePoint < highStart) { + // Supplemental code point, use two-level lookup. + ix = (UTRIE2_INDEX_1_OFFSET - UTRIE2_OMITTED_BMP_INDEX_1_LENGTH) + (codePoint >> UTRIE2_SHIFT_1); + ix = index[ix]; + ix += (codePoint >> UTRIE2_SHIFT_2) & UTRIE2_INDEX_2_MASK; + ix = index[ix]; + ix = (ix << UTRIE2_INDEX_SHIFT) + (codePoint & UTRIE2_DATA_MASK); + value = index[ix]; + return value; + } + if (codePoint <= 0x10ffff) { + value = index[highValueIndex]; + return value; + } + } + + // Fall through. The code point is outside of the legal range of 0..0x10ffff. + return errorValue; + } + + + /** + * Get a Trie2 value for a UTF-16 code unit. + * + * This function returns the same value as get() if the input + * character is outside of the lead surrogate range + * + * There are two values stored in a Trie2 for inputs in the lead + * surrogate range. This function returns the alternate value, + * while Trie2.get() returns the main value. + * + * @param codeUnit a 16 bit code unit or lead surrogate value. + * @return the value + */ + @Override + public int getFromU16SingleLead(char codeUnit) { + int value; + int ix; + + // Because the input is a 16 bit char, we can skip the tests for it being in + // the BMP range. It is. + ix = index[codeUnit >> UTRIE2_SHIFT_2]; + ix = (ix << UTRIE2_INDEX_SHIFT) + (codeUnit & UTRIE2_DATA_MASK); + value = index[ix]; + return value; + } + + /** + * @return the number of bytes of the serialized trie + */ + public int getSerializedLength() { + return 16+(header.indexLength+dataLength)*2; + } +} --- old/src/java.base/share/classes/sun/text/normalizer/UBiDiProps.java 2020-01-10 15:57:40.000000000 -0800 +++ /dev/null 2020-01-10 15:57:40.000000000 -0800 @@ -1,267 +0,0 @@ -/* - * Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved. - * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. - * - * This code is free software; you can redistribute it and/or modify it - * under the terms of the GNU General Public License version 2 only, as - * published by the Free Software Foundation. Oracle designates this - * particular file as subject to the "Classpath" exception as provided - * by Oracle in the LICENSE file that accompanied this code. - * - * This code is distributed in the hope that it will be useful, but WITHOUT - * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or - * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License - * version 2 for more details (a copy is included in the LICENSE file that - * accompanied this code). - * - * You should have received a copy of the GNU General Public License version - * 2 along with this work; if not, write to the Free Software Foundation, - * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. - * - * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA - * or visit www.oracle.com if you need additional information or have any - * questions. - */ -/* - ******************************************************************************* - * - * Copyright (C) 2004-2014, International Business Machines - * Corporation and others. All Rights Reserved. - * - ******************************************************************************* - * file name: UBiDiProps.java - * encoding: US-ASCII - * tab size: 8 (not used) - * indentation:4 - * - * created on: 2005jan16 - * created by: Markus W. Scherer - * - * Low-level Unicode bidi/shaping properties access. - * Java port of ubidi_props.h/.c. - */ - -package sun.text.normalizer; - -import java.io.IOException; -import java.nio.ByteBuffer; -import java.util.MissingResourceException; - -public final class UBiDiProps { - // constructors etc. --------------------------------------------------- *** - - // port of ubidi_openProps() - private UBiDiProps() throws IOException{ - ByteBuffer bytes=ICUBinary.getRequiredData(DATA_FILE_NAME); - readData(bytes); - } - - private void readData(ByteBuffer bytes) throws IOException { - // read the header - ICUBinary.readHeader(bytes, FMT, new IsAcceptable()); - - // read indexes[] - int i, count; - count=bytes.getInt(); - if(countexpectedTrieLength) { - throw new IOException(DATA_FILE_NAME+": not enough bytes for the trie"); - } - // skip padding after trie bytes - ICUBinary.skipBytes(bytes, expectedTrieLength-trieLength); - - // read mirrors[] - count=indexes[IX_MIRROR_LENGTH]; - if(count>0) { - mirrors=new int[count]; - for(i=0; i>JT_SHIFT; - } - - public final int getJoiningGroup(int c) { - int start, limit; - - start=indexes[IX_JG_START]; - limit=indexes[IX_JG_LIMIT]; - if(start<=c && c>BPT_SHIFT; - } - - public final int getPairedBracket(int c) { - int props=trie.get(c); - if((props&BPT_MASK)==0) { - return c; - } else { - return getMirror(c, props); - } - } - - // data members -------------------------------------------------------- *** - private int indexes[]; - private int mirrors[]; - private byte jgArray[]; - private byte jgArray2[]; - - private Trie2_16 trie; - - // data format constants ----------------------------------------------- *** - private static final String DATA_FILE_NAME = "/sun/text/resources/ubidi.icu"; - - /* format "BiDi" */ - private static final int FMT=0x42694469; - - /* indexes into indexes[] */ - private static final int IX_TRIE_SIZE=2; - private static final int IX_MIRROR_LENGTH=3; - - private static final int IX_JG_START=4; - private static final int IX_JG_LIMIT=5; - private static final int IX_JG_START2=6; /* new in format version 2.2, ICU 54 */ - private static final int IX_JG_LIMIT2=7; - - private static final int IX_TOP=16; - - // definitions for 16-bit bidi/shaping properties word ----------------- *** - - /* CLASS_SHIFT=0, */ /* bidi class: 5 bits (4..0) */ - private static final int JT_SHIFT=5; /* joining type: 3 bits (7..5) */ - - private static final int BPT_SHIFT=8; /* Bidi_Paired_Bracket_Type(bpt): 2 bits (9..8) */ - - private static final int MIRROR_DELTA_SHIFT=13; /* bidi mirroring delta: 3 bits (15..13) */ - - private static final int CLASS_MASK= 0x0000001f; - private static final int JT_MASK= 0x000000e0; - private static final int BPT_MASK= 0x00000300; - - private static final int getClassFromProps(int props) { - return props&CLASS_MASK; - } - private static final boolean getFlagFromProps(int props, int shift) { - return ((props>>shift)&1)!=0; - } - private static final int getMirrorDeltaFromProps(int props) { - return (short)props>>MIRROR_DELTA_SHIFT; - } - - private static final int ESC_MIRROR_DELTA=-4; - - // definitions for 32-bit mirror table entry --------------------------- *** - - /* the source Unicode code point takes 21 bits (20..0) */ - private static final int MIRROR_INDEX_SHIFT=21; - - private static final int getMirrorCodePoint(int m) { - return m&0x1fffff; - } - private static final int getMirrorIndex(int m) { - return m>>>MIRROR_INDEX_SHIFT; - } - - - /* - * public singleton instance - */ - public static final UBiDiProps INSTANCE; - - // This static initializer block must be placed after - // other static member initialization - static { - try { - INSTANCE = new UBiDiProps(); - } catch (IOException e) { - throw new MissingResourceException(e.getMessage(),DATA_FILE_NAME,""); - } - } -} --- /dev/null 2020-01-10 15:57:40.000000000 -0800 +++ new/src/java.base/share/classes/jdk/internal/icu/impl/UBiDiProps.java 2020-01-10 15:57:40.000000000 -0800 @@ -0,0 +1,274 @@ +/* + * Copyright (c) 2005, 2020, Oracle and/or its affiliates. All rights reserved. + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This code is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License version 2 only, as + * published by the Free Software Foundation. Oracle designates this + * particular file as subject to the "Classpath" exception as provided + * by Oracle in the LICENSE file that accompanied this code. + * + * This code is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License + * version 2 for more details (a copy is included in the LICENSE file that + * accompanied this code). + * + * You should have received a copy of the GNU General Public License version + * 2 along with this work; if not, write to the Free Software Foundation, + * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. + * + * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA + * or visit www.oracle.com if you need additional information or have any + * questions. + */ +/* + ******************************************************************************* + * + * Copyright (C) 2004-2014, International Business Machines + * Corporation and others. All Rights Reserved. + * + ******************************************************************************* + * file name: UBiDiProps.java + * encoding: US-ASCII + * tab size: 8 (not used) + * indentation:4 + * + * created on: 2005jan16 + * created by: Markus W. Scherer + * + * Low-level Unicode bidi/shaping properties access. + * Java port of ubidi_props.h/.c. + */ + +package jdk.internal.icu.impl; + +import jdk.internal.icu.lang.UCharacter; +import jdk.internal.icu.util.VersionInfo; + +import java.io.IOException; +import java.nio.ByteBuffer; +import java.util.MissingResourceException; + +public final class UBiDiProps { + // constructors etc. --------------------------------------------------- *** + + // port of ubidi_openProps() + private UBiDiProps() throws IOException{ + ByteBuffer bytes=ICUBinary.getRequiredData(DATA_FILE_NAME); + readData(bytes); + } + + private void readData(ByteBuffer bytes) throws IOException { + // read the header + ICUBinary.readHeader(bytes, FMT, new IsAcceptable()); + + // read indexes[] + int i, count; + count=bytes.getInt(); + if(countexpectedTrieLength) { + throw new IOException(DATA_FILE_NAME+": not enough bytes for the trie"); + } + // skip padding after trie bytes + ICUBinary.skipBytes(bytes, expectedTrieLength-trieLength); + + // read mirrors[] + count=indexes[IX_MIRROR_LENGTH]; + if(count>0) { + mirrors=new int[count]; + for(i=0; i>JT_SHIFT; + } + + public final int getJoiningGroup(int c) { + int start, limit; + + start=indexes[IX_JG_START]; + limit=indexes[IX_JG_LIMIT]; + if(start<=c && c>BPT_SHIFT; + } + + public final int getPairedBracket(int c) { + int props=trie.get(c); + if((props&BPT_MASK)==0) { + return c; + } else { + return getMirror(c, props); + } + } + + // data members -------------------------------------------------------- *** + private int indexes[]; + private int mirrors[]; + private byte jgArray[]; + private byte jgArray2[]; + + private Trie2_16 trie; + + // data format constants ----------------------------------------------- *** + @SuppressWarnings("deprecation") + private static final String DATA_FILE_NAME = + "/jdk/internal/icu/impl/data/icudt" + + VersionInfo.ICU_DATA_VERSION_PATH + + "/ubidi.icu"; + + /* format "BiDi" */ + private static final int FMT=0x42694469; + + /* indexes into indexes[] */ + private static final int IX_TRIE_SIZE=2; + private static final int IX_MIRROR_LENGTH=3; + + private static final int IX_JG_START=4; + private static final int IX_JG_LIMIT=5; + private static final int IX_JG_START2=6; /* new in format version 2.2, ICU 54 */ + private static final int IX_JG_LIMIT2=7; + + private static final int IX_TOP=16; + + // definitions for 16-bit bidi/shaping properties word ----------------- *** + + /* CLASS_SHIFT=0, */ /* bidi class: 5 bits (4..0) */ + private static final int JT_SHIFT=5; /* joining type: 3 bits (7..5) */ + + private static final int BPT_SHIFT=8; /* Bidi_Paired_Bracket_Type(bpt): 2 bits (9..8) */ + + private static final int MIRROR_DELTA_SHIFT=13; /* bidi mirroring delta: 3 bits (15..13) */ + + private static final int CLASS_MASK= 0x0000001f; + private static final int JT_MASK= 0x000000e0; + private static final int BPT_MASK= 0x00000300; + + private static final int getClassFromProps(int props) { + return props&CLASS_MASK; + } + private static final boolean getFlagFromProps(int props, int shift) { + return ((props>>shift)&1)!=0; + } + private static final int getMirrorDeltaFromProps(int props) { + return (short)props>>MIRROR_DELTA_SHIFT; + } + + private static final int ESC_MIRROR_DELTA=-4; + + // definitions for 32-bit mirror table entry --------------------------- *** + + /* the source Unicode code point takes 21 bits (20..0) */ + private static final int MIRROR_INDEX_SHIFT=21; + + private static final int getMirrorCodePoint(int m) { + return m&0x1fffff; + } + private static final int getMirrorIndex(int m) { + return m>>>MIRROR_INDEX_SHIFT; + } + + + /* + * public singleton instance + */ + public static final UBiDiProps INSTANCE; + + // This static initializer block must be placed after + // other static member initialization + static { + try { + INSTANCE = new UBiDiProps(); + } catch (IOException e) { + throw new MissingResourceException(e.getMessage(),DATA_FILE_NAME,""); + } + } +} --- old/src/java.base/share/classes/sun/text/normalizer/UCharacterProperty.java 2020-01-10 15:57:41.000000000 -0800 +++ /dev/null 2020-01-10 15:57:41.000000000 -0800 @@ -1,607 +0,0 @@ -/* - * Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved. - * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. - * - * This code is free software; you can redistribute it and/or modify it - * under the terms of the GNU General Public License version 2 only, as - * published by the Free Software Foundation. Oracle designates this - * particular file as subject to the "Classpath" exception as provided - * by Oracle in the LICENSE file that accompanied this code. - * - * This code is distributed in the hope that it will be useful, but WITHOUT - * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or - * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License - * version 2 for more details (a copy is included in the LICENSE file that - * accompanied this code). - * - * You should have received a copy of the GNU General Public License version - * 2 along with this work; if not, write to the Free Software Foundation, - * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. - * - * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA - * or visit www.oracle.com if you need additional information or have any - * questions. - */ -/* - ******************************************************************************* - * Copyright (C) 1996-2014, International Business Machines Corporation and - * others. All Rights Reserved. - ******************************************************************************* - */ - -package sun.text.normalizer; - -import java.io.IOException; -import java.nio.ByteBuffer; -import java.util.Iterator; -import java.util.MissingResourceException; - -import sun.text.normalizer.UCharacter.HangulSyllableType; -import sun.text.normalizer.UCharacter.NumericType; - -/** -*

Internal class used for Unicode character property database.

-*

This classes store binary data read from uprops.icu. -* It does not have the capability to parse the data into more high-level -* information. It only returns bytes of information when required.

-*

Due to the form most commonly used for retrieval, array of char is used -* to store the binary data.

-*

UCharacterPropertyDB also contains information on accessing indexes to -* significant points in the binary data.

-*

Responsibility for molding the binary data into more meaning form lies on -* UCharacter.

-* @author Syn Wee Quek -* @since release 2.1, february 1st 2002 -*/ - -final class UCharacterProperty -{ - // public data members ----------------------------------------------- - - /* - * public singleton instance - */ - public static final UCharacterProperty INSTANCE; - - /** - * Trie data - */ - public Trie2_16 m_trie_; - - /** - * Unicode version - */ - public VersionInfo m_unicodeVersion_; - - /** - * Character type mask - */ - public static final int TYPE_MASK = 0x1F; - - // uprops.h enum UPropertySource --------------------------------------- *** - - /** From uchar.c/uprops.icu main trie */ - public static final int SRC_CHAR=1; - /** From uchar.c/uprops.icu properties vectors trie */ - public static final int SRC_PROPSVEC=2; - /** From ubidi_props.c/ubidi.icu */ - public static final int SRC_BIDI=5; - /** From normalizer2impl.cpp/nfc.nrm */ - public static final int SRC_NFC=8; - /** From normalizer2impl.cpp/nfkc.nrm */ - public static final int SRC_NFKC=9; - - // public methods ---------------------------------------------------- - - /** - * Gets the main property value for code point ch. - * @param ch code point whose property value is to be retrieved - * @return property value of code point - */ - public final int getProperty(int ch) - { - return m_trie_.get(ch); - } - - /** - * Gets the unicode additional properties. - * Java version of C u_getUnicodeProperties(). - * @param codepoint codepoint whose additional properties is to be - * retrieved - * @param column The column index. - * @return unicode properties - */ - public int getAdditional(int codepoint, int column) { - assert column >= 0; - if (column >= m_additionalColumnsCount_) { - return 0; - } - return m_additionalVectors_[m_additionalTrie_.get(codepoint) + column]; - } - - /** - *

Get the "age" of the code point.

- *

The "age" is the Unicode version when the code point was first - * designated (as a non-character or for Private Use) or assigned a - * character.

- *

This can be useful to avoid emitting code points to receiving - * processes that do not accept newer characters.

- *

The data is from the UCD file DerivedAge.txt.

- *

This API does not check the validity of the codepoint.

- * @param codepoint The code point. - * @return the Unicode version number - */ - public VersionInfo getAge(int codepoint) - { - int version = getAdditional(codepoint, 0) >> AGE_SHIFT_; - return VersionInfo.getInstance( - (version >> FIRST_NIBBLE_SHIFT_) & LAST_NIBBLE_MASK_, - version & LAST_NIBBLE_MASK_, 0, 0); - } - - // int-value and enumerated properties --------------------------------- *** - - public int getType(int c) { - return getProperty(c)&TYPE_MASK; - } - - /* - * Map some of the Grapheme Cluster Break values to Hangul Syllable Types. - * Hangul_Syllable_Type is fully redundant with a subset of Grapheme_Cluster_Break. - */ - private static final int /* UHangulSyllableType */ gcbToHst[]={ - HangulSyllableType.NOT_APPLICABLE, /* U_GCB_OTHER */ - HangulSyllableType.NOT_APPLICABLE, /* U_GCB_CONTROL */ - HangulSyllableType.NOT_APPLICABLE, /* U_GCB_CR */ - HangulSyllableType.NOT_APPLICABLE, /* U_GCB_EXTEND */ - HangulSyllableType.LEADING_JAMO, /* U_GCB_L */ - HangulSyllableType.NOT_APPLICABLE, /* U_GCB_LF */ - HangulSyllableType.LV_SYLLABLE, /* U_GCB_LV */ - HangulSyllableType.LVT_SYLLABLE, /* U_GCB_LVT */ - HangulSyllableType.TRAILING_JAMO, /* U_GCB_T */ - HangulSyllableType.VOWEL_JAMO /* U_GCB_V */ - /* - * Omit GCB values beyond what we need for hst. - * The code below checks for the array length. - */ - }; - - private class IntProperty { - int column; // SRC_PROPSVEC column, or "source" if mask==0 - int mask; - int shift; - - IntProperty(int column, int mask, int shift) { - this.column=column; - this.mask=mask; - this.shift=shift; - } - - IntProperty(int source) { - this.column=source; - this.mask=0; - } - - int getValue(int c) { - // systematic, directly stored properties - return (getAdditional(c, column)&mask)>>>shift; - } - } - - private class BiDiIntProperty extends IntProperty { - BiDiIntProperty() { - super(SRC_BIDI); - } - } - - private class CombiningClassIntProperty extends IntProperty { - CombiningClassIntProperty(int source) { - super(source); - } - } - - private class NormQuickCheckIntProperty extends IntProperty { // UCHAR_NF*_QUICK_CHECK properties - int which; - int max; - - NormQuickCheckIntProperty(int source, int which, int max) { - super(source); - this.which=which; - this.max=max; - } - } - - private IntProperty intProp = new BiDiIntProperty() { // BIDI_PAIRED_BRACKET_TYPE - int getValue(int c) { - return UBiDiProps.INSTANCE.getPairedBracketType(c); - } - }; - - public int getIntPropertyValue(int c, int which) { - if (which == BIDI_PAIRED_BRACKET_TYPE) { - return intProp.getValue(c); - } - return 0; // undefined - } - - /** - * Forms a supplementary code point from the argument character
- * Note this is for internal use hence no checks for the validity of the - * surrogate characters are done - * @param lead lead surrogate character - * @param trail trailing surrogate character - * @return code point of the supplementary character - */ - public static int getRawSupplementary(char lead, char trail) - { - return (lead << LEAD_SURROGATE_SHIFT_) + trail + SURROGATE_OFFSET_; - } - - /** - * Gets the type mask - * @param type character type - * @return mask - */ - public static final int getMask(int type) - { - return 1 << type; - } - - /** - * Returns the digit values of characters like 'A' - 'Z', normal, - * half-width and full-width. This method assumes that the other digit - * characters are checked by the calling method. - * @param ch character to test - * @return -1 if ch is not a character of the form 'A' - 'Z', otherwise - * its corresponding digit will be returned. - */ - public static int getEuropeanDigit(int ch) { - if ((ch > 0x7a && ch < 0xff21) - || ch < 0x41 || (ch > 0x5a && ch < 0x61) - || ch > 0xff5a || (ch > 0xff3a && ch < 0xff41)) { - return -1; - } - if (ch <= 0x7a) { - // ch >= 0x41 or ch < 0x61 - return ch + 10 - ((ch <= 0x5a) ? 0x41 : 0x61); - } - // ch >= 0xff21 - if (ch <= 0xff3a) { - return ch + 10 - 0xff21; - } - // ch >= 0xff41 && ch <= 0xff5a - return ch + 10 - 0xff41; - } - - public int digit(int c) { - int value = getNumericTypeValue(getProperty(c)) - NTV_DECIMAL_START_; - if(value<=9) { - return value; - } else { - return -1; - } - } - - // protected variables ----------------------------------------------- - - /** - * Extra property trie - */ - Trie2_16 m_additionalTrie_; - /** - * Extra property vectors, 1st column for age and second for binary - * properties. - */ - int m_additionalVectors_[]; - /** - * Number of additional columns - */ - int m_additionalColumnsCount_; - /** - * Maximum values for block, bits used as in vector word - * 0 - */ - int m_maxBlockScriptValue_; - /** - * Maximum values for script, bits used as in vector word - * 0 - */ - int m_maxJTGValue_; - /** - * Script_Extensions data - */ - public char[] m_scriptExtensions_; - - // private variables ------------------------------------------------- - - /** - * Default name of the datafile - */ - private static final String DATA_FILE_NAME_ = "/sun/text/resources/uprops.icu"; - - /** - * Shift value for lead surrogate to form a supplementary character. - */ - private static final int LEAD_SURROGATE_SHIFT_ = 10; - /** - * Offset to add to combined surrogate pair to avoid masking. - */ - private static final int SURROGATE_OFFSET_ = - UTF16.SUPPLEMENTARY_MIN_VALUE - - (UTF16.SURROGATE_MIN_VALUE << - LEAD_SURROGATE_SHIFT_) - - UTF16.TRAIL_SURROGATE_MIN_VALUE; - - - // property data constants ------------------------------------------------- - - /** - * Numeric types and values in the main properties words. - */ - private static final int NUMERIC_TYPE_VALUE_SHIFT_ = 6; - private static final int getNumericTypeValue(int props) { - return props >> NUMERIC_TYPE_VALUE_SHIFT_; - } - - /* constants for the storage form of numeric types and values */ - /** No numeric value. */ - private static final int NTV_NONE_ = 0; - /** Decimal digits: nv=0..9 */ - private static final int NTV_DECIMAL_START_ = 1; - /** Other digits: nv=0..9 */ - private static final int NTV_DIGIT_START_ = 11; - /** Small integers: nv=0..154 */ - private static final int NTV_NUMERIC_START_ = 21; - - private static final int ntvGetType(int ntv) { - return - (ntv==NTV_NONE_) ? NumericType.NONE : - (ntv expectedTrieLength) { - throw new IOException("uprops.icu: not enough bytes for main trie"); - } - // skip padding after trie bytes - ICUBinary.skipBytes(bytes, expectedTrieLength - trieLength); - - // skip unused intervening data structures - ICUBinary.skipBytes(bytes, (additionalOffset - propertyOffset) * 4); - - if(m_additionalColumnsCount_ > 0) { - // reads the additional property block - m_additionalTrie_ = Trie2_16.createFromSerialized(bytes); - expectedTrieLength = (additionalVectorsOffset-additionalOffset)*4; - trieLength = m_additionalTrie_.getSerializedLength(); - if(trieLength > expectedTrieLength) { - throw new IOException("uprops.icu: not enough bytes for additional-properties trie"); - } - // skip padding after trie bytes - ICUBinary.skipBytes(bytes, expectedTrieLength - trieLength); - - // additional properties - int size = scriptExtensionsOffset - additionalVectorsOffset; - m_additionalVectors_ = new int[size]; - for (int i = 0; i < size; i ++) { - m_additionalVectors_[i] = bytes.getInt(); - } - } - - // Script_Extensions - int numChars = (reservedOffset7 - scriptExtensionsOffset) * 2; - if(numChars > 0) { - m_scriptExtensions_ = new char[numChars]; - for(int i = 0; i < numChars; ++i) { - m_scriptExtensions_[i] = bytes.getChar(); - } - } - } - - private static final class IsAcceptable implements ICUBinary.Authenticate { - // @Override when we switch to Java 6 - public boolean isDataVersionAcceptable(byte version[]) { - return version[0] == 7; - } - } - - private static final int DATA_FORMAT = 0x5550726F; // "UPro" - - public void upropsvec_addPropertyStarts(UnicodeSet set) { - /* add the start code point of each same-value range of the properties vectors trie */ - if(m_additionalColumnsCount_>0) { - /* if m_additionalColumnsCount_==0 then the properties vectors trie may not be there at all */ - Iterator trieIterator = m_additionalTrie_.iterator(); - Trie2.Range range; - while(trieIterator.hasNext() && !(range=trieIterator.next()).leadSurrogate) { - set.add(range.startCodePoint); - } - } - } - - // This static initializer block must be placed after - // other static member initialization - static { - try { - INSTANCE = new UCharacterProperty(); - } - catch (IOException e) { - throw new MissingResourceException(e.getMessage(),DATA_FILE_NAME_,""); - } - } - - - // Moved from UProperty.java - /** - * Enumerated property Bidi_Paired_Bracket_Type (new in Unicode 6.3). - * Used in UAX #9: Unicode Bidirectional Algorithm - * (http://www.unicode.org/reports/tr9/) - * Returns UCharacter.BidiPairedBracketType values. - * @stable ICU 52 - */ - public static final int BIDI_PAIRED_BRACKET_TYPE = 0x1015; - -} --- /dev/null 2020-01-10 15:57:41.000000000 -0800 +++ new/src/java.base/share/classes/jdk/internal/icu/impl/UCharacterProperty.java 2020-01-10 15:57:41.000000000 -0800 @@ -0,0 +1,614 @@ +/* + * Copyright (c) 2005, 2020, Oracle and/or its affiliates. All rights reserved. + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This code is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License version 2 only, as + * published by the Free Software Foundation. Oracle designates this + * particular file as subject to the "Classpath" exception as provided + * by Oracle in the LICENSE file that accompanied this code. + * + * This code is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License + * version 2 for more details (a copy is included in the LICENSE file that + * accompanied this code). + * + * You should have received a copy of the GNU General Public License version + * 2 along with this work; if not, write to the Free Software Foundation, + * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. + * + * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA + * or visit www.oracle.com if you need additional information or have any + * questions. + */ +/* + ******************************************************************************* + * Copyright (C) 1996-2014, International Business Machines Corporation and + * others. All Rights Reserved. + ******************************************************************************* + */ + +package jdk.internal.icu.impl; + +import java.io.IOException; +import java.nio.ByteBuffer; +import java.util.Iterator; +import java.util.MissingResourceException; + +import jdk.internal.icu.lang.UCharacter.HangulSyllableType; +import jdk.internal.icu.lang.UCharacter.NumericType; +import jdk.internal.icu.text.UTF16; +import jdk.internal.icu.text.UnicodeSet; +import jdk.internal.icu.util.VersionInfo; + +/** +*

Internal class used for Unicode character property database.

+*

This classes store binary data read from uprops.icu. +* It does not have the capability to parse the data into more high-level +* information. It only returns bytes of information when required.

+*

Due to the form most commonly used for retrieval, array of char is used +* to store the binary data.

+*

UCharacterPropertyDB also contains information on accessing indexes to +* significant points in the binary data.

+*

Responsibility for molding the binary data into more meaning form lies on +* UCharacter.

+* @author Syn Wee Quek +* @since release 2.1, february 1st 2002 +*/ + +public final class UCharacterProperty +{ + // public data members ----------------------------------------------- + + /* + * public singleton instance + */ + public static final UCharacterProperty INSTANCE; + + /** + * Trie data + */ + public Trie2_16 m_trie_; + + /** + * Unicode version + */ + public VersionInfo m_unicodeVersion_; + + /** + * Character type mask + */ + public static final int TYPE_MASK = 0x1F; + + // uprops.h enum UPropertySource --------------------------------------- *** + + /** From uchar.c/uprops.icu main trie */ + public static final int SRC_CHAR=1; + /** From uchar.c/uprops.icu properties vectors trie */ + public static final int SRC_PROPSVEC=2; + /** From ubidi_props.c/ubidi.icu */ + public static final int SRC_BIDI=5; + /** From normalizer2impl.cpp/nfc.nrm */ + public static final int SRC_NFC=8; + /** From normalizer2impl.cpp/nfkc.nrm */ + public static final int SRC_NFKC=9; + + // public methods ---------------------------------------------------- + + /** + * Gets the main property value for code point ch. + * @param ch code point whose property value is to be retrieved + * @return property value of code point + */ + public final int getProperty(int ch) + { + return m_trie_.get(ch); + } + + /** + * Gets the unicode additional properties. + * Java version of C u_getUnicodeProperties(). + * @param codepoint codepoint whose additional properties is to be + * retrieved + * @param column The column index. + * @return unicode properties + */ + public int getAdditional(int codepoint, int column) { + assert column >= 0; + if (column >= m_additionalColumnsCount_) { + return 0; + } + return m_additionalVectors_[m_additionalTrie_.get(codepoint) + column]; + } + + /** + *

Get the "age" of the code point.

+ *

The "age" is the Unicode version when the code point was first + * designated (as a non-character or for Private Use) or assigned a + * character.

+ *

This can be useful to avoid emitting code points to receiving + * processes that do not accept newer characters.

+ *

The data is from the UCD file DerivedAge.txt.

+ *

This API does not check the validity of the codepoint.

+ * @param codepoint The code point. + * @return the Unicode version number + */ + public VersionInfo getAge(int codepoint) + { + int version = getAdditional(codepoint, 0) >> AGE_SHIFT_; + return VersionInfo.getInstance( + (version >> FIRST_NIBBLE_SHIFT_) & LAST_NIBBLE_MASK_, + version & LAST_NIBBLE_MASK_, 0, 0); + } + + // int-value and enumerated properties --------------------------------- *** + + public int getType(int c) { + return getProperty(c)&TYPE_MASK; + } + + /* + * Map some of the Grapheme Cluster Break values to Hangul Syllable Types. + * Hangul_Syllable_Type is fully redundant with a subset of Grapheme_Cluster_Break. + */ + private static final int /* UHangulSyllableType */ gcbToHst[]={ + HangulSyllableType.NOT_APPLICABLE, /* U_GCB_OTHER */ + HangulSyllableType.NOT_APPLICABLE, /* U_GCB_CONTROL */ + HangulSyllableType.NOT_APPLICABLE, /* U_GCB_CR */ + HangulSyllableType.NOT_APPLICABLE, /* U_GCB_EXTEND */ + HangulSyllableType.LEADING_JAMO, /* U_GCB_L */ + HangulSyllableType.NOT_APPLICABLE, /* U_GCB_LF */ + HangulSyllableType.LV_SYLLABLE, /* U_GCB_LV */ + HangulSyllableType.LVT_SYLLABLE, /* U_GCB_LVT */ + HangulSyllableType.TRAILING_JAMO, /* U_GCB_T */ + HangulSyllableType.VOWEL_JAMO /* U_GCB_V */ + /* + * Omit GCB values beyond what we need for hst. + * The code below checks for the array length. + */ + }; + + private class IntProperty { + int column; // SRC_PROPSVEC column, or "source" if mask==0 + int mask; + int shift; + + IntProperty(int column, int mask, int shift) { + this.column=column; + this.mask=mask; + this.shift=shift; + } + + IntProperty(int source) { + this.column=source; + this.mask=0; + } + + int getValue(int c) { + // systematic, directly stored properties + return (getAdditional(c, column)&mask)>>>shift; + } + } + + private class BiDiIntProperty extends IntProperty { + BiDiIntProperty() { + super(SRC_BIDI); + } + } + + private class CombiningClassIntProperty extends IntProperty { + CombiningClassIntProperty(int source) { + super(source); + } + } + + private class NormQuickCheckIntProperty extends IntProperty { // UCHAR_NF*_QUICK_CHECK properties + int which; + int max; + + NormQuickCheckIntProperty(int source, int which, int max) { + super(source); + this.which=which; + this.max=max; + } + } + + private IntProperty intProp = new BiDiIntProperty() { // BIDI_PAIRED_BRACKET_TYPE + int getValue(int c) { + return UBiDiProps.INSTANCE.getPairedBracketType(c); + } + }; + + public int getIntPropertyValue(int c, int which) { + if (which == BIDI_PAIRED_BRACKET_TYPE) { + return intProp.getValue(c); + } + return 0; // undefined + } + + /** + * Forms a supplementary code point from the argument character
+ * Note this is for internal use hence no checks for the validity of the + * surrogate characters are done + * @param lead lead surrogate character + * @param trail trailing surrogate character + * @return code point of the supplementary character + */ + public static int getRawSupplementary(char lead, char trail) + { + return (lead << LEAD_SURROGATE_SHIFT_) + trail + SURROGATE_OFFSET_; + } + + /** + * Gets the type mask + * @param type character type + * @return mask + */ + public static final int getMask(int type) + { + return 1 << type; + } + + /** + * Returns the digit values of characters like 'A' - 'Z', normal, + * half-width and full-width. This method assumes that the other digit + * characters are checked by the calling method. + * @param ch character to test + * @return -1 if ch is not a character of the form 'A' - 'Z', otherwise + * its corresponding digit will be returned. + */ + public static int getEuropeanDigit(int ch) { + if ((ch > 0x7a && ch < 0xff21) + || ch < 0x41 || (ch > 0x5a && ch < 0x61) + || ch > 0xff5a || (ch > 0xff3a && ch < 0xff41)) { + return -1; + } + if (ch <= 0x7a) { + // ch >= 0x41 or ch < 0x61 + return ch + 10 - ((ch <= 0x5a) ? 0x41 : 0x61); + } + // ch >= 0xff21 + if (ch <= 0xff3a) { + return ch + 10 - 0xff21; + } + // ch >= 0xff41 && ch <= 0xff5a + return ch + 10 - 0xff41; + } + + public int digit(int c) { + int value = getNumericTypeValue(getProperty(c)) - NTV_DECIMAL_START_; + if(value<=9) { + return value; + } else { + return -1; + } + } + + // protected variables ----------------------------------------------- + + /** + * Extra property trie + */ + Trie2_16 m_additionalTrie_; + /** + * Extra property vectors, 1st column for age and second for binary + * properties. + */ + int m_additionalVectors_[]; + /** + * Number of additional columns + */ + int m_additionalColumnsCount_; + /** + * Maximum values for block, bits used as in vector word + * 0 + */ + int m_maxBlockScriptValue_; + /** + * Maximum values for script, bits used as in vector word + * 0 + */ + int m_maxJTGValue_; + /** + * Script_Extensions data + */ + public char[] m_scriptExtensions_; + + // private variables ------------------------------------------------- + + /** + * Default name of the datafile + */ + @SuppressWarnings("deprecation") + private static final String DATA_FILE_NAME_ = + "/jdk/internal/icu/impl/data/icudt" + + VersionInfo.ICU_DATA_VERSION_PATH + + "/uprops.icu"; + + /** + * Shift value for lead surrogate to form a supplementary character. + */ + private static final int LEAD_SURROGATE_SHIFT_ = 10; + /** + * Offset to add to combined surrogate pair to avoid masking. + */ + private static final int SURROGATE_OFFSET_ = + UTF16.SUPPLEMENTARY_MIN_VALUE - + (UTF16.SURROGATE_MIN_VALUE << + LEAD_SURROGATE_SHIFT_) - + UTF16.TRAIL_SURROGATE_MIN_VALUE; + + + // property data constants ------------------------------------------------- + + /** + * Numeric types and values in the main properties words. + */ + private static final int NUMERIC_TYPE_VALUE_SHIFT_ = 6; + private static final int getNumericTypeValue(int props) { + return props >> NUMERIC_TYPE_VALUE_SHIFT_; + } + + /* constants for the storage form of numeric types and values */ + /** No numeric value. */ + private static final int NTV_NONE_ = 0; + /** Decimal digits: nv=0..9 */ + private static final int NTV_DECIMAL_START_ = 1; + /** Other digits: nv=0..9 */ + private static final int NTV_DIGIT_START_ = 11; + /** Small integers: nv=0..154 */ + private static final int NTV_NUMERIC_START_ = 21; + + private static final int ntvGetType(int ntv) { + return + (ntv==NTV_NONE_) ? NumericType.NONE : + (ntv expectedTrieLength) { + throw new IOException("uprops.icu: not enough bytes for main trie"); + } + // skip padding after trie bytes + ICUBinary.skipBytes(bytes, expectedTrieLength - trieLength); + + // skip unused intervening data structures + ICUBinary.skipBytes(bytes, (additionalOffset - propertyOffset) * 4); + + if(m_additionalColumnsCount_ > 0) { + // reads the additional property block + m_additionalTrie_ = Trie2_16.createFromSerialized(bytes); + expectedTrieLength = (additionalVectorsOffset-additionalOffset)*4; + trieLength = m_additionalTrie_.getSerializedLength(); + if(trieLength > expectedTrieLength) { + throw new IOException("uprops.icu: not enough bytes for additional-properties trie"); + } + // skip padding after trie bytes + ICUBinary.skipBytes(bytes, expectedTrieLength - trieLength); + + // additional properties + int size = scriptExtensionsOffset - additionalVectorsOffset; + m_additionalVectors_ = new int[size]; + for (int i = 0; i < size; i ++) { + m_additionalVectors_[i] = bytes.getInt(); + } + } + + // Script_Extensions + int numChars = (reservedOffset7 - scriptExtensionsOffset) * 2; + if(numChars > 0) { + m_scriptExtensions_ = new char[numChars]; + for(int i = 0; i < numChars; ++i) { + m_scriptExtensions_[i] = bytes.getChar(); + } + } + } + + private static final class IsAcceptable implements ICUBinary.Authenticate { + // @Override when we switch to Java 6 + public boolean isDataVersionAcceptable(byte version[]) { + return version[0] == 7; + } + } + + private static final int DATA_FORMAT = 0x5550726F; // "UPro" + + public void upropsvec_addPropertyStarts(UnicodeSet set) { + /* add the start code point of each same-value range of the properties vectors trie */ + if(m_additionalColumnsCount_>0) { + /* if m_additionalColumnsCount_==0 then the properties vectors trie may not be there at all */ + Iterator trieIterator = m_additionalTrie_.iterator(); + Trie2.Range range; + while(trieIterator.hasNext() && !(range=trieIterator.next()).leadSurrogate) { + set.add(range.startCodePoint); + } + } + } + + // This static initializer block must be placed after + // other static member initialization + static { + try { + INSTANCE = new UCharacterProperty(); + } + catch (IOException e) { + throw new MissingResourceException(e.getMessage(),DATA_FILE_NAME_,""); + } + } + + + // Moved from UProperty.java + /** + * Enumerated property Bidi_Paired_Bracket_Type (new in Unicode 6.3). + * Used in UAX #9: Unicode Bidirectional Algorithm + * (http://www.unicode.org/reports/tr9/) + * Returns UCharacter.BidiPairedBracketType values. + * @stable ICU 52 + */ + public static final int BIDI_PAIRED_BRACKET_TYPE = 0x1015; + +} --- old/src/java.base/share/classes/sun/text/normalizer/UnicodeSetStringSpan.java 2020-01-10 15:57:42.000000000 -0800 +++ /dev/null 2020-01-10 15:57:43.000000000 -0800 @@ -1,1165 +0,0 @@ -/* - * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved. - * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. - * - * This code is free software; you can redistribute it and/or modify it - * under the terms of the GNU General Public License version 2 only, as - * published by the Free Software Foundation. Oracle designates this - * particular file as subject to the "Classpath" exception as provided - * by Oracle in the LICENSE file that accompanied this code. - * - * This code is distributed in the hope that it will be useful, but WITHOUT - * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or - * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License - * version 2 for more details (a copy is included in the LICENSE file that - * accompanied this code). - * - * You should have received a copy of the GNU General Public License version - * 2 along with this work; if not, write to the Free Software Foundation, - * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. - * - * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA - * or visit www.oracle.com if you need additional information or have any - * questions. - */ - -/* - ****************************************************************************** - * - * Copyright (C) 2009-2014, International Business Machines - * Corporation and others. All Rights Reserved. - * - ****************************************************************************** - */ - -package sun.text.normalizer; - -import java.util.ArrayList; - -import sun.text.normalizer.UnicodeSet.SpanCondition; - -/* - * Implement span() etc. for a set with strings. - * Avoid recursion because of its exponential complexity. - * Instead, try multiple paths at once and track them with an IndexList. - */ -class UnicodeSetStringSpan { - - /* - * Which span() variant will be used? The object is either built for one variant and used once, - * or built for all and may be used many times. - */ - public static final int WITH_COUNT = 0x40; // spanAndCount() may be called - public static final int FWD = 0x20; - public static final int BACK = 0x10; - // public static final int UTF16 = 8; - public static final int CONTAINED = 2; - public static final int NOT_CONTAINED = 1; - - public static final int ALL = 0x7f; - - public static final int FWD_UTF16_CONTAINED = FWD | /* UTF16 | */ CONTAINED; - public static final int FWD_UTF16_NOT_CONTAINED = FWD | /* UTF16 | */NOT_CONTAINED; - public static final int BACK_UTF16_CONTAINED = BACK | /* UTF16 | */ CONTAINED; - public static final int BACK_UTF16_NOT_CONTAINED = BACK | /* UTF16 | */NOT_CONTAINED; - - /** - * Special spanLength short values. (since Java has not unsigned byte type) - * All code points in the string are contained in the parent set. - */ - static final short ALL_CP_CONTAINED = 0xff; - - /** The spanLength is >=0xfe. */ - static final short LONG_SPAN = ALL_CP_CONTAINED - 1; - - /** Set for span(). Same as parent but without strings. */ - private UnicodeSet spanSet; - - /** - * Set for span(not contained). - * Same as spanSet, plus characters that start or end strings. - */ - private UnicodeSet spanNotSet; - - /** The strings of the parent set. */ - private ArrayList strings; - - /** The lengths of span(), spanBack() etc. for each string. */ - private short[] spanLengths; - - /** Maximum lengths of relevant strings. */ - private int maxLength16; - - /** Are there strings that are not fully contained in the code point set? */ - private boolean someRelevant; - - /** Set up for all variants of span()? */ - private boolean all; - - /** Span helper */ - private OffsetList offsets; - - /** - * Constructs for all variants of span(), or only for any one variant. - * Initializes as little as possible, for single use. - */ - public UnicodeSetStringSpan(final UnicodeSet set, final ArrayList setStrings, int which) { - spanSet = new UnicodeSet(0, 0x10ffff); - // TODO: With Java 6, just take the parent set's strings as is, - // as a NavigableSet, rather than as an ArrayList copy of the set of strings. - // Then iterate via the first() and higher() methods. - // (We do not want to create multiple Iterator objects in each span().) - // See ICU ticket #7454. - strings = setStrings; - all = (which == ALL); - spanSet.retainAll(set); - if (0 != (which & NOT_CONTAINED)) { - // Default to the same sets. - // addToSpanNotSet() will create a separate set if necessary. - spanNotSet = spanSet; - } - offsets = new OffsetList(); - - // Determine if the strings even need to be taken into account at all for span() etc. - // If any string is relevant, then all strings need to be used for - // span(longest match) but only the relevant ones for span(while contained). - // TODO: Possible optimization: Distinguish CONTAINED vs. LONGEST_MATCH - // and do not store UTF-8 strings if !thisRelevant and CONTAINED. - // (Only store irrelevant UTF-8 strings for LONGEST_MATCH where they are relevant after all.) - // Also count the lengths of the UTF-8 versions of the strings for memory allocation. - int stringsLength = strings.size(); - - int i, spanLength; - someRelevant = false; - for (i = 0; i < stringsLength; ++i) { - String string = strings.get(i); - int length16 = string.length(); - spanLength = spanSet.span(string, SpanCondition.CONTAINED); - if (spanLength < length16) { // Relevant string. - someRelevant = true; - } - if (/* (0 != (which & UTF16)) && */ length16 > maxLength16) { - maxLength16 = length16; - } - } - if (!someRelevant && (which & WITH_COUNT) == 0) { - return; - } - - // Freeze after checking for the need to use strings at all because freezing - // a set takes some time and memory which are wasted if there are no relevant strings. - if (all) { - spanSet.freeze(); - } - - int spanBackLengthsOffset; - - // Allocate a block of meta data. - int allocSize; - if (all) { - // 2 sets of span lengths - allocSize = stringsLength * (2); - } else { - allocSize = stringsLength; // One set of span lengths. - } - spanLengths = new short[allocSize]; - - if (all) { - // Store span lengths for all span() variants. - spanBackLengthsOffset = stringsLength; - } else { - // Store span lengths for only one span() variant. - spanBackLengthsOffset = 0; - } - - // Set the meta data and spanNotSet and write the UTF-8 strings. - - for (i = 0; i < stringsLength; ++i) { - String string = strings.get(i); - int length16 = string.length(); - spanLength = spanSet.span(string, SpanCondition.CONTAINED); - if (spanLength < length16) { // Relevant string. - if (true /* 0 != (which & UTF16) */) { - if (0 != (which & CONTAINED)) { - if (0 != (which & FWD)) { - spanLengths[i] = makeSpanLengthByte(spanLength); - } - if (0 != (which & BACK)) { - spanLength = length16 - - spanSet.spanBack(string, length16, SpanCondition.CONTAINED); - spanLengths[spanBackLengthsOffset + i] = makeSpanLengthByte(spanLength); - } - } else /* not CONTAINED, not all, but NOT_CONTAINED */{ - spanLengths[i] = spanLengths[spanBackLengthsOffset + i] = 0; // Only store a relevant/irrelevant - // flag. - } - } - if (0 != (which & NOT_CONTAINED)) { - // Add string start and end code points to the spanNotSet so that - // a span(while not contained) stops before any string. - int c; - if (0 != (which & FWD)) { - c = string.codePointAt(0); - addToSpanNotSet(c); - } - if (0 != (which & BACK)) { - c = string.codePointBefore(length16); - addToSpanNotSet(c); - } - } - } else { // Irrelevant string. - if (all) { - spanLengths[i] = spanLengths[spanBackLengthsOffset + i] = ALL_CP_CONTAINED; - } else { - // All spanXYZLengths pointers contain the same address. - spanLengths[i] = ALL_CP_CONTAINED; - } - } - } - - // Finish. - if (all) { - spanNotSet.freeze(); - } - } - - /** - * Do the strings need to be checked in span() etc.? - * - * @return true if strings need to be checked (call span() here), - * false if not (use a BMPSet for best performance). - */ - public boolean needsStringSpanUTF16() { - return someRelevant; - } - - /** For fast UnicodeSet::contains(c). */ - public boolean contains(int c) { - return spanSet.contains(c); - } - - /** - * Adds a starting or ending string character to the spanNotSet - * so that a character span ends before any string. - */ - private void addToSpanNotSet(int c) { - if (spanNotSet == null || spanNotSet == spanSet) { - if (spanSet.contains(c)) { - return; // Nothing to do. - } - spanNotSet = spanSet.cloneAsThawed(); - } - spanNotSet.add(c); - } - - /* - * Note: In span() when spanLength==0 - * (after a string match, or at the beginning after an empty code point span) - * and in spanNot() and spanNotUTF8(), - * string matching could use a binary search because all string matches are done - * from the same start index. - * - * For UTF-8, this would require a comparison function that returns UTF-16 order. - * - * This optimization should not be necessary for normal UnicodeSets because most sets have no strings, and most sets - * with strings have very few very short strings. For cases with many strings, it might be better to use a different - * API and implementation with a DFA (state machine). - */ - - /* - * Algorithm for span(SpanCondition.CONTAINED) - * - * Theoretical algorithm: - * - Iterate through the string, and at each code point boundary: - * + If the code point there is in the set, then remember to continue after it. - * + If a set string matches at the current position, then remember to continue after it. - * + Either recursively span for each code point or string match, or recursively span - * for all but the shortest one and iteratively continue the span with the shortest local match. - * + Remember the longest recursive span (the farthest end point). - * + If there is no match at the current position, - * neither for the code point there nor for any set string, - * then stop and return the longest recursive span length. - * - * Optimized implementation: - * - * (We assume that most sets will have very few very short strings. - * A span using a string-less set is extremely fast.) - * - * Create and cache a spanSet which contains all of the single code points of the original set - * but none of its strings. - * - * - Start with spanLength=spanSet.span(SpanCondition.CONTAINED). - * - Loop: - * + Try to match each set string at the end of the spanLength. - * ~ Set strings that start with set-contained code points - * must be matched with a partial overlap - * because the recursive algorithm would have tried to match them at every position. - * ~ Set strings that entirely consist of set-contained code points - * are irrelevant for span(SpanCondition.CONTAINED) - * because the recursive algorithm would continue after them anyway and - * find the longest recursive match from their end. - * ~ Rather than recursing, note each end point of a set string match. - * + If no set string matched after spanSet.span(), - * then return with where the spanSet.span() ended. - * + If at least one set string matched after spanSet.span(), - * then pop the shortest string match end point and continue the loop, - * trying to match all set strings from there. - * + If at least one more set string matched after a previous string match, then test if the - * code point after the previous string match is also contained in the set. - * Continue the loop with the shortest end point of - * either this code point or a matching set string. - * + If no more set string matched after a previous string match, - * then try another spanLength=spanSet.span(SpanCondition.CONTAINED). - * Stop if spanLength==0, otherwise continue the loop. - * - * By noting each end point of a set string match, the function visits each string position at most once and - * finishes in linear time. - * - * The recursive algorithm may visit the same string position many times - * if multiple paths lead to it and finishes in exponential time. - */ - - /* - * Algorithm for span(SIMPLE) - * - * Theoretical algorithm: - * - Iterate through the string, and at each code point boundary: - * + If the code point there is in the set, then remember to continue after it. - * + If a set string matches at the current position, then remember to continue after it. - * + Continue from the farthest match position and ignore all others. - * + If there is no match at the current position, then stop and return the current position. - * - * Optimized implementation: - * - * (Same assumption and spanSet as above.) - * - * - Start with spanLength=spanSet.span(SpanCondition.CONTAINED). - * - Loop: - * + Try to match each set string at the end of the spanLength. - * ~ Set strings that start with set-contained code points - * must be matched with a partial overlap - * because the standard algorithm would have tried to match them earlier. - * ~ Set strings that entirely consist of set-contained code points - * must be matched with a full overlap because the longest-match algorithm - * would hide set string matches that end earlier. - * Such set strings need not be matched earlier inside the code point span - * because the standard algorithm would then have - * continued after the set string match anyway. - * ~ Remember the longest set string match (farthest end point) - * from the earliest starting point. - * + If no set string matched after spanSet.span(), - * then return with where the spanSet.span() ended. - * + If at least one set string matched, - * then continue the loop after the longest match from the earliest position. - * + If no more set string matched after a previous string match, - * then try another spanLength=spanSet.span(SpanCondition.CONTAINED). - * Stop if spanLength==0, otherwise continue the loop. - */ - /** - * Spans a string. - * - * @param s The string to be spanned - * @param start The start index that the span begins - * @param spanCondition The span condition - * @return the limit (exclusive end) of the span - */ - public int span(CharSequence s, int start, SpanCondition spanCondition) { - if (spanCondition == SpanCondition.NOT_CONTAINED) { - return spanNot(s, start, null); - } - int spanLimit = spanSet.span(s, start, SpanCondition.CONTAINED); - if (spanLimit == s.length()) { - return spanLimit; - } - return spanWithStrings(s, start, spanLimit, spanCondition); - } - - /** - * Synchronized method for complicated spans using the offsets. - * Avoids synchronization for simple cases. - * - * @param spanLimit = spanSet.span(s, start, CONTAINED) - */ - private synchronized int spanWithStrings(CharSequence s, int start, int spanLimit, - SpanCondition spanCondition) { - // Consider strings; they may overlap with the span. - int initSize = 0; - if (spanCondition == SpanCondition.CONTAINED) { - // Use offset list to try all possibilities. - initSize = maxLength16; - } - offsets.setMaxLength(initSize); - int length = s.length(); - int pos = spanLimit, rest = length - spanLimit; - int spanLength = spanLimit - start; - int i, stringsLength = strings.size(); - for (;;) { - if (spanCondition == SpanCondition.CONTAINED) { - for (i = 0; i < stringsLength; ++i) { - int overlap = spanLengths[i]; - if (overlap == ALL_CP_CONTAINED) { - continue; // Irrelevant string. - } - String string = strings.get(i); - - int length16 = string.length(); - - // Try to match this string at pos-overlap..pos. - if (overlap >= LONG_SPAN) { - overlap = length16; - // While contained: No point matching fully inside the code point span. - overlap = string.offsetByCodePoints(overlap, -1); // Length of the string minus the last code - // point. - } - if (overlap > spanLength) { - overlap = spanLength; - } - int inc = length16 - overlap; // Keep overlap+inc==length16. - for (;;) { - if (inc > rest) { - break; - } - // Try to match if the increment is not listed already. - if (!offsets.containsOffset(inc) && matches16CPB(s, pos - overlap, length, string, length16)) { - if (inc == rest) { - return length; // Reached the end of the string. - } - offsets.addOffset(inc); - } - if (overlap == 0) { - break; - } - --overlap; - ++inc; - } - } - } else /* SIMPLE */{ - int maxInc = 0, maxOverlap = 0; - for (i = 0; i < stringsLength; ++i) { - int overlap = spanLengths[i]; - // For longest match, we do need to try to match even an all-contained string - // to find the match from the earliest start. - - String string = strings.get(i); - - int length16 = string.length(); - - // Try to match this string at pos-overlap..pos. - if (overlap >= LONG_SPAN) { - overlap = length16; - // Longest match: Need to match fully inside the code point span - // to find the match from the earliest start. - } - if (overlap > spanLength) { - overlap = spanLength; - } - int inc = length16 - overlap; // Keep overlap+inc==length16. - for (;;) { - if (inc > rest || overlap < maxOverlap) { - break; - } - // Try to match if the string is longer or starts earlier. - if ((overlap > maxOverlap || /* redundant overlap==maxOverlap && */inc > maxInc) - && matches16CPB(s, pos - overlap, length, string, length16)) { - maxInc = inc; // Longest match from earliest start. - maxOverlap = overlap; - break; - } - --overlap; - ++inc; - } - } - - if (maxInc != 0 || maxOverlap != 0) { - // Longest-match algorithm, and there was a string match. - // Simply continue after it. - pos += maxInc; - rest -= maxInc; - if (rest == 0) { - return length; // Reached the end of the string. - } - spanLength = 0; // Match strings from after a string match. - continue; - } - } - // Finished trying to match all strings at pos. - - if (spanLength != 0 || pos == 0) { - // The position is after an unlimited code point span (spanLength!=0), - // not after a string match. - // The only position where spanLength==0 after a span is pos==0. - // Otherwise, an unlimited code point span is only tried again when no - // strings match, and if such a non-initial span fails we stop. - if (offsets.isEmpty()) { - return pos; // No strings matched after a span. - } - // Match strings from after the next string match. - } else { - // The position is after a string match (or a single code point). - if (offsets.isEmpty()) { - // No more strings matched after a previous string match. - // Try another code point span from after the last string match. - spanLimit = spanSet.span(s, pos, SpanCondition.CONTAINED); - spanLength = spanLimit - pos; - if (spanLength == rest || // Reached the end of the string, or - spanLength == 0 // neither strings nor span progressed. - ) { - return spanLimit; - } - pos += spanLength; - rest -= spanLength; - continue; // spanLength>0: Match strings from after a span. - } else { - // Try to match only one code point from after a string match if some - // string matched beyond it, so that we try all possible positions - // and don't overshoot. - spanLength = spanOne(spanSet, s, pos, rest); - if (spanLength > 0) { - if (spanLength == rest) { - return length; // Reached the end of the string. - } - // Match strings after this code point. - // There cannot be any increments below it because UnicodeSet strings - // contain multiple code points. - pos += spanLength; - rest -= spanLength; - offsets.shift(spanLength); - spanLength = 0; - continue; // Match strings from after a single code point. - } - // Match strings from after the next string match. - } - } - int minOffset = offsets.popMinimum(null); - pos += minOffset; - rest -= minOffset; - spanLength = 0; // Match strings from after a string match. - } - } - - /** - * Spans a string and counts the smallest number of set elements on any path across the span. - * - *

For proper counting, we cannot ignore strings that are fully contained in code point spans. - * - *

If the set does not have any fully-contained strings, then we could optimize this - * like span(), but such sets are likely rare, and this is at least still linear. - * - * @param s The string to be spanned - * @param start The start index that the span begins - * @param spanCondition The span condition - * @param outCount The count - * @return the limit (exclusive end) of the span - */ - public int spanAndCount(CharSequence s, int start, SpanCondition spanCondition, - OutputInt outCount) { - if (spanCondition == SpanCondition.NOT_CONTAINED) { - return spanNot(s, start, outCount); - } - // Consider strings; they may overlap with the span, - // and they may result in a smaller count that with just code points. - if (spanCondition == SpanCondition.CONTAINED) { - return spanContainedAndCount(s, start, outCount); - } - // SIMPLE (not synchronized, does not use offsets) - int stringsLength = strings.size(); - int length = s.length(); - int pos = start; - int rest = length - start; - int count = 0; - while (rest != 0) { - // Try to match the next code point. - int cpLength = spanOne(spanSet, s, pos, rest); - int maxInc = (cpLength > 0) ? cpLength : 0; - // Try to match all of the strings. - for (int i = 0; i < stringsLength; ++i) { - String string = strings.get(i); - int length16 = string.length(); - if (maxInc < length16 && length16 <= rest && - matches16CPB(s, pos, length, string, length16)) { - maxInc = length16; - } - } - // We are done if there is no match beyond pos. - if (maxInc == 0) { - outCount.value = count; - return pos; - } - // Continue from the longest match. - ++count; - pos += maxInc; - rest -= maxInc; - } - outCount.value = count; - return pos; - } - - private synchronized int spanContainedAndCount(CharSequence s, int start, OutputInt outCount) { - // Use offset list to try all possibilities. - offsets.setMaxLength(maxLength16); - int stringsLength = strings.size(); - int length = s.length(); - int pos = start; - int rest = length - start; - int count = 0; - while (rest != 0) { - // Try to match the next code point. - int cpLength = spanOne(spanSet, s, pos, rest); - if (cpLength > 0) { - offsets.addOffsetAndCount(cpLength, count + 1); - } - // Try to match all of the strings. - for (int i = 0; i < stringsLength; ++i) { - String string = strings.get(i); - int length16 = string.length(); - // Note: If the strings were sorted by length, then we could also - // avoid trying to match if there is already a match of the same length. - if (length16 <= rest && !offsets.hasCountAtOffset(length16, count + 1) && - matches16CPB(s, pos, length, string, length16)) { - offsets.addOffsetAndCount(length16, count + 1); - } - } - // We are done if there is no match beyond pos. - if (offsets.isEmpty()) { - outCount.value = count; - return pos; - } - // Continue from the nearest match. - int minOffset = offsets.popMinimum(outCount); - count = outCount.value; - pos += minOffset; - rest -= minOffset; - } - outCount.value = count; - return pos; - } - - /** - * Span a string backwards. - * - * @param s The string to be spanned - * @param spanCondition The span condition - * @return The string index which starts the span (i.e. inclusive). - */ - public synchronized int spanBack(CharSequence s, int length, SpanCondition spanCondition) { - if (spanCondition == SpanCondition.NOT_CONTAINED) { - return spanNotBack(s, length); - } - int pos = spanSet.spanBack(s, length, SpanCondition.CONTAINED); - if (pos == 0) { - return 0; - } - int spanLength = length - pos; - - // Consider strings; they may overlap with the span. - int initSize = 0; - if (spanCondition == SpanCondition.CONTAINED) { - // Use offset list to try all possibilities. - initSize = maxLength16; - } - offsets.setMaxLength(initSize); - int i, stringsLength = strings.size(); - int spanBackLengthsOffset = 0; - if (all) { - spanBackLengthsOffset = stringsLength; - } - for (;;) { - if (spanCondition == SpanCondition.CONTAINED) { - for (i = 0; i < stringsLength; ++i) { - int overlap = spanLengths[spanBackLengthsOffset + i]; - if (overlap == ALL_CP_CONTAINED) { - continue; // Irrelevant string. - } - String string = strings.get(i); - - int length16 = string.length(); - - // Try to match this string at pos-(length16-overlap)..pos-length16. - if (overlap >= LONG_SPAN) { - overlap = length16; - // While contained: No point matching fully inside the code point span. - int len1 = 0; - len1 = string.offsetByCodePoints(0, 1); - overlap -= len1; // Length of the string minus the first code point. - } - if (overlap > spanLength) { - overlap = spanLength; - } - int dec = length16 - overlap; // Keep dec+overlap==length16. - for (;;) { - if (dec > pos) { - break; - } - // Try to match if the decrement is not listed already. - if (!offsets.containsOffset(dec) && matches16CPB(s, pos - dec, length, string, length16)) { - if (dec == pos) { - return 0; // Reached the start of the string. - } - offsets.addOffset(dec); - } - if (overlap == 0) { - break; - } - --overlap; - ++dec; - } - } - } else /* SIMPLE */{ - int maxDec = 0, maxOverlap = 0; - for (i = 0; i < stringsLength; ++i) { - int overlap = spanLengths[spanBackLengthsOffset + i]; - // For longest match, we do need to try to match even an all-contained string - // to find the match from the latest end. - - String string = strings.get(i); - - int length16 = string.length(); - - // Try to match this string at pos-(length16-overlap)..pos-length16. - if (overlap >= LONG_SPAN) { - overlap = length16; - // Longest match: Need to match fully inside the code point span - // to find the match from the latest end. - } - if (overlap > spanLength) { - overlap = spanLength; - } - int dec = length16 - overlap; // Keep dec+overlap==length16. - for (;;) { - if (dec > pos || overlap < maxOverlap) { - break; - } - // Try to match if the string is longer or ends later. - if ((overlap > maxOverlap || /* redundant overlap==maxOverlap && */dec > maxDec) - && matches16CPB(s, pos - dec, length, string, length16)) { - maxDec = dec; // Longest match from latest end. - maxOverlap = overlap; - break; - } - --overlap; - ++dec; - } - } - - if (maxDec != 0 || maxOverlap != 0) { - // Longest-match algorithm, and there was a string match. - // Simply continue before it. - pos -= maxDec; - if (pos == 0) { - return 0; // Reached the start of the string. - } - spanLength = 0; // Match strings from before a string match. - continue; - } - } - // Finished trying to match all strings at pos. - - if (spanLength != 0 || pos == length) { - // The position is before an unlimited code point span (spanLength!=0), - // not before a string match. - // The only position where spanLength==0 before a span is pos==length. - // Otherwise, an unlimited code point span is only tried again when no - // strings match, and if such a non-initial span fails we stop. - if (offsets.isEmpty()) { - return pos; // No strings matched before a span. - } - // Match strings from before the next string match. - } else { - // The position is before a string match (or a single code point). - if (offsets.isEmpty()) { - // No more strings matched before a previous string match. - // Try another code point span from before the last string match. - int oldPos = pos; - pos = spanSet.spanBack(s, oldPos, SpanCondition.CONTAINED); - spanLength = oldPos - pos; - if (pos == 0 || // Reached the start of the string, or - spanLength == 0 // neither strings nor span progressed. - ) { - return pos; - } - continue; // spanLength>0: Match strings from before a span. - } else { - // Try to match only one code point from before a string match if some - // string matched beyond it, so that we try all possible positions - // and don't overshoot. - spanLength = spanOneBack(spanSet, s, pos); - if (spanLength > 0) { - if (spanLength == pos) { - return 0; // Reached the start of the string. - } - // Match strings before this code point. - // There cannot be any decrements below it because UnicodeSet strings - // contain multiple code points. - pos -= spanLength; - offsets.shift(spanLength); - spanLength = 0; - continue; // Match strings from before a single code point. - } - // Match strings from before the next string match. - } - } - pos -= offsets.popMinimum(null); - spanLength = 0; // Match strings from before a string match. - } - } - - /** - * Algorithm for spanNot()==span(SpanCondition.NOT_CONTAINED) - * - * Theoretical algorithm: - * - Iterate through the string, and at each code point boundary: - * + If the code point there is in the set, then return with the current position. - * + If a set string matches at the current position, then return with the current position. - * - * Optimized implementation: - * - * (Same assumption as for span() above.) - * - * Create and cache a spanNotSet which contains - * all of the single code points of the original set but none of its strings. - * For each set string add its initial code point to the spanNotSet. - * (Also add its final code point for spanNotBack().) - * - * - Loop: - * + Do spanLength=spanNotSet.span(SpanCondition.NOT_CONTAINED). - * + If the current code point is in the original set, then return the current position. - * + If any set string matches at the current position, then return the current position. - * + If there is no match at the current position, neither for the code point - * there nor for any set string, then skip this code point and continue the loop. - * This happens for set-string-initial code points that were added to spanNotSet - * when there is not actually a match for such a set string. - * - * @param s The string to be spanned - * @param start The start index that the span begins - * @param outCount If not null: Receives the number of code points across the span. - * @return the limit (exclusive end) of the span - */ - private int spanNot(CharSequence s, int start, OutputInt outCount) { - int length = s.length(); - int pos = start, rest = length - start; - int stringsLength = strings.size(); - int count = 0; - do { - // Span until we find a code point from the set, - // or a code point that starts or ends some string. - int spanLimit; - if (outCount == null) { - spanLimit = spanNotSet.span(s, pos, SpanCondition.NOT_CONTAINED); - } else { - spanLimit = spanNotSet.spanAndCount(s, pos, SpanCondition.NOT_CONTAINED, outCount); - outCount.value = count = count + outCount.value; - } - if (spanLimit == length) { - return length; // Reached the end of the string. - } - pos = spanLimit; - rest = length - spanLimit; - - // Check whether the current code point is in the original set, - // without the string starts and ends. - int cpLength = spanOne(spanSet, s, pos, rest); - if (cpLength > 0) { - return pos; // There is a set element at pos. - } - - // Try to match the strings at pos. - for (int i = 0; i < stringsLength; ++i) { - if (spanLengths[i] == ALL_CP_CONTAINED) { - continue; // Irrelevant string. - } - String string = strings.get(i); - - int length16 = string.length(); - if (length16 <= rest && matches16CPB(s, pos, length, string, length16)) { - return pos; // There is a set element at pos. - } - } - - // The span(while not contained) ended on a string start/end which is - // not in the original set. Skip this code point and continue. - // cpLength<0 - pos -= cpLength; - rest += cpLength; - ++count; - } while (rest != 0); - if (outCount != null) { - outCount.value = count; - } - return length; // Reached the end of the string. - } - - private int spanNotBack(CharSequence s, int length) { - int pos = length; - int i, stringsLength = strings.size(); - do { - // Span until we find a code point from the set, - // or a code point that starts or ends some string. - pos = spanNotSet.spanBack(s, pos, SpanCondition.NOT_CONTAINED); - if (pos == 0) { - return 0; // Reached the start of the string. - } - - // Check whether the current code point is in the original set, - // without the string starts and ends. - int cpLength = spanOneBack(spanSet, s, pos); - if (cpLength > 0) { - return pos; // There is a set element at pos. - } - - // Try to match the strings at pos. - for (i = 0; i < stringsLength; ++i) { - // Use spanLengths rather than a spanLengths pointer because - // it is easier and we only need to know whether the string is irrelevant - // which is the same in either array. - if (spanLengths[i] == ALL_CP_CONTAINED) { - continue; // Irrelevant string. - } - String string = strings.get(i); - - int length16 = string.length(); - if (length16 <= pos && matches16CPB(s, pos - length16, length, string, length16)) { - return pos; // There is a set element at pos. - } - } - - // The span(while not contained) ended on a string start/end which is - // not in the original set. Skip this code point and continue. - // cpLength<0 - pos += cpLength; - } while (pos != 0); - return 0; // Reached the start of the string. - } - - static short makeSpanLengthByte(int spanLength) { - // 0xfe==UnicodeSetStringSpan::LONG_SPAN - return spanLength < LONG_SPAN ? (short) spanLength : LONG_SPAN; - } - - // Compare strings without any argument checks. Requires length>0. - private static boolean matches16(CharSequence s, int start, final String t, int length) { - int end = start + length; - while (length-- > 0) { - if (s.charAt(--end) != t.charAt(length)) { - return false; - } - } - return true; - } - - /** - * Compare 16-bit Unicode strings (which may be malformed UTF-16) - * at code point boundaries. - * That is, each edge of a match must not be in the middle of a surrogate pair. - * @param s The string to match in. - * @param start The start index of s. - * @param limit The limit of the subsequence of s being spanned. - * @param t The substring to be matched in s. - * @param tlength The length of t. - */ - static boolean matches16CPB(CharSequence s, int start, int limit, final String t, int tlength) { - return matches16(s, start, t, tlength) - && !(0 < start && Character.isHighSurrogate(s.charAt(start - 1)) && - Character.isLowSurrogate(s.charAt(start))) - && !((start + tlength) < limit && Character.isHighSurrogate(s.charAt(start + tlength - 1)) && - Character.isLowSurrogate(s.charAt(start + tlength))); - } - - /** - * Does the set contain the next code point? - * If so, return its length; otherwise return its negative length. - */ - static int spanOne(final UnicodeSet set, CharSequence s, int start, int length) { - char c = s.charAt(start); - if (c >= 0xd800 && c <= 0xdbff && length >= 2) { - char c2 = s.charAt(start + 1); - if (UTF16.isTrailSurrogate(c2)) { - int supplementary = UCharacterProperty.getRawSupplementary(c, c2); - return set.contains(supplementary) ? 2 : -2; - } - } - return set.contains(c) ? 1 : -1; - } - - static int spanOneBack(final UnicodeSet set, CharSequence s, int length) { - char c = s.charAt(length - 1); - if (c >= 0xdc00 && c <= 0xdfff && length >= 2) { - char c2 = s.charAt(length - 2); - if (UTF16.isLeadSurrogate(c2)) { - int supplementary = UCharacterProperty.getRawSupplementary(c2, c); - return set.contains(supplementary) ? 2 : -2; - } - } - return set.contains(c) ? 1 : -1; - } - - /** - * Helper class for UnicodeSetStringSpan. - * - *

List of offsets from the current position from where to try matching - * a code point or a string. - * Stores offsets rather than indexes to simplify the code and use the same list - * for both increments (in span()) and decrements (in spanBack()). - * - *

Assumption: The maximum offset is limited, and the offsets that are stored at any one time - * are relatively dense, that is, - * there are normally no gaps of hundreds or thousands of offset values. - * - *

This class optionally also tracks the minimum non-negative count for each position, - * intended to count the smallest number of elements of any path leading to that position. - * - *

The implementation uses a circular buffer of count integers, - * each indicating whether the corresponding offset is in the list, - * and its path element count. - * This avoids inserting into a sorted list of offsets (or absolute indexes) - * and physically moving part of the list. - * - *

Note: In principle, the caller should setMaxLength() to - * the maximum of the max string length and U16_LENGTH/U8_LENGTH - * to account for "long" single code points. - * - *

Note: An earlier version did not track counts and stored only byte flags. - * With boolean flags, if maxLength were guaranteed to be no more than 32 or 64, - * the list could be stored as bit flags in a single integer. - * Rather than handling a circular buffer with a start list index, - * the integer would simply be shifted when lower offsets are removed. - * UnicodeSet does not have a limit on the lengths of strings. - */ - private static final class OffsetList { - private int[] list; - private int length; - private int start; - - public OffsetList() { - list = new int[16]; // default size - } - - public void setMaxLength(int maxLength) { - if (maxLength > list.length) { - list = new int[maxLength]; - } - clear(); - } - - public void clear() { - for (int i = list.length; i-- > 0;) { - list[i] = 0; - } - start = length = 0; - } - - public boolean isEmpty() { - return (length == 0); - } - - /** - * Reduces all stored offsets by delta, used when the current position moves by delta. - * There must not be any offsets lower than delta. - * If there is an offset equal to delta, it is removed. - * - * @param delta [1..maxLength] - */ - public void shift(int delta) { - int i = start + delta; - if (i >= list.length) { - i -= list.length; - } - if (list[i] != 0) { - list[i] = 0; - --length; - } - start = i; - } - - /** - * Adds an offset. The list must not contain it yet. - * @param offset [1..maxLength] - */ - public void addOffset(int offset) { - int i = start + offset; - if (i >= list.length) { - i -= list.length; - } - assert list[i] == 0; - list[i] = 1; - ++length; - } - - /** - * Adds an offset and updates its count. - * The list may already contain the offset. - * @param offset [1..maxLength] - */ - public void addOffsetAndCount(int offset, int count) { - assert count > 0; - int i = start + offset; - if (i >= list.length) { - i -= list.length; - } - if (list[i] == 0) { - list[i] = count; - ++length; - } else if (count < list[i]) { - list[i] = count; - } - } - - /** - * @param offset [1..maxLength] - */ - public boolean containsOffset(int offset) { - int i = start + offset; - if (i >= list.length) { - i -= list.length; - } - return list[i] != 0; - } - - /** - * @param offset [1..maxLength] - */ - public boolean hasCountAtOffset(int offset, int count) { - int i = start + offset; - if (i >= list.length) { - i -= list.length; - } - int oldCount = list[i]; - return oldCount != 0 && oldCount <= count; - } - - /** - * Finds the lowest stored offset from a non-empty list, removes it, - * and reduces all other offsets by this minimum. - * @return min=[1..maxLength] - */ - public int popMinimum(OutputInt outCount) { - // Look for the next offset in list[start+1..list.length-1]. - int i = start, result; - while (++i < list.length) { - int count = list[i]; - if (count != 0) { - list[i] = 0; - --length; - result = i - start; - start = i; - if (outCount != null) { outCount.value = count; } - return result; - } - } - // i==list.length - - // Wrap around and look for the next offset in list[0..start]. - // Since the list is not empty, there will be one. - result = list.length - start; - i = 0; - int count; - while ((count = list[i]) == 0) { - ++i; - } - list[i] = 0; - --length; - start = i; - if (outCount != null) { outCount.value = count; } - return result + i; - } - } -} --- /dev/null 2020-01-10 15:57:43.000000000 -0800 +++ new/src/java.base/share/classes/jdk/internal/icu/impl/UnicodeSetStringSpan.java 2020-01-10 15:57:42.000000000 -0800 @@ -0,0 +1,1168 @@ +/* + * Copyright (c) 2015, 2020, Oracle and/or its affiliates. All rights reserved. + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This code is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License version 2 only, as + * published by the Free Software Foundation. Oracle designates this + * particular file as subject to the "Classpath" exception as provided + * by Oracle in the LICENSE file that accompanied this code. + * + * This code is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License + * version 2 for more details (a copy is included in the LICENSE file that + * accompanied this code). + * + * You should have received a copy of the GNU General Public License version + * 2 along with this work; if not, write to the Free Software Foundation, + * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. + * + * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA + * or visit www.oracle.com if you need additional information or have any + * questions. + */ + +/* + ****************************************************************************** + * + * Copyright (C) 2009-2014, International Business Machines + * Corporation and others. All Rights Reserved. + * + ****************************************************************************** + */ + +package jdk.internal.icu.impl; + +import java.util.ArrayList; + +import jdk.internal.icu.text.UTF16; +import jdk.internal.icu.text.UnicodeSet; +import jdk.internal.icu.text.UnicodeSet.SpanCondition; +import jdk.internal.icu.util.OutputInt; + +/* + * Implement span() etc. for a set with strings. + * Avoid recursion because of its exponential complexity. + * Instead, try multiple paths at once and track them with an IndexList. + */ +public class UnicodeSetStringSpan { + + /* + * Which span() variant will be used? The object is either built for one variant and used once, + * or built for all and may be used many times. + */ + public static final int WITH_COUNT = 0x40; // spanAndCount() may be called + public static final int FWD = 0x20; + public static final int BACK = 0x10; + // public static final int UTF16 = 8; + public static final int CONTAINED = 2; + public static final int NOT_CONTAINED = 1; + + public static final int ALL = 0x7f; + + public static final int FWD_UTF16_CONTAINED = FWD | /* UTF16 | */ CONTAINED; + public static final int FWD_UTF16_NOT_CONTAINED = FWD | /* UTF16 | */NOT_CONTAINED; + public static final int BACK_UTF16_CONTAINED = BACK | /* UTF16 | */ CONTAINED; + public static final int BACK_UTF16_NOT_CONTAINED = BACK | /* UTF16 | */NOT_CONTAINED; + + /** + * Special spanLength short values. (since Java has not unsigned byte type) + * All code points in the string are contained in the parent set. + */ + static final short ALL_CP_CONTAINED = 0xff; + + /** The spanLength is >=0xfe. */ + static final short LONG_SPAN = ALL_CP_CONTAINED - 1; + + /** Set for span(). Same as parent but without strings. */ + private UnicodeSet spanSet; + + /** + * Set for span(not contained). + * Same as spanSet, plus characters that start or end strings. + */ + private UnicodeSet spanNotSet; + + /** The strings of the parent set. */ + private ArrayList strings; + + /** The lengths of span(), spanBack() etc. for each string. */ + private short[] spanLengths; + + /** Maximum lengths of relevant strings. */ + private int maxLength16; + + /** Are there strings that are not fully contained in the code point set? */ + private boolean someRelevant; + + /** Set up for all variants of span()? */ + private boolean all; + + /** Span helper */ + private OffsetList offsets; + + /** + * Constructs for all variants of span(), or only for any one variant. + * Initializes as little as possible, for single use. + */ + public UnicodeSetStringSpan(final UnicodeSet set, final ArrayList setStrings, int which) { + spanSet = new UnicodeSet(0, 0x10ffff); + // TODO: With Java 6, just take the parent set's strings as is, + // as a NavigableSet, rather than as an ArrayList copy of the set of strings. + // Then iterate via the first() and higher() methods. + // (We do not want to create multiple Iterator objects in each span().) + // See ICU ticket #7454. + strings = setStrings; + all = (which == ALL); + spanSet.retainAll(set); + if (0 != (which & NOT_CONTAINED)) { + // Default to the same sets. + // addToSpanNotSet() will create a separate set if necessary. + spanNotSet = spanSet; + } + offsets = new OffsetList(); + + // Determine if the strings even need to be taken into account at all for span() etc. + // If any string is relevant, then all strings need to be used for + // span(longest match) but only the relevant ones for span(while contained). + // TODO: Possible optimization: Distinguish CONTAINED vs. LONGEST_MATCH + // and do not store UTF-8 strings if !thisRelevant and CONTAINED. + // (Only store irrelevant UTF-8 strings for LONGEST_MATCH where they are relevant after all.) + // Also count the lengths of the UTF-8 versions of the strings for memory allocation. + int stringsLength = strings.size(); + + int i, spanLength; + someRelevant = false; + for (i = 0; i < stringsLength; ++i) { + String string = strings.get(i); + int length16 = string.length(); + spanLength = spanSet.span(string, SpanCondition.CONTAINED); + if (spanLength < length16) { // Relevant string. + someRelevant = true; + } + if (/* (0 != (which & UTF16)) && */ length16 > maxLength16) { + maxLength16 = length16; + } + } + if (!someRelevant && (which & WITH_COUNT) == 0) { + return; + } + + // Freeze after checking for the need to use strings at all because freezing + // a set takes some time and memory which are wasted if there are no relevant strings. + if (all) { + spanSet.freeze(); + } + + int spanBackLengthsOffset; + + // Allocate a block of meta data. + int allocSize; + if (all) { + // 2 sets of span lengths + allocSize = stringsLength * (2); + } else { + allocSize = stringsLength; // One set of span lengths. + } + spanLengths = new short[allocSize]; + + if (all) { + // Store span lengths for all span() variants. + spanBackLengthsOffset = stringsLength; + } else { + // Store span lengths for only one span() variant. + spanBackLengthsOffset = 0; + } + + // Set the meta data and spanNotSet and write the UTF-8 strings. + + for (i = 0; i < stringsLength; ++i) { + String string = strings.get(i); + int length16 = string.length(); + spanLength = spanSet.span(string, SpanCondition.CONTAINED); + if (spanLength < length16) { // Relevant string. + if (true /* 0 != (which & UTF16) */) { + if (0 != (which & CONTAINED)) { + if (0 != (which & FWD)) { + spanLengths[i] = makeSpanLengthByte(spanLength); + } + if (0 != (which & BACK)) { + spanLength = length16 + - spanSet.spanBack(string, length16, SpanCondition.CONTAINED); + spanLengths[spanBackLengthsOffset + i] = makeSpanLengthByte(spanLength); + } + } else /* not CONTAINED, not all, but NOT_CONTAINED */{ + spanLengths[i] = spanLengths[spanBackLengthsOffset + i] = 0; // Only store a relevant/irrelevant + // flag. + } + } + if (0 != (which & NOT_CONTAINED)) { + // Add string start and end code points to the spanNotSet so that + // a span(while not contained) stops before any string. + int c; + if (0 != (which & FWD)) { + c = string.codePointAt(0); + addToSpanNotSet(c); + } + if (0 != (which & BACK)) { + c = string.codePointBefore(length16); + addToSpanNotSet(c); + } + } + } else { // Irrelevant string. + if (all) { + spanLengths[i] = spanLengths[spanBackLengthsOffset + i] = ALL_CP_CONTAINED; + } else { + // All spanXYZLengths pointers contain the same address. + spanLengths[i] = ALL_CP_CONTAINED; + } + } + } + + // Finish. + if (all) { + spanNotSet.freeze(); + } + } + + /** + * Do the strings need to be checked in span() etc.? + * + * @return true if strings need to be checked (call span() here), + * false if not (use a BMPSet for best performance). + */ + public boolean needsStringSpanUTF16() { + return someRelevant; + } + + /** For fast UnicodeSet::contains(c). */ + public boolean contains(int c) { + return spanSet.contains(c); + } + + /** + * Adds a starting or ending string character to the spanNotSet + * so that a character span ends before any string. + */ + private void addToSpanNotSet(int c) { + if (spanNotSet == null || spanNotSet == spanSet) { + if (spanSet.contains(c)) { + return; // Nothing to do. + } + spanNotSet = spanSet.cloneAsThawed(); + } + spanNotSet.add(c); + } + + /* + * Note: In span() when spanLength==0 + * (after a string match, or at the beginning after an empty code point span) + * and in spanNot() and spanNotUTF8(), + * string matching could use a binary search because all string matches are done + * from the same start index. + * + * For UTF-8, this would require a comparison function that returns UTF-16 order. + * + * This optimization should not be necessary for normal UnicodeSets because most sets have no strings, and most sets + * with strings have very few very short strings. For cases with many strings, it might be better to use a different + * API and implementation with a DFA (state machine). + */ + + /* + * Algorithm for span(SpanCondition.CONTAINED) + * + * Theoretical algorithm: + * - Iterate through the string, and at each code point boundary: + * + If the code point there is in the set, then remember to continue after it. + * + If a set string matches at the current position, then remember to continue after it. + * + Either recursively span for each code point or string match, or recursively span + * for all but the shortest one and iteratively continue the span with the shortest local match. + * + Remember the longest recursive span (the farthest end point). + * + If there is no match at the current position, + * neither for the code point there nor for any set string, + * then stop and return the longest recursive span length. + * + * Optimized implementation: + * + * (We assume that most sets will have very few very short strings. + * A span using a string-less set is extremely fast.) + * + * Create and cache a spanSet which contains all of the single code points of the original set + * but none of its strings. + * + * - Start with spanLength=spanSet.span(SpanCondition.CONTAINED). + * - Loop: + * + Try to match each set string at the end of the spanLength. + * ~ Set strings that start with set-contained code points + * must be matched with a partial overlap + * because the recursive algorithm would have tried to match them at every position. + * ~ Set strings that entirely consist of set-contained code points + * are irrelevant for span(SpanCondition.CONTAINED) + * because the recursive algorithm would continue after them anyway and + * find the longest recursive match from their end. + * ~ Rather than recursing, note each end point of a set string match. + * + If no set string matched after spanSet.span(), + * then return with where the spanSet.span() ended. + * + If at least one set string matched after spanSet.span(), + * then pop the shortest string match end point and continue the loop, + * trying to match all set strings from there. + * + If at least one more set string matched after a previous string match, then test if the + * code point after the previous string match is also contained in the set. + * Continue the loop with the shortest end point of + * either this code point or a matching set string. + * + If no more set string matched after a previous string match, + * then try another spanLength=spanSet.span(SpanCondition.CONTAINED). + * Stop if spanLength==0, otherwise continue the loop. + * + * By noting each end point of a set string match, the function visits each string position at most once and + * finishes in linear time. + * + * The recursive algorithm may visit the same string position many times + * if multiple paths lead to it and finishes in exponential time. + */ + + /* + * Algorithm for span(SIMPLE) + * + * Theoretical algorithm: + * - Iterate through the string, and at each code point boundary: + * + If the code point there is in the set, then remember to continue after it. + * + If a set string matches at the current position, then remember to continue after it. + * + Continue from the farthest match position and ignore all others. + * + If there is no match at the current position, then stop and return the current position. + * + * Optimized implementation: + * + * (Same assumption and spanSet as above.) + * + * - Start with spanLength=spanSet.span(SpanCondition.CONTAINED). + * - Loop: + * + Try to match each set string at the end of the spanLength. + * ~ Set strings that start with set-contained code points + * must be matched with a partial overlap + * because the standard algorithm would have tried to match them earlier. + * ~ Set strings that entirely consist of set-contained code points + * must be matched with a full overlap because the longest-match algorithm + * would hide set string matches that end earlier. + * Such set strings need not be matched earlier inside the code point span + * because the standard algorithm would then have + * continued after the set string match anyway. + * ~ Remember the longest set string match (farthest end point) + * from the earliest starting point. + * + If no set string matched after spanSet.span(), + * then return with where the spanSet.span() ended. + * + If at least one set string matched, + * then continue the loop after the longest match from the earliest position. + * + If no more set string matched after a previous string match, + * then try another spanLength=spanSet.span(SpanCondition.CONTAINED). + * Stop if spanLength==0, otherwise continue the loop. + */ + /** + * Spans a string. + * + * @param s The string to be spanned + * @param start The start index that the span begins + * @param spanCondition The span condition + * @return the limit (exclusive end) of the span + */ + public int span(CharSequence s, int start, SpanCondition spanCondition) { + if (spanCondition == SpanCondition.NOT_CONTAINED) { + return spanNot(s, start, null); + } + int spanLimit = spanSet.span(s, start, SpanCondition.CONTAINED); + if (spanLimit == s.length()) { + return spanLimit; + } + return spanWithStrings(s, start, spanLimit, spanCondition); + } + + /** + * Synchronized method for complicated spans using the offsets. + * Avoids synchronization for simple cases. + * + * @param spanLimit = spanSet.span(s, start, CONTAINED) + */ + private synchronized int spanWithStrings(CharSequence s, int start, int spanLimit, + SpanCondition spanCondition) { + // Consider strings; they may overlap with the span. + int initSize = 0; + if (spanCondition == SpanCondition.CONTAINED) { + // Use offset list to try all possibilities. + initSize = maxLength16; + } + offsets.setMaxLength(initSize); + int length = s.length(); + int pos = spanLimit, rest = length - spanLimit; + int spanLength = spanLimit - start; + int i, stringsLength = strings.size(); + for (;;) { + if (spanCondition == SpanCondition.CONTAINED) { + for (i = 0; i < stringsLength; ++i) { + int overlap = spanLengths[i]; + if (overlap == ALL_CP_CONTAINED) { + continue; // Irrelevant string. + } + String string = strings.get(i); + + int length16 = string.length(); + + // Try to match this string at pos-overlap..pos. + if (overlap >= LONG_SPAN) { + overlap = length16; + // While contained: No point matching fully inside the code point span. + overlap = string.offsetByCodePoints(overlap, -1); // Length of the string minus the last code + // point. + } + if (overlap > spanLength) { + overlap = spanLength; + } + int inc = length16 - overlap; // Keep overlap+inc==length16. + for (;;) { + if (inc > rest) { + break; + } + // Try to match if the increment is not listed already. + if (!offsets.containsOffset(inc) && matches16CPB(s, pos - overlap, length, string, length16)) { + if (inc == rest) { + return length; // Reached the end of the string. + } + offsets.addOffset(inc); + } + if (overlap == 0) { + break; + } + --overlap; + ++inc; + } + } + } else /* SIMPLE */{ + int maxInc = 0, maxOverlap = 0; + for (i = 0; i < stringsLength; ++i) { + int overlap = spanLengths[i]; + // For longest match, we do need to try to match even an all-contained string + // to find the match from the earliest start. + + String string = strings.get(i); + + int length16 = string.length(); + + // Try to match this string at pos-overlap..pos. + if (overlap >= LONG_SPAN) { + overlap = length16; + // Longest match: Need to match fully inside the code point span + // to find the match from the earliest start. + } + if (overlap > spanLength) { + overlap = spanLength; + } + int inc = length16 - overlap; // Keep overlap+inc==length16. + for (;;) { + if (inc > rest || overlap < maxOverlap) { + break; + } + // Try to match if the string is longer or starts earlier. + if ((overlap > maxOverlap || /* redundant overlap==maxOverlap && */inc > maxInc) + && matches16CPB(s, pos - overlap, length, string, length16)) { + maxInc = inc; // Longest match from earliest start. + maxOverlap = overlap; + break; + } + --overlap; + ++inc; + } + } + + if (maxInc != 0 || maxOverlap != 0) { + // Longest-match algorithm, and there was a string match. + // Simply continue after it. + pos += maxInc; + rest -= maxInc; + if (rest == 0) { + return length; // Reached the end of the string. + } + spanLength = 0; // Match strings from after a string match. + continue; + } + } + // Finished trying to match all strings at pos. + + if (spanLength != 0 || pos == 0) { + // The position is after an unlimited code point span (spanLength!=0), + // not after a string match. + // The only position where spanLength==0 after a span is pos==0. + // Otherwise, an unlimited code point span is only tried again when no + // strings match, and if such a non-initial span fails we stop. + if (offsets.isEmpty()) { + return pos; // No strings matched after a span. + } + // Match strings from after the next string match. + } else { + // The position is after a string match (or a single code point). + if (offsets.isEmpty()) { + // No more strings matched after a previous string match. + // Try another code point span from after the last string match. + spanLimit = spanSet.span(s, pos, SpanCondition.CONTAINED); + spanLength = spanLimit - pos; + if (spanLength == rest || // Reached the end of the string, or + spanLength == 0 // neither strings nor span progressed. + ) { + return spanLimit; + } + pos += spanLength; + rest -= spanLength; + continue; // spanLength>0: Match strings from after a span. + } else { + // Try to match only one code point from after a string match if some + // string matched beyond it, so that we try all possible positions + // and don't overshoot. + spanLength = spanOne(spanSet, s, pos, rest); + if (spanLength > 0) { + if (spanLength == rest) { + return length; // Reached the end of the string. + } + // Match strings after this code point. + // There cannot be any increments below it because UnicodeSet strings + // contain multiple code points. + pos += spanLength; + rest -= spanLength; + offsets.shift(spanLength); + spanLength = 0; + continue; // Match strings from after a single code point. + } + // Match strings from after the next string match. + } + } + int minOffset = offsets.popMinimum(null); + pos += minOffset; + rest -= minOffset; + spanLength = 0; // Match strings from after a string match. + } + } + + /** + * Spans a string and counts the smallest number of set elements on any path across the span. + * + *

For proper counting, we cannot ignore strings that are fully contained in code point spans. + * + *

If the set does not have any fully-contained strings, then we could optimize this + * like span(), but such sets are likely rare, and this is at least still linear. + * + * @param s The string to be spanned + * @param start The start index that the span begins + * @param spanCondition The span condition + * @param outCount The count + * @return the limit (exclusive end) of the span + */ + public int spanAndCount(CharSequence s, int start, SpanCondition spanCondition, + OutputInt outCount) { + if (spanCondition == SpanCondition.NOT_CONTAINED) { + return spanNot(s, start, outCount); + } + // Consider strings; they may overlap with the span, + // and they may result in a smaller count that with just code points. + if (spanCondition == SpanCondition.CONTAINED) { + return spanContainedAndCount(s, start, outCount); + } + // SIMPLE (not synchronized, does not use offsets) + int stringsLength = strings.size(); + int length = s.length(); + int pos = start; + int rest = length - start; + int count = 0; + while (rest != 0) { + // Try to match the next code point. + int cpLength = spanOne(spanSet, s, pos, rest); + int maxInc = (cpLength > 0) ? cpLength : 0; + // Try to match all of the strings. + for (int i = 0; i < stringsLength; ++i) { + String string = strings.get(i); + int length16 = string.length(); + if (maxInc < length16 && length16 <= rest && + matches16CPB(s, pos, length, string, length16)) { + maxInc = length16; + } + } + // We are done if there is no match beyond pos. + if (maxInc == 0) { + outCount.value = count; + return pos; + } + // Continue from the longest match. + ++count; + pos += maxInc; + rest -= maxInc; + } + outCount.value = count; + return pos; + } + + private synchronized int spanContainedAndCount(CharSequence s, int start, OutputInt outCount) { + // Use offset list to try all possibilities. + offsets.setMaxLength(maxLength16); + int stringsLength = strings.size(); + int length = s.length(); + int pos = start; + int rest = length - start; + int count = 0; + while (rest != 0) { + // Try to match the next code point. + int cpLength = spanOne(spanSet, s, pos, rest); + if (cpLength > 0) { + offsets.addOffsetAndCount(cpLength, count + 1); + } + // Try to match all of the strings. + for (int i = 0; i < stringsLength; ++i) { + String string = strings.get(i); + int length16 = string.length(); + // Note: If the strings were sorted by length, then we could also + // avoid trying to match if there is already a match of the same length. + if (length16 <= rest && !offsets.hasCountAtOffset(length16, count + 1) && + matches16CPB(s, pos, length, string, length16)) { + offsets.addOffsetAndCount(length16, count + 1); + } + } + // We are done if there is no match beyond pos. + if (offsets.isEmpty()) { + outCount.value = count; + return pos; + } + // Continue from the nearest match. + int minOffset = offsets.popMinimum(outCount); + count = outCount.value; + pos += minOffset; + rest -= minOffset; + } + outCount.value = count; + return pos; + } + + /** + * Span a string backwards. + * + * @param s The string to be spanned + * @param spanCondition The span condition + * @return The string index which starts the span (i.e. inclusive). + */ + public synchronized int spanBack(CharSequence s, int length, SpanCondition spanCondition) { + if (spanCondition == SpanCondition.NOT_CONTAINED) { + return spanNotBack(s, length); + } + int pos = spanSet.spanBack(s, length, SpanCondition.CONTAINED); + if (pos == 0) { + return 0; + } + int spanLength = length - pos; + + // Consider strings; they may overlap with the span. + int initSize = 0; + if (spanCondition == SpanCondition.CONTAINED) { + // Use offset list to try all possibilities. + initSize = maxLength16; + } + offsets.setMaxLength(initSize); + int i, stringsLength = strings.size(); + int spanBackLengthsOffset = 0; + if (all) { + spanBackLengthsOffset = stringsLength; + } + for (;;) { + if (spanCondition == SpanCondition.CONTAINED) { + for (i = 0; i < stringsLength; ++i) { + int overlap = spanLengths[spanBackLengthsOffset + i]; + if (overlap == ALL_CP_CONTAINED) { + continue; // Irrelevant string. + } + String string = strings.get(i); + + int length16 = string.length(); + + // Try to match this string at pos-(length16-overlap)..pos-length16. + if (overlap >= LONG_SPAN) { + overlap = length16; + // While contained: No point matching fully inside the code point span. + int len1 = 0; + len1 = string.offsetByCodePoints(0, 1); + overlap -= len1; // Length of the string minus the first code point. + } + if (overlap > spanLength) { + overlap = spanLength; + } + int dec = length16 - overlap; // Keep dec+overlap==length16. + for (;;) { + if (dec > pos) { + break; + } + // Try to match if the decrement is not listed already. + if (!offsets.containsOffset(dec) && matches16CPB(s, pos - dec, length, string, length16)) { + if (dec == pos) { + return 0; // Reached the start of the string. + } + offsets.addOffset(dec); + } + if (overlap == 0) { + break; + } + --overlap; + ++dec; + } + } + } else /* SIMPLE */{ + int maxDec = 0, maxOverlap = 0; + for (i = 0; i < stringsLength; ++i) { + int overlap = spanLengths[spanBackLengthsOffset + i]; + // For longest match, we do need to try to match even an all-contained string + // to find the match from the latest end. + + String string = strings.get(i); + + int length16 = string.length(); + + // Try to match this string at pos-(length16-overlap)..pos-length16. + if (overlap >= LONG_SPAN) { + overlap = length16; + // Longest match: Need to match fully inside the code point span + // to find the match from the latest end. + } + if (overlap > spanLength) { + overlap = spanLength; + } + int dec = length16 - overlap; // Keep dec+overlap==length16. + for (;;) { + if (dec > pos || overlap < maxOverlap) { + break; + } + // Try to match if the string is longer or ends later. + if ((overlap > maxOverlap || /* redundant overlap==maxOverlap && */dec > maxDec) + && matches16CPB(s, pos - dec, length, string, length16)) { + maxDec = dec; // Longest match from latest end. + maxOverlap = overlap; + break; + } + --overlap; + ++dec; + } + } + + if (maxDec != 0 || maxOverlap != 0) { + // Longest-match algorithm, and there was a string match. + // Simply continue before it. + pos -= maxDec; + if (pos == 0) { + return 0; // Reached the start of the string. + } + spanLength = 0; // Match strings from before a string match. + continue; + } + } + // Finished trying to match all strings at pos. + + if (spanLength != 0 || pos == length) { + // The position is before an unlimited code point span (spanLength!=0), + // not before a string match. + // The only position where spanLength==0 before a span is pos==length. + // Otherwise, an unlimited code point span is only tried again when no + // strings match, and if such a non-initial span fails we stop. + if (offsets.isEmpty()) { + return pos; // No strings matched before a span. + } + // Match strings from before the next string match. + } else { + // The position is before a string match (or a single code point). + if (offsets.isEmpty()) { + // No more strings matched before a previous string match. + // Try another code point span from before the last string match. + int oldPos = pos; + pos = spanSet.spanBack(s, oldPos, SpanCondition.CONTAINED); + spanLength = oldPos - pos; + if (pos == 0 || // Reached the start of the string, or + spanLength == 0 // neither strings nor span progressed. + ) { + return pos; + } + continue; // spanLength>0: Match strings from before a span. + } else { + // Try to match only one code point from before a string match if some + // string matched beyond it, so that we try all possible positions + // and don't overshoot. + spanLength = spanOneBack(spanSet, s, pos); + if (spanLength > 0) { + if (spanLength == pos) { + return 0; // Reached the start of the string. + } + // Match strings before this code point. + // There cannot be any decrements below it because UnicodeSet strings + // contain multiple code points. + pos -= spanLength; + offsets.shift(spanLength); + spanLength = 0; + continue; // Match strings from before a single code point. + } + // Match strings from before the next string match. + } + } + pos -= offsets.popMinimum(null); + spanLength = 0; // Match strings from before a string match. + } + } + + /** + * Algorithm for spanNot()==span(SpanCondition.NOT_CONTAINED) + * + * Theoretical algorithm: + * - Iterate through the string, and at each code point boundary: + * + If the code point there is in the set, then return with the current position. + * + If a set string matches at the current position, then return with the current position. + * + * Optimized implementation: + * + * (Same assumption as for span() above.) + * + * Create and cache a spanNotSet which contains + * all of the single code points of the original set but none of its strings. + * For each set string add its initial code point to the spanNotSet. + * (Also add its final code point for spanNotBack().) + * + * - Loop: + * + Do spanLength=spanNotSet.span(SpanCondition.NOT_CONTAINED). + * + If the current code point is in the original set, then return the current position. + * + If any set string matches at the current position, then return the current position. + * + If there is no match at the current position, neither for the code point + * there nor for any set string, then skip this code point and continue the loop. + * This happens for set-string-initial code points that were added to spanNotSet + * when there is not actually a match for such a set string. + * + * @param s The string to be spanned + * @param start The start index that the span begins + * @param outCount If not null: Receives the number of code points across the span. + * @return the limit (exclusive end) of the span + */ + private int spanNot(CharSequence s, int start, OutputInt outCount) { + int length = s.length(); + int pos = start, rest = length - start; + int stringsLength = strings.size(); + int count = 0; + do { + // Span until we find a code point from the set, + // or a code point that starts or ends some string. + int spanLimit; + if (outCount == null) { + spanLimit = spanNotSet.span(s, pos, SpanCondition.NOT_CONTAINED); + } else { + spanLimit = spanNotSet.spanAndCount(s, pos, SpanCondition.NOT_CONTAINED, outCount); + outCount.value = count = count + outCount.value; + } + if (spanLimit == length) { + return length; // Reached the end of the string. + } + pos = spanLimit; + rest = length - spanLimit; + + // Check whether the current code point is in the original set, + // without the string starts and ends. + int cpLength = spanOne(spanSet, s, pos, rest); + if (cpLength > 0) { + return pos; // There is a set element at pos. + } + + // Try to match the strings at pos. + for (int i = 0; i < stringsLength; ++i) { + if (spanLengths[i] == ALL_CP_CONTAINED) { + continue; // Irrelevant string. + } + String string = strings.get(i); + + int length16 = string.length(); + if (length16 <= rest && matches16CPB(s, pos, length, string, length16)) { + return pos; // There is a set element at pos. + } + } + + // The span(while not contained) ended on a string start/end which is + // not in the original set. Skip this code point and continue. + // cpLength<0 + pos -= cpLength; + rest += cpLength; + ++count; + } while (rest != 0); + if (outCount != null) { + outCount.value = count; + } + return length; // Reached the end of the string. + } + + private int spanNotBack(CharSequence s, int length) { + int pos = length; + int i, stringsLength = strings.size(); + do { + // Span until we find a code point from the set, + // or a code point that starts or ends some string. + pos = spanNotSet.spanBack(s, pos, SpanCondition.NOT_CONTAINED); + if (pos == 0) { + return 0; // Reached the start of the string. + } + + // Check whether the current code point is in the original set, + // without the string starts and ends. + int cpLength = spanOneBack(spanSet, s, pos); + if (cpLength > 0) { + return pos; // There is a set element at pos. + } + + // Try to match the strings at pos. + for (i = 0; i < stringsLength; ++i) { + // Use spanLengths rather than a spanLengths pointer because + // it is easier and we only need to know whether the string is irrelevant + // which is the same in either array. + if (spanLengths[i] == ALL_CP_CONTAINED) { + continue; // Irrelevant string. + } + String string = strings.get(i); + + int length16 = string.length(); + if (length16 <= pos && matches16CPB(s, pos - length16, length, string, length16)) { + return pos; // There is a set element at pos. + } + } + + // The span(while not contained) ended on a string start/end which is + // not in the original set. Skip this code point and continue. + // cpLength<0 + pos += cpLength; + } while (pos != 0); + return 0; // Reached the start of the string. + } + + static short makeSpanLengthByte(int spanLength) { + // 0xfe==UnicodeSetStringSpan::LONG_SPAN + return spanLength < LONG_SPAN ? (short) spanLength : LONG_SPAN; + } + + // Compare strings without any argument checks. Requires length>0. + private static boolean matches16(CharSequence s, int start, final String t, int length) { + int end = start + length; + while (length-- > 0) { + if (s.charAt(--end) != t.charAt(length)) { + return false; + } + } + return true; + } + + /** + * Compare 16-bit Unicode strings (which may be malformed UTF-16) + * at code point boundaries. + * That is, each edge of a match must not be in the middle of a surrogate pair. + * @param s The string to match in. + * @param start The start index of s. + * @param limit The limit of the subsequence of s being spanned. + * @param t The substring to be matched in s. + * @param tlength The length of t. + */ + static boolean matches16CPB(CharSequence s, int start, int limit, final String t, int tlength) { + return matches16(s, start, t, tlength) + && !(0 < start && Character.isHighSurrogate(s.charAt(start - 1)) && + Character.isLowSurrogate(s.charAt(start))) + && !((start + tlength) < limit && Character.isHighSurrogate(s.charAt(start + tlength - 1)) && + Character.isLowSurrogate(s.charAt(start + tlength))); + } + + /** + * Does the set contain the next code point? + * If so, return its length; otherwise return its negative length. + */ + static int spanOne(final UnicodeSet set, CharSequence s, int start, int length) { + char c = s.charAt(start); + if (c >= 0xd800 && c <= 0xdbff && length >= 2) { + char c2 = s.charAt(start + 1); + if (UTF16.isTrailSurrogate(c2)) { + int supplementary = UCharacterProperty.getRawSupplementary(c, c2); + return set.contains(supplementary) ? 2 : -2; + } + } + return set.contains(c) ? 1 : -1; + } + + static int spanOneBack(final UnicodeSet set, CharSequence s, int length) { + char c = s.charAt(length - 1); + if (c >= 0xdc00 && c <= 0xdfff && length >= 2) { + char c2 = s.charAt(length - 2); + if (UTF16.isLeadSurrogate(c2)) { + int supplementary = UCharacterProperty.getRawSupplementary(c2, c); + return set.contains(supplementary) ? 2 : -2; + } + } + return set.contains(c) ? 1 : -1; + } + + /** + * Helper class for UnicodeSetStringSpan. + * + *

List of offsets from the current position from where to try matching + * a code point or a string. + * Stores offsets rather than indexes to simplify the code and use the same list + * for both increments (in span()) and decrements (in spanBack()). + * + *

Assumption: The maximum offset is limited, and the offsets that are stored at any one time + * are relatively dense, that is, + * there are normally no gaps of hundreds or thousands of offset values. + * + *

This class optionally also tracks the minimum non-negative count for each position, + * intended to count the smallest number of elements of any path leading to that position. + * + *

The implementation uses a circular buffer of count integers, + * each indicating whether the corresponding offset is in the list, + * and its path element count. + * This avoids inserting into a sorted list of offsets (or absolute indexes) + * and physically moving part of the list. + * + *

Note: In principle, the caller should setMaxLength() to + * the maximum of the max string length and U16_LENGTH/U8_LENGTH + * to account for "long" single code points. + * + *

Note: An earlier version did not track counts and stored only byte flags. + * With boolean flags, if maxLength were guaranteed to be no more than 32 or 64, + * the list could be stored as bit flags in a single integer. + * Rather than handling a circular buffer with a start list index, + * the integer would simply be shifted when lower offsets are removed. + * UnicodeSet does not have a limit on the lengths of strings. + */ + private static final class OffsetList { + private int[] list; + private int length; + private int start; + + public OffsetList() { + list = new int[16]; // default size + } + + public void setMaxLength(int maxLength) { + if (maxLength > list.length) { + list = new int[maxLength]; + } + clear(); + } + + public void clear() { + for (int i = list.length; i-- > 0;) { + list[i] = 0; + } + start = length = 0; + } + + public boolean isEmpty() { + return (length == 0); + } + + /** + * Reduces all stored offsets by delta, used when the current position moves by delta. + * There must not be any offsets lower than delta. + * If there is an offset equal to delta, it is removed. + * + * @param delta [1..maxLength] + */ + public void shift(int delta) { + int i = start + delta; + if (i >= list.length) { + i -= list.length; + } + if (list[i] != 0) { + list[i] = 0; + --length; + } + start = i; + } + + /** + * Adds an offset. The list must not contain it yet. + * @param offset [1..maxLength] + */ + public void addOffset(int offset) { + int i = start + offset; + if (i >= list.length) { + i -= list.length; + } + assert list[i] == 0; + list[i] = 1; + ++length; + } + + /** + * Adds an offset and updates its count. + * The list may already contain the offset. + * @param offset [1..maxLength] + */ + public void addOffsetAndCount(int offset, int count) { + assert count > 0; + int i = start + offset; + if (i >= list.length) { + i -= list.length; + } + if (list[i] == 0) { + list[i] = count; + ++length; + } else if (count < list[i]) { + list[i] = count; + } + } + + /** + * @param offset [1..maxLength] + */ + public boolean containsOffset(int offset) { + int i = start + offset; + if (i >= list.length) { + i -= list.length; + } + return list[i] != 0; + } + + /** + * @param offset [1..maxLength] + */ + public boolean hasCountAtOffset(int offset, int count) { + int i = start + offset; + if (i >= list.length) { + i -= list.length; + } + int oldCount = list[i]; + return oldCount != 0 && oldCount <= count; + } + + /** + * Finds the lowest stored offset from a non-empty list, removes it, + * and reduces all other offsets by this minimum. + * @return min=[1..maxLength] + */ + public int popMinimum(OutputInt outCount) { + // Look for the next offset in list[start+1..list.length-1]. + int i = start, result; + while (++i < list.length) { + int count = list[i]; + if (count != 0) { + list[i] = 0; + --length; + result = i - start; + start = i; + if (outCount != null) { outCount.value = count; } + return result; + } + } + // i==list.length + + // Wrap around and look for the next offset in list[0..start]. + // Since the list is not empty, there will be one. + result = list.length - start; + i = 0; + int count; + while ((count = list[i]) == 0) { + ++i; + } + list[i] = 0; + --length; + start = i; + if (outCount != null) { outCount.value = count; } + return result + i; + } + } +} --- old/src/java.base/share/classes/sun/text/normalizer/Utility.java 2020-01-10 15:57:44.000000000 -0800 +++ /dev/null 2020-01-10 15:57:44.000000000 -0800 @@ -1,273 +0,0 @@ -/* - * Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved. - * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. - * - * This code is free software; you can redistribute it and/or modify it - * under the terms of the GNU General Public License version 2 only, as - * published by the Free Software Foundation. Oracle designates this - * particular file as subject to the "Classpath" exception as provided - * by Oracle in the LICENSE file that accompanied this code. - * - * This code is distributed in the hope that it will be useful, but WITHOUT - * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or - * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License - * version 2 for more details (a copy is included in the LICENSE file that - * accompanied this code). - * - * You should have received a copy of the GNU General Public License version - * 2 along with this work; if not, write to the Free Software Foundation, - * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. - * - * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA - * or visit www.oracle.com if you need additional information or have any - * questions. - */ -/* - ******************************************************************************* - * Copyright (C) 1996-2011, International Business Machines Corporation and * - * others. All Rights Reserved. * - ******************************************************************************* - */ - -package sun.text.normalizer; - -import java.io.IOException; -import java.util.Locale; - -final class Utility { - - /** - * Convert characters outside the range U+0020 to U+007F to - * Unicode escapes, and convert backslash to a double backslash. - */ - public static final String escape(String s) { - StringBuilder buf = new StringBuilder(); - for (int i=0; i= ' ' && c <= 0x007F) { - if (c == '\\') { - buf.append("\\\\"); // That is, "\\" - } else { - buf.append((char)c); - } - } else { - boolean four = c <= 0xFFFF; - buf.append(four ? "\\u" : "\\U"); - buf.append(hex(c, four ? 4 : 8)); - } - } - return buf.toString(); - } - - /* This map must be in ASCENDING ORDER OF THE ESCAPE CODE */ - private static final char[] UNESCAPE_MAP = { - /*" 0x22, 0x22 */ - /*' 0x27, 0x27 */ - /*? 0x3F, 0x3F */ - /*\ 0x5C, 0x5C */ - /*a*/ 0x61, 0x07, - /*b*/ 0x62, 0x08, - /*e*/ 0x65, 0x1b, - /*f*/ 0x66, 0x0c, - /*n*/ 0x6E, 0x0a, - /*r*/ 0x72, 0x0d, - /*t*/ 0x74, 0x09, - /*v*/ 0x76, 0x0b - }; - - /** - * Convert an escape to a 32-bit code point value. We attempt - * to parallel the icu4c unescapeAt() function. - * @param offset16 an array containing offset to the character - * after the backslash. Upon return offset16[0] will - * be updated to point after the escape sequence. - * @return character value from 0 to 10FFFF, or -1 on error. - */ - public static int unescapeAt(String s, int[] offset16) { - int c; - int result = 0; - int n = 0; - int minDig = 0; - int maxDig = 0; - int bitsPerDigit = 4; - int dig; - int i; - boolean braces = false; - - /* Check that offset is in range */ - int offset = offset16[0]; - int length = s.length(); - if (offset < 0 || offset >= length) { - return -1; - } - - /* Fetch first UChar after '\\' */ - c = Character.codePointAt(s, offset); - offset += UTF16.getCharCount(c); - - /* Convert hexadecimal and octal escapes */ - switch (c) { - case 'u': - minDig = maxDig = 4; - break; - case 'U': - minDig = maxDig = 8; - break; - case 'x': - minDig = 1; - if (offset < length && UTF16.charAt(s, offset) == 0x7B /*{*/) { - ++offset; - braces = true; - maxDig = 8; - } else { - maxDig = 2; - } - break; - default: - dig = UCharacter.digit(c, 8); - if (dig >= 0) { - minDig = 1; - maxDig = 3; - n = 1; /* Already have first octal digit */ - bitsPerDigit = 3; - result = dig; - } - break; - } - if (minDig != 0) { - while (offset < length && n < maxDig) { - c = UTF16.charAt(s, offset); - dig = UCharacter.digit(c, (bitsPerDigit == 3) ? 8 : 16); - if (dig < 0) { - break; - } - result = (result << bitsPerDigit) | dig; - offset += UTF16.getCharCount(c); - ++n; - } - if (n < minDig) { - return -1; - } - if (braces) { - if (c != 0x7D /*}*/) { - return -1; - } - ++offset; - } - if (result < 0 || result >= 0x110000) { - return -1; - } - // If an escape sequence specifies a lead surrogate, see - // if there is a trail surrogate after it, either as an - // escape or as a literal. If so, join them up into a - // supplementary. - if (offset < length && - UTF16.isLeadSurrogate((char) result)) { - int ahead = offset+1; - c = s.charAt(offset); // [sic] get 16-bit code unit - if (c == '\\' && ahead < length) { - int o[] = new int[] { ahead }; - c = unescapeAt(s, o); - ahead = o[0]; - } - if (UTF16.isTrailSurrogate((char) c)) { - offset = ahead; - result = UCharacterProperty.getRawSupplementary( - (char) result, (char) c); - } - } - offset16[0] = offset; - return result; - } - - /* Convert C-style escapes in table */ - for (i=0; i= 0x20 && c <= 0x7E); - } - - /** - * Escape unprintable characters using uxxxx notation - * for U+0000 to U+FFFF and Uxxxxxxxx for U+10000 and - * above. If the character is printable ASCII, then do nothing - * and return FALSE. Otherwise, append the escaped notation and - * return TRUE. - */ - public static boolean escapeUnprintable(T result, int c) { - try { - if (isUnprintable(c)) { - result.append('\\'); - if ((c & ~0xFFFF) != 0) { - result.append('U'); - result.append(DIGITS[0xF&(c>>28)]); - result.append(DIGITS[0xF&(c>>24)]); - result.append(DIGITS[0xF&(c>>20)]); - result.append(DIGITS[0xF&(c>>16)]); - } else { - result.append('u'); - } - result.append(DIGITS[0xF&(c>>12)]); - result.append(DIGITS[0xF&(c>>8)]); - result.append(DIGITS[0xF&(c>>4)]); - result.append(DIGITS[0xF&c]); - return true; - } - return false; - } catch (IOException e) { - throw new IllegalArgumentException(e); - } - } -} --- /dev/null 2020-01-10 15:57:44.000000000 -0800 +++ new/src/java.base/share/classes/jdk/internal/icu/impl/Utility.java 2020-01-10 15:57:43.000000000 -0800 @@ -0,0 +1,276 @@ +/* + * Copyright (c) 2005, 2020, Oracle and/or its affiliates. All rights reserved. + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This code is free software; you can redistribute it and/or modify it + * under the terms of the GNU General Public License version 2 only, as + * published by the Free Software Foundation. Oracle designates this + * particular file as subject to the "Classpath" exception as provided + * by Oracle in the LICENSE file that accompanied this code. + * + * This code is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License + * version 2 for more details (a copy is included in the LICENSE file that + * accompanied this code). + * + * You should have received a copy of the GNU General Public License version + * 2 along with this work; if not, write to the Free Software Foundation, + * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. + * + * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA + * or visit www.oracle.com if you need additional information or have any + * questions. + */ +/* + ******************************************************************************* + * Copyright (C) 1996-2011, International Business Machines Corporation and * + * others. All Rights Reserved. * + ******************************************************************************* + */ + +package jdk.internal.icu.impl; + +import jdk.internal.icu.lang.UCharacter; +import jdk.internal.icu.text.UTF16; + +import java.io.IOException; +import java.util.Locale; + +public final class Utility { + + /** + * Convert characters outside the range U+0020 to U+007F to + * Unicode escapes, and convert backslash to a double backslash. + */ + public static final String escape(String s) { + StringBuilder buf = new StringBuilder(); + for (int i=0; i= ' ' && c <= 0x007F) { + if (c == '\\') { + buf.append("\\\\"); // That is, "\\" + } else { + buf.append((char)c); + } + } else { + boolean four = c <= 0xFFFF; + buf.append(four ? "\\u" : "\\U"); + buf.append(hex(c, four ? 4 : 8)); + } + } + return buf.toString(); + } + + /* This map must be in ASCENDING ORDER OF THE ESCAPE CODE */ + private static final char[] UNESCAPE_MAP = { + /*" 0x22, 0x22 */ + /*' 0x27, 0x27 */ + /*? 0x3F, 0x3F */ + /*\ 0x5C, 0x5C */ + /*a*/ 0x61, 0x07, + /*b*/ 0x62, 0x08, + /*e*/ 0x65, 0x1b, + /*f*/ 0x66, 0x0c, + /*n*/ 0x6E, 0x0a, + /*r*/ 0x72, 0x0d, + /*t*/ 0x74, 0x09, + /*v*/ 0x76, 0x0b + }; + + /** + * Convert an escape to a 32-bit code point value. We attempt + * to parallel the icu4c unescapeAt() function. + * @param offset16 an array containing offset to the character + * after the backslash. Upon return offset16[0] will + * be updated to point after the escape sequence. + * @return character value from 0 to 10FFFF, or -1 on error. + */ + public static int unescapeAt(String s, int[] offset16) { + int c; + int result = 0; + int n = 0; + int minDig = 0; + int maxDig = 0; + int bitsPerDigit = 4; + int dig; + int i; + boolean braces = false; + + /* Check that offset is in range */ + int offset = offset16[0]; + int length = s.length(); + if (offset < 0 || offset >= length) { + return -1; + } + + /* Fetch first UChar after '\\' */ + c = Character.codePointAt(s, offset); + offset += UTF16.getCharCount(c); + + /* Convert hexadecimal and octal escapes */ + switch (c) { + case 'u': + minDig = maxDig = 4; + break; + case 'U': + minDig = maxDig = 8; + break; + case 'x': + minDig = 1; + if (offset < length && UTF16.charAt(s, offset) == 0x7B /*{*/) { + ++offset; + braces = true; + maxDig = 8; + } else { + maxDig = 2; + } + break; + default: + dig = UCharacter.digit(c, 8); + if (dig >= 0) { + minDig = 1; + maxDig = 3; + n = 1; /* Already have first octal digit */ + bitsPerDigit = 3; + result = dig; + } + break; + } + if (minDig != 0) { + while (offset < length && n < maxDig) { + c = UTF16.charAt(s, offset); + dig = UCharacter.digit(c, (bitsPerDigit == 3) ? 8 : 16); + if (dig < 0) { + break; + } + result = (result << bitsPerDigit) | dig; + offset += UTF16.getCharCount(c); + ++n; + } + if (n < minDig) { + return -1; + } + if (braces) { + if (c != 0x7D /*}*/) { + return -1; + } + ++offset; + } + if (result < 0 || result >= 0x110000) { + return -1; + } + // If an escape sequence specifies a lead surrogate, see + // if there is a trail surrogate after it, either as an + // escape or as a literal. If so, join them up into a + // supplementary. + if (offset < length && + UTF16.isLeadSurrogate((char) result)) { + int ahead = offset+1; + c = s.charAt(offset); // [sic] get 16-bit code unit + if (c == '\\' && ahead < length) { + int o[] = new int[] { ahead }; + c = unescapeAt(s, o); + ahead = o[0]; + } + if (UTF16.isTrailSurrogate((char) c)) { + offset = ahead; + result = UCharacterProperty.getRawSupplementary( + (char) result, (char) c); + } + } + offset16[0] = offset; + return result; + } + + /* Convert C-style escapes in table */ + for (i=0; i= 0x20 && c <= 0x7E); + } + + /** + * Escape unprintable characters using uxxxx notation + * for U+0000 to U+FFFF and Uxxxxxxxx for U+10000 and + * above. If the character is printable ASCII, then do nothing + * and return FALSE. Otherwise, append the escaped notation and + * return TRUE. + */ + public static boolean escapeUnprintable(T result, int c) { + try { + if (isUnprintable(c)) { + result.append('\\'); + if ((c & ~0xFFFF) != 0) { + result.append('U'); + result.append(DIGITS[0xF&(c>>28)]); + result.append(DIGITS[0xF&(c>>24)]); + result.append(DIGITS[0xF&(c>>20)]); + result.append(DIGITS[0xF&(c>>16)]); + } else { + result.append('u'); + } + result.append(DIGITS[0xF&(c>>12)]); + result.append(DIGITS[0xF&(c>>8)]); + result.append(DIGITS[0xF&(c>>4)]); + result.append(DIGITS[0xF&c]); + return true; + } + return false; + } catch (IOException e) { + throw new IllegalArgumentException(e); + } + } +} Binary files old/src/java.base/share/classes/sun/text/resources/nfc.nrm and /dev/null differ Binary files /dev/null and new/src/java.base/share/classes/jdk/internal/icu/impl/data/icudt64b/nfc.nrm differ Binary files old/src/java.base/share/classes/sun/text/resources/nfkc.nrm and /dev/null differ Binary files /dev/null and new/src/java.base/share/classes/jdk/internal/icu/impl/data/icudt64b/nfkc.nrm differ Binary files old/src/java.base/share/classes/sun/text/resources/ubidi.icu and /dev/null differ Binary files /dev/null and new/src/java.base/share/classes/jdk/internal/icu/impl/data/icudt64b/ubidi.icu differ Binary files old/src/java.base/share/classes/sun/text/resources/uprops.icu and /dev/null differ Binary files /dev/null and new/src/java.base/share/classes/jdk/internal/icu/impl/data/icudt64b/uprops.icu differ --- old/src/java.base/share/classes/sun/text/normalizer/UCharacter.java 2020-01-10 15:57:48.000000000 -0800 +++ /dev/null 2020-01-10 15:57:48.000000000 -0800 @@ -1,539 +0,0 @@ -/* - * Copyright (c) 2009, 2015, Oracle and/or its affiliates. All rights reserved. - * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. - * - * This code is free software; you can redistribute it and/or modify it - * under the terms of the GNU General Public License version 2 only, as - * published by the Free Software Foundation. Oracle designates this - * particular file as subject to the "Classpath" exception as provided - * by Oracle in the LICENSE file that accompanied this code. - * - * This code is distributed in the hope that it will be useful, but WITHOUT - * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or - * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License - * version 2 for more details (a copy is included in the LICENSE file that - * accompanied this code). - * - * You should have received a copy of the GNU General Public License version - * 2 along with this work; if not, write to the Free Software Foundation, - * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. - * - * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA - * or visit www.oracle.com if you need additional information or have any - * questions. - */ - -/** -******************************************************************************* -* Copyright (C) 1996-2014, International Business Machines Corporation and -* others. All Rights Reserved. -******************************************************************************* -*/ - -package sun.text.normalizer; - -/** - *

The UCharacter class provides extensions to the - * - * java.lang.Character class. These extensions provide support for - * more Unicode properties and together with the UTF16 - * class, provide support for supplementary characters (those with code - * points above U+FFFF). - * Each ICU release supports the latest version of Unicode available at that time. - * - *

Code points are represented in these API using ints. While it would be - * more convenient in Java to have a separate primitive datatype for them, - * ints suffice in the meantime. - * - *

To use this class please add the jar file name icu4j.jar to the - * class path, since it contains data files which supply the information used - * by this file.
- * E.g. In Windows
- * set CLASSPATH=%CLASSPATH%;$JAR_FILE_PATH/ucharacter.jar.
- * Otherwise, another method would be to copy the files uprops.dat and - * unames.icu from the icu4j source subdirectory - * $ICU4J_SRC/src/com.ibm.icu.impl.data to your class directory - * $ICU4J_CLASS/com.ibm.icu.impl.data. - * - *

Aside from the additions for UTF-16 support, and the updated Unicode - * properties, the main differences between UCharacter and Character are: - *

    - *
  • UCharacter is not designed to be a char wrapper and does not have - * APIs to which involves management of that single char.
    - * These include: - *
      - *
    • char charValue(), - *
    • int compareTo(java.lang.Character, java.lang.Character), etc. - *
    - *
  • UCharacter does not include Character APIs that are deprecated, nor - * does it include the Java-specific character information, such as - * boolean isJavaIdentifierPart(char ch). - *
  • Character maps characters 'A' - 'Z' and 'a' - 'z' to the numeric - * values '10' - '35'. UCharacter also does this in digit and - * getNumericValue, to adhere to the java semantics of these - * methods. New methods unicodeDigit, and - * getUnicodeNumericValue do not treat the above code points - * as having numeric values. This is a semantic change from ICU4J 1.3.1. - *
- *

- * Further detail on differences can be determined using the program - * - * com.ibm.icu.dev.test.lang.UCharacterCompare - *

- *

- * In addition to Java compatibility functions, which calculate derived properties, - * this API provides low-level access to the Unicode Character Database. - *

- *

- * Unicode assigns each code point (not just assigned character) values for - * many properties. - * Most of them are simple boolean flags, or constants from a small enumerated list. - * For some properties, values are strings or other relatively more complex types. - *

- *

- * For more information see - * "About the Unicode Character Database" - * (http://www.unicode.org/ucd/) - * and the ICU - * User Guide chapter on Properties - * (http://www.icu-project.org/userguide/properties.html). - *

- *

- * There are also functions that provide easy migration from C/POSIX functions - * like isblank(). Their use is generally discouraged because the C/POSIX - * standards do not define their semantics beyond the ASCII range, which means - * that different implementations exhibit very different behavior. - * Instead, Unicode properties should be used directly. - *

- *

- * There are also only a few, broad C/POSIX character classes, and they tend - * to be used for conflicting purposes. For example, the "isalpha()" class - * is sometimes used to determine word boundaries, while a more sophisticated - * approach would at least distinguish initial letters from continuation - * characters (the latter including combining marks). - * (In ICU, BreakIterator is the most sophisticated API for word boundaries.) - * Another example: There is no "istitle()" class for titlecase characters. - *

- *

- * ICU 3.4 and later provides API access for all twelve C/POSIX character classes. - * ICU implements them according to the Standard Recommendations in - * Annex C: Compatibility Properties of UTS #18 Unicode Regular Expressions - * (http://www.unicode.org/reports/tr18/#Compatibility_Properties). - *

- *

- * API access for C/POSIX character classes is as follows: - * 

{@code
- * - alpha:     isUAlphabetic(c) or hasBinaryProperty(c, UProperty.ALPHABETIC)
- * - lower:     isULowercase(c) or hasBinaryProperty(c, UProperty.LOWERCASE)
- * - upper:     isUUppercase(c) or hasBinaryProperty(c, UProperty.UPPERCASE)
- * - punct:     ((1<
- * 

- * 

- * The C/POSIX character classes are also available in UnicodeSet patterns,
- * using patterns like [:graph:] or \p{graph}.
- * 

- *
- * There are several ICU (and Java) whitespace functions.
- * Comparison:
    
- * 
  •  isUWhiteSpace=UCHAR_WHITE_SPACE: Unicode White_Space property;
- *       most of general categories "Z" (separators) + most whitespace ISO controls
- *       (including no-break spaces, but excluding IS1..IS4 and ZWSP)
- * 
  •  isWhitespace: Java isWhitespace; Z + whitespace ISO controls but excluding no-break spaces
- * 
  •  isSpaceChar: just Z (including no-break spaces)

- * 

- * 

- * This class is not subclassable.
- * 

- * @author Syn Wee Quek
- * @stable ICU 2.1
- * @see com.ibm.icu.lang.UCharacterEnums
- */
-
-public final class UCharacter
-{
-
-    /**
-     * Joining Group constants.
-     * @see UProperty#JOINING_GROUP
-     * @stable ICU 2.4
-     */
-    public static interface JoiningGroup
-    {
-        /**
-         * @stable ICU 2.4
-         */
-        public static final int NO_JOINING_GROUP = 0;
-    }
-
-    /**
-     * Numeric Type constants.
-     * @see UProperty#NUMERIC_TYPE
-     * @stable ICU 2.4
-     */
-    public static interface NumericType
-    {
-        /**
-         * @stable ICU 2.4
-         */
-        public static final int NONE = 0;
-        /**
-         * @stable ICU 2.4
-         */
-        public static final int DECIMAL = 1;
-        /**
-         * @stable ICU 2.4
-         */
-        public static final int DIGIT = 2;
-        /**
-         * @stable ICU 2.4
-         */
-        public static final int NUMERIC = 3;
-        /**
-         * @stable ICU 2.4
-         */
-        public static final int COUNT = 4;
-    }
-
-    /**
-     * Hangul Syllable Type constants.
-     *
-     * @see UProperty#HANGUL_SYLLABLE_TYPE
-     * @stable ICU 2.6
-     */
-    public static interface HangulSyllableType
-    {
-        /**
-         * @stable ICU 2.6
-         */
-        public static final int NOT_APPLICABLE      = 0;   /*[NA]*/ /*See note !!*/
-        /**
-         * @stable ICU 2.6
-         */
-        public static final int LEADING_JAMO        = 1;   /*[L]*/
-        /**
-         * @stable ICU 2.6
-         */
-        public static final int VOWEL_JAMO          = 2;   /*[V]*/
-        /**
-         * @stable ICU 2.6
-         */
-        public static final int TRAILING_JAMO       = 3;   /*[T]*/
-        /**
-         * @stable ICU 2.6
-         */
-        public static final int LV_SYLLABLE         = 4;   /*[LV]*/
-        /**
-         * @stable ICU 2.6
-         */
-        public static final int LVT_SYLLABLE        = 5;   /*[LVT]*/
-        /**
-         * @stable ICU 2.6
-         */
-        public static final int COUNT               = 6;
-    }
-
-    // public data members -----------------------------------------------
-
-    /**
-     * The lowest Unicode code point value.
-     * @stable ICU 2.1
-     */
-    public static final int MIN_VALUE = UTF16.CODEPOINT_MIN_VALUE;
-
-    /**
-     * The highest Unicode code point value (scalar value) according to the
-     * Unicode Standard.
-     * This is a 21-bit value (21 bits, rounded up).

-     * Up-to-date Unicode implementation of java.lang.Character.MAX_VALUE
-     * @stable ICU 2.1
-     */
-    public static final int MAX_VALUE = UTF16.CODEPOINT_MAX_VALUE;
-
-    // public methods ----------------------------------------------------
-
-    /**
-     * Returns the numeric value of a decimal digit code point.
-     * 
This method observes the semantics of
-     * java.lang.Character.digit().  Note that this
-     * will return positive values for code points for which isDigit
-     * returns false, just like java.lang.Character.
-     * 
Semantic Change: In release 1.3.1 and
-     * prior, this did not treat the European letters as having a
-     * digit value, and also treated numeric letters and other numbers as
-     * digits.
-     * This has been changed to conform to the java semantics.
-     * 
A code point is a valid digit if and only if:
-     * 
    
-     *   
  • ch is a decimal digit or one of the european letters, and
-     *   
  • the value of ch is less than the specified radix.
-     * 

-     * @param ch the code point to query
-     * @param radix the radix
-     * @return the numeric value represented by the code point in the
-     * specified radix, or -1 if the code point is not a decimal digit
-     * or if its value is too large for the radix
-     * @stable ICU 2.1
-     */
-    public static int digit(int ch, int radix)
-    {
-        if (2 <= radix && radix <= 36) {
-            int value = digit(ch);
-            if (value < 0) {
-                // ch is not a decimal digit, try latin letters
-                value = UCharacterProperty.getEuropeanDigit(ch);
-            }
-            return (value < radix) ? value : -1;
-        } else {
-            return -1;  // invalid radix
-        }
-    }
-
-    /**
-     * Returns the numeric value of a decimal digit code point.
-     * 
This is a convenience overload of digit(int, int)
-     * that provides a decimal radix.
-     * 
Semantic Change: In release 1.3.1 and prior, this
-     * treated numeric letters and other numbers as digits.  This has
-     * been changed to conform to the java semantics.
-     * @param ch the code point to query
-     * @return the numeric value represented by the code point,
-     * or -1 if the code point is not a decimal digit or if its
-     * value is too large for a decimal radix
-     * @stable ICU 2.1
-     */
-    public static int digit(int ch)
-    {
-        return UCharacterProperty.INSTANCE.digit(ch);
-    }
-
-    /**
-     * Returns a value indicating a code point's Unicode category.
-     * Up-to-date Unicode implementation of java.lang.Character.getType()
-     * except for the above mentioned code points that had their category
-     * changed.

-     * Return results are constants from the interface
-     * UCharacterCategory

-     * NOTE: the UCharacterCategory values are not compatible with
-     * those returned by java.lang.Character.getType.  UCharacterCategory values
-     * match the ones used in ICU4C, while java.lang.Character type
-     * values, though similar, skip the value 17.

-     * @param ch code point whose type is to be determined
-     * @return category which is a value of UCharacterCategory
-     * @stable ICU 2.1
-     */
-    public static int getType(int ch)
-    {
-        return UCharacterProperty.INSTANCE.getType(ch);
-    }
-
-    /**
-     * Returns the Bidirection property of a code point.
-     * For example, 0x0041 (letter A) has the LEFT_TO_RIGHT directional
-     * property.

-     * Result returned belongs to the interface
-     * UCharacterDirection
-     * @param ch the code point to be determined its direction
-     * @return direction constant from UCharacterDirection.
-     * @stable ICU 2.1
-     */
-    public static int getDirection(int ch)
-    {
-        return UBiDiProps.INSTANCE.getClass(ch);
-    }
-
-    /**
-     * Maps the specified code point to a "mirror-image" code point.
-     * For code points with the "mirrored" property, implementations sometimes
-     * need a "poor man's" mapping to another code point such that the default
-     * glyph may serve as the mirror-image of the default glyph of the
-     * specified code point.

-     * This is useful for text conversion to and from codepages with visual
-     * order, and for displays without glyph selection capabilities.
-     * @param ch code point whose mirror is to be retrieved
-     * @return another code point that may serve as a mirror-image substitute,
-     *         or ch itself if there is no such mapping or ch does not have the
-     *         "mirrored" property
-     * @stable ICU 2.1
-     */
-    public static int getMirror(int ch)
-    {
-        return UBiDiProps.INSTANCE.getMirror(ch);
-    }
-
-    /**
-     * Maps the specified character to its paired bracket character.
-     * For Bidi_Paired_Bracket_Type!=None, this is the same as getMirror(int).
-     * Otherwise c itself is returned.
-     * See http://www.unicode.org/reports/tr9/
-     *
-     * @param c the code point to be mapped
-     * @return the paired bracket code point,
-     *         or c itself if there is no such mapping
-     *         (Bidi_Paired_Bracket_Type=None)
-     *
-     * @see UProperty#BIDI_PAIRED_BRACKET
-     * @see UProperty#BIDI_PAIRED_BRACKET_TYPE
-     * @see #getMirror(int)
-     * @stable ICU 52
-     */
-    public static int getBidiPairedBracket(int c) {
-        return UBiDiProps.INSTANCE.getPairedBracket(c);
-    }
-
-    /**
-     * Returns the combining class of the argument codepoint
-     * @param ch code point whose combining is to be retrieved
-     * @return the combining class of the codepoint
-     * @stable ICU 2.1
-     */
-    public static int getCombiningClass(int ch)
-    {
-        return Normalizer2.getNFDInstance().getCombiningClass(ch);
-    }
-
-    /**
-     * Returns the version of Unicode data used.
-     * @return the unicode version number used
-     * @stable ICU 2.1
-     */
-    public static VersionInfo getUnicodeVersion()
-    {
-        return UCharacterProperty.INSTANCE.m_unicodeVersion_;
-    }
-
-    /**
-     * Returns a code point corresponding to the two UTF16 characters.
-     * @param lead the lead char
-     * @param trail the trail char
-     * @return code point if surrogate characters are valid.
-     * @exception IllegalArgumentException thrown when argument characters do
-     *            not form a valid codepoint
-     * @stable ICU 2.1
-     */
-    public static int getCodePoint(char lead, char trail)
-    {
-        if (UTF16.isLeadSurrogate(lead) && UTF16.isTrailSurrogate(trail)) {
-            return UCharacterProperty.getRawSupplementary(lead, trail);
-        }
-        throw new IllegalArgumentException("Illegal surrogate characters");
-    }
-
-    /**
-     * Returns the "age" of the code point.

-     * 
The "age" is the Unicode version when the code point was first
-     * designated (as a non-character or for Private Use) or assigned a
-     * character.
-     * 
This can be useful to avoid emitting code points to receiving
-     * processes that do not accept newer characters.

-     * 
The data is from the UCD file DerivedAge.txt.

-     * @param ch The code point.
-     * @return the Unicode version number
-     * @stable ICU 2.6
-     */
-    public static VersionInfo getAge(int ch)
-    {
-        if (ch < MIN_VALUE || ch > MAX_VALUE) {
-            throw new IllegalArgumentException("Codepoint out of bounds");
-        }
-        return UCharacterProperty.INSTANCE.getAge(ch);
-    }
-
-    /**
-     * Returns the property value for an Unicode property type of a code point.
-     * Also returns binary and mask property values.

-     * 
Unicode, especially in version 3.2, defines many more properties than
-     * the original set in UnicodeData.txt.

-     * 
The properties APIs are intended to reflect Unicode properties as
-     * defined in the Unicode Character Database (UCD) and Unicode Technical
-     * Reports (UTR). For details about the properties see
-     * http://www.unicode.org/.

-     * 
For names of Unicode properties see the UCD file PropertyAliases.txt.
-     * 

-     * 
-     * Sample usage:
-     * int ea = UCharacter.getIntPropertyValue(c, UProperty.EAST_ASIAN_WIDTH);
-     * int ideo = UCharacter.getIntPropertyValue(c, UProperty.IDEOGRAPHIC);
-     * boolean b = (ideo == 1) ? true : false;
-     * 

-     * @param ch code point to test.
-     * @param type UProperty selector constant, identifies which binary
-     *        property to check. Must be
-     *        UProperty.BINARY_START <= type < UProperty.BINARY_LIMIT or
-     *        UProperty.INT_START <= type < UProperty.INT_LIMIT or
-     *        UProperty.MASK_START <= type < UProperty.MASK_LIMIT.
-     * @return numeric value that is directly the property value or,
-     *         for enumerated properties, corresponds to the numeric value of
-     *         the enumerated constant of the respective property value
-     *         enumeration type (cast to enum type if necessary).
-     *         Returns 0 or 1 (for false / true) for binary Unicode properties.
-     *         Returns a bit-mask for mask properties.
-     *         Returns 0 if 'type' is out of bounds or if the Unicode version
-     *         does not have data for the property at all, or not for this code
-     *         point.
-     * @see UProperty
-     * @see #hasBinaryProperty
-     * @see #getIntPropertyMinValue
-     * @see #getIntPropertyMaxValue
-     * @see #getUnicodeVersion
-     * @stable ICU 2.4
-     */
-     // for BiDiBase.java
-    public static int getIntPropertyValue(int ch, int type) {
-        return UCharacterProperty.INSTANCE.getIntPropertyValue(ch, type);
-    }
-
-    // private constructor -----------------------------------------------
-
-    /**
-     * Private constructor to prevent instantiation
-     */
-    private UCharacter() { }
-
-      /*
-       * Copied from UCharacterEnums.java
-       */
-
-        /**
-         * Character type Mn
-         * @stable ICU 2.1
-         */
-        public static final byte NON_SPACING_MARK        = 6;
-        /**
-         * Character type Me
-         * @stable ICU 2.1
-         */
-        public static final byte ENCLOSING_MARK          = 7;
-        /**
-         * Character type Mc
-         * @stable ICU 2.1
-         */
-        public static final byte COMBINING_SPACING_MARK  = 8;
-        /**
-         * Character type count
-         * @stable ICU 2.1
-         */
-        public static final byte CHAR_CATEGORY_COUNT     = 30;
-
-        /**
-         * Directional type R
-         * @stable ICU 2.1
-         */
-        public static final int RIGHT_TO_LEFT              = 1;
-        /**
-         * Directional type AL
-         * @stable ICU 2.1
-         */
-        public static final int RIGHT_TO_LEFT_ARABIC       = 13;
-}
--- /dev/null	2020-01-10 15:57:48.000000000 -0800
+++ new/src/java.base/share/classes/jdk/internal/icu/lang/UCharacter.java	2020-01-10 15:57:48.000000000 -0800
@@ -0,0 +1,545 @@
+/*
+ * Copyright (c) 2009, 2020, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.  Oracle designates this
+ * particular file as subject to the "Classpath" exception as provided
+ * by Oracle in the LICENSE file that accompanied this code.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+/**
+*******************************************************************************
+* Copyright (C) 1996-2014, International Business Machines Corporation and
+* others. All Rights Reserved.
+*******************************************************************************
+*/
+
+package jdk.internal.icu.lang;
+
+import jdk.internal.icu.impl.UBiDiProps;
+import jdk.internal.icu.impl.UCharacterProperty;
+import jdk.internal.icu.text.Normalizer2;
+import jdk.internal.icu.text.UTF16;
+import jdk.internal.icu.util.VersionInfo;
+
+/**
+ * 
The UCharacter class provides extensions to the
+ * 
+ * java.lang.Character class. These extensions provide support for
+ * more Unicode properties and together with the UTF16
+ * class, provide support for supplementary characters (those with code
+ * points above U+FFFF).
+ * Each ICU release supports the latest version of Unicode available at that time.
+ *
+ * 
Code points are represented in these API using ints. While it would be
+ * more convenient in Java to have a separate primitive datatype for them,
+ * ints suffice in the meantime.
+ *
+ * 
To use this class please add the jar file name icu4j.jar to the
+ * class path, since it contains data files which supply the information used
+ * by this file.

+ * E.g. In Windows 

+ * set CLASSPATH=%CLASSPATH%;$JAR_FILE_PATH/ucharacter.jar.

+ * Otherwise, another method would be to copy the files uprops.dat and
+ * unames.icu from the icu4j source subdirectory
+ * $ICU4J_SRC/src/com.ibm.icu.impl.data to your class directory
+ * $ICU4J_CLASS/com.ibm.icu.impl.data.
+ *
+ * 
Aside from the additions for UTF-16 support, and the updated Unicode
+ * properties, the main differences between UCharacter and Character are:
+ * 
    
+ * 
  •  UCharacter is not designed to be a char wrapper and does not have
+ *      APIs to which involves management of that single char.
    
+ *      These include:
+ *      
      
+ *        
    •  char charValue(),
+ *        
    •  int compareTo(java.lang.Character, java.lang.Character), etc.
+ *      
    
+ * 
  •  UCharacter does not include Character APIs that are deprecated, nor
+ *      does it include the Java-specific character information, such as
+ *      boolean isJavaIdentifierPart(char ch).
+ * 
  •  Character maps characters 'A' - 'Z' and 'a' - 'z' to the numeric
+ *      values '10' - '35'. UCharacter also does this in digit and
+ *      getNumericValue, to adhere to the java semantics of these
+ *      methods.  New methods unicodeDigit, and
+ *      getUnicodeNumericValue do not treat the above code points
+ *      as having numeric values.  This is a semantic change from ICU4J 1.3.1.
+ * 

+ * 

+ * Further detail on differences can be determined using the program
+ *        
+ *        com.ibm.icu.dev.test.lang.UCharacterCompare
+ * 

+ * 

+ * In addition to Java compatibility functions, which calculate derived properties,
+ * this API provides low-level access to the Unicode Character Database.
+ * 

+ * 

+ * Unicode assigns each code point (not just assigned character) values for
+ * many properties.
+ * Most of them are simple boolean flags, or constants from a small enumerated list.
+ * For some properties, values are strings or other relatively more complex types.
+ * 

+ * 

+ * For more information see
+ * "About the Unicode Character Database"
+ * (http://www.unicode.org/ucd/)
+ * and the ICU
+ * User Guide chapter on Properties
+ * (http://www.icu-project.org/userguide/properties.html).
+ * 

+ * 

+ * There are also functions that provide easy migration from C/POSIX functions
+ * like isblank(). Their use is generally discouraged because the C/POSIX
+ * standards do not define their semantics beyond the ASCII range, which means
+ * that different implementations exhibit very different behavior.
+ * Instead, Unicode properties should be used directly.
+ * 

+ * 

+ * There are also only a few, broad C/POSIX character classes, and they tend
+ * to be used for conflicting purposes. For example, the "isalpha()" class
+ * is sometimes used to determine word boundaries, while a more sophisticated
+ * approach would at least distinguish initial letters from continuation
+ * characters (the latter including combining marks).
+ * (In ICU, BreakIterator is the most sophisticated API for word boundaries.)
+ * Another example: There is no "istitle()" class for titlecase characters.
+ * 

+ * 

+ * ICU 3.4 and later provides API access for all twelve C/POSIX character classes.
+ * ICU implements them according to the Standard Recommendations in
+ * Annex C: Compatibility Properties of UTS #18 Unicode Regular Expressions
+ * (http://www.unicode.org/reports/tr18/#Compatibility_Properties).
+ * 

+ * 

+ * API access for C/POSIX character classes is as follows:
+ * 
{@code
+ * - alpha:     isUAlphabetic(c) or hasBinaryProperty(c, UProperty.ALPHABETIC)
+ * - lower:     isULowercase(c) or hasBinaryProperty(c, UProperty.LOWERCASE)
+ * - upper:     isUUppercase(c) or hasBinaryProperty(c, UProperty.UPPERCASE)
+ * - punct:     ((1<
+ * 

+ * 

+ * The C/POSIX character classes are also available in UnicodeSet patterns,
+ * using patterns like [:graph:] or \p{graph}.
+ * 

+ *
+ * There are several ICU (and Java) whitespace functions.
+ * Comparison:
    
+ * 
  •  isUWhiteSpace=UCHAR_WHITE_SPACE: Unicode White_Space property;
+ *       most of general categories "Z" (separators) + most whitespace ISO controls
+ *       (including no-break spaces, but excluding IS1..IS4 and ZWSP)
+ * 
  •  isWhitespace: Java isWhitespace; Z + whitespace ISO controls but excluding no-break spaces
+ * 
  •  isSpaceChar: just Z (including no-break spaces)

+ * 

+ * 

+ * This class is not subclassable.
+ * 

+ * @author Syn Wee Quek
+ * @stable ICU 2.1
+ * @see com.ibm.icu.lang.UCharacterEnums
+ */
+
+public final class UCharacter
+{
+
+    /**
+     * Joining Group constants.
+     * @see UProperty#JOINING_GROUP
+     * @stable ICU 2.4
+     */
+    public static interface JoiningGroup
+    {
+        /**
+         * @stable ICU 2.4
+         */
+        public static final int NO_JOINING_GROUP = 0;
+    }
+
+    /**
+     * Numeric Type constants.
+     * @see UProperty#NUMERIC_TYPE
+     * @stable ICU 2.4
+     */
+    public static interface NumericType
+    {
+        /**
+         * @stable ICU 2.4
+         */
+        public static final int NONE = 0;
+        /**
+         * @stable ICU 2.4
+         */
+        public static final int DECIMAL = 1;
+        /**
+         * @stable ICU 2.4
+         */
+        public static final int DIGIT = 2;
+        /**
+         * @stable ICU 2.4
+         */
+        public static final int NUMERIC = 3;
+        /**
+         * @stable ICU 2.4
+         */
+        public static final int COUNT = 4;
+    }
+
+    /**
+     * Hangul Syllable Type constants.
+     *
+     * @see UProperty#HANGUL_SYLLABLE_TYPE
+     * @stable ICU 2.6
+     */
+    public static interface HangulSyllableType
+    {
+        /**
+         * @stable ICU 2.6
+         */
+        public static final int NOT_APPLICABLE      = 0;   /*[NA]*/ /*See note !!*/
+        /**
+         * @stable ICU 2.6
+         */
+        public static final int LEADING_JAMO        = 1;   /*[L]*/
+        /**
+         * @stable ICU 2.6
+         */
+        public static final int VOWEL_JAMO          = 2;   /*[V]*/
+        /**
+         * @stable ICU 2.6
+         */
+        public static final int TRAILING_JAMO       = 3;   /*[T]*/
+        /**
+         * @stable ICU 2.6
+         */
+        public static final int LV_SYLLABLE         = 4;   /*[LV]*/
+        /**
+         * @stable ICU 2.6
+         */
+        public static final int LVT_SYLLABLE        = 5;   /*[LVT]*/
+        /**
+         * @stable ICU 2.6
+         */
+        public static final int COUNT               = 6;
+    }
+
+    // public data members -----------------------------------------------
+
+    /**
+     * The lowest Unicode code point value.
+     * @stable ICU 2.1
+     */
+    public static final int MIN_VALUE = UTF16.CODEPOINT_MIN_VALUE;
+
+    /**
+     * The highest Unicode code point value (scalar value) according to the
+     * Unicode Standard.
+     * This is a 21-bit value (21 bits, rounded up).

+     * Up-to-date Unicode implementation of java.lang.Character.MAX_VALUE
+     * @stable ICU 2.1
+     */
+    public static final int MAX_VALUE = UTF16.CODEPOINT_MAX_VALUE;
+
+    // public methods ----------------------------------------------------
+
+    /**
+     * Returns the numeric value of a decimal digit code point.
+     * 
This method observes the semantics of
+     * java.lang.Character.digit().  Note that this
+     * will return positive values for code points for which isDigit
+     * returns false, just like java.lang.Character.
+     * 
Semantic Change: In release 1.3.1 and
+     * prior, this did not treat the European letters as having a
+     * digit value, and also treated numeric letters and other numbers as
+     * digits.
+     * This has been changed to conform to the java semantics.
+     * 
A code point is a valid digit if and only if:
+     * 
    
+     *   
  • ch is a decimal digit or one of the european letters, and
+     *   
  • the value of ch is less than the specified radix.
+     * 

+     * @param ch the code point to query
+     * @param radix the radix
+     * @return the numeric value represented by the code point in the
+     * specified radix, or -1 if the code point is not a decimal digit
+     * or if its value is too large for the radix
+     * @stable ICU 2.1
+     */
+    public static int digit(int ch, int radix)
+    {
+        if (2 <= radix && radix <= 36) {
+            int value = digit(ch);
+            if (value < 0) {
+                // ch is not a decimal digit, try latin letters
+                value = UCharacterProperty.getEuropeanDigit(ch);
+            }
+            return (value < radix) ? value : -1;
+        } else {
+            return -1;  // invalid radix
+        }
+    }
+
+    /**
+     * Returns the numeric value of a decimal digit code point.
+     * 
This is a convenience overload of digit(int, int)
+     * that provides a decimal radix.
+     * 
Semantic Change: In release 1.3.1 and prior, this
+     * treated numeric letters and other numbers as digits.  This has
+     * been changed to conform to the java semantics.
+     * @param ch the code point to query
+     * @return the numeric value represented by the code point,
+     * or -1 if the code point is not a decimal digit or if its
+     * value is too large for a decimal radix
+     * @stable ICU 2.1
+     */
+    public static int digit(int ch)
+    {
+        return UCharacterProperty.INSTANCE.digit(ch);
+    }
+
+    /**
+     * Returns a value indicating a code point's Unicode category.
+     * Up-to-date Unicode implementation of java.lang.Character.getType()
+     * except for the above mentioned code points that had their category
+     * changed.

+     * Return results are constants from the interface
+     * UCharacterCategory

+     * NOTE: the UCharacterCategory values are not compatible with
+     * those returned by java.lang.Character.getType.  UCharacterCategory values
+     * match the ones used in ICU4C, while java.lang.Character type
+     * values, though similar, skip the value 17.

+     * @param ch code point whose type is to be determined
+     * @return category which is a value of UCharacterCategory
+     * @stable ICU 2.1
+     */
+    public static int getType(int ch)
+    {
+        return UCharacterProperty.INSTANCE.getType(ch);
+    }
+
+    /**
+     * Returns the Bidirection property of a code point.
+     * For example, 0x0041 (letter A) has the LEFT_TO_RIGHT directional
+     * property.

+     * Result returned belongs to the interface
+     * UCharacterDirection
+     * @param ch the code point to be determined its direction
+     * @return direction constant from UCharacterDirection.
+     * @stable ICU 2.1
+     */
+    public static int getDirection(int ch)
+    {
+        return UBiDiProps.INSTANCE.getClass(ch);
+    }
+
+    /**
+     * Maps the specified code point to a "mirror-image" code point.
+     * For code points with the "mirrored" property, implementations sometimes
+     * need a "poor man's" mapping to another code point such that the default
+     * glyph may serve as the mirror-image of the default glyph of the
+     * specified code point.

+     * This is useful for text conversion to and from codepages with visual
+     * order, and for displays without glyph selection capabilities.
+     * @param ch code point whose mirror is to be retrieved
+     * @return another code point that may serve as a mirror-image substitute,
+     *         or ch itself if there is no such mapping or ch does not have the
+     *         "mirrored" property
+     * @stable ICU 2.1
+     */
+    public static int getMirror(int ch)
+    {
+        return UBiDiProps.INSTANCE.getMirror(ch);
+    }
+
+    /**
+     * Maps the specified character to its paired bracket character.
+     * For Bidi_Paired_Bracket_Type!=None, this is the same as getMirror(int).
+     * Otherwise c itself is returned.
+     * See http://www.unicode.org/reports/tr9/
+     *
+     * @param c the code point to be mapped
+     * @return the paired bracket code point,
+     *         or c itself if there is no such mapping
+     *         (Bidi_Paired_Bracket_Type=None)
+     *
+     * @see UProperty#BIDI_PAIRED_BRACKET
+     * @see UProperty#BIDI_PAIRED_BRACKET_TYPE
+     * @see #getMirror(int)
+     * @stable ICU 52
+     */
+    public static int getBidiPairedBracket(int c) {
+        return UBiDiProps.INSTANCE.getPairedBracket(c);
+    }
+
+    /**
+     * Returns the combining class of the argument codepoint
+     * @param ch code point whose combining is to be retrieved
+     * @return the combining class of the codepoint
+     * @stable ICU 2.1
+     */
+    public static int getCombiningClass(int ch)
+    {
+        return Normalizer2.getNFDInstance().getCombiningClass(ch);
+    }
+
+    /**
+     * Returns the version of Unicode data used.
+     * @return the unicode version number used
+     * @stable ICU 2.1
+     */
+    public static VersionInfo getUnicodeVersion()
+    {
+        return UCharacterProperty.INSTANCE.m_unicodeVersion_;
+    }
+
+    /**
+     * Returns a code point corresponding to the two UTF16 characters.
+     * @param lead the lead char
+     * @param trail the trail char
+     * @return code point if surrogate characters are valid.
+     * @exception IllegalArgumentException thrown when argument characters do
+     *            not form a valid codepoint
+     * @stable ICU 2.1
+     */
+    public static int getCodePoint(char lead, char trail)
+    {
+        if (UTF16.isLeadSurrogate(lead) && UTF16.isTrailSurrogate(trail)) {
+            return UCharacterProperty.getRawSupplementary(lead, trail);
+        }
+        throw new IllegalArgumentException("Illegal surrogate characters");
+    }
+
+    /**
+     * Returns the "age" of the code point.

+     * 
The "age" is the Unicode version when the code point was first
+     * designated (as a non-character or for Private Use) or assigned a
+     * character.
+     * 
This can be useful to avoid emitting code points to receiving
+     * processes that do not accept newer characters.

+     * 
The data is from the UCD file DerivedAge.txt.

+     * @param ch The code point.
+     * @return the Unicode version number
+     * @stable ICU 2.6
+     */
+    public static VersionInfo getAge(int ch)
+    {
+        if (ch < MIN_VALUE || ch > MAX_VALUE) {
+            throw new IllegalArgumentException("Codepoint out of bounds");
+        }
+        return UCharacterProperty.INSTANCE.getAge(ch);
+    }
+
+    /**
+     * Returns the property value for an Unicode property type of a code point.
+     * Also returns binary and mask property values.

+     * 
Unicode, especially in version 3.2, defines many more properties than
+     * the original set in UnicodeData.txt.

+     * 
The properties APIs are intended to reflect Unicode properties as
+     * defined in the Unicode Character Database (UCD) and Unicode Technical
+     * Reports (UTR). For details about the properties see
+     * http://www.unicode.org/.

+     * 
For names of Unicode properties see the UCD file PropertyAliases.txt.
+     * 

+     * 
+     * Sample usage:
+     * int ea = UCharacter.getIntPropertyValue(c, UProperty.EAST_ASIAN_WIDTH);
+     * int ideo = UCharacter.getIntPropertyValue(c, UProperty.IDEOGRAPHIC);
+     * boolean b = (ideo == 1) ? true : false;
+     * 

+     * @param ch code point to test.
+     * @param type UProperty selector constant, identifies which binary
+     *        property to check. Must be
+     *        UProperty.BINARY_START <= type < UProperty.BINARY_LIMIT or
+     *        UProperty.INT_START <= type < UProperty.INT_LIMIT or
+     *        UProperty.MASK_START <= type < UProperty.MASK_LIMIT.
+     * @return numeric value that is directly the property value or,
+     *         for enumerated properties, corresponds to the numeric value of
+     *         the enumerated constant of the respective property value
+     *         enumeration type (cast to enum type if necessary).
+     *         Returns 0 or 1 (for false / true) for binary Unicode properties.
+     *         Returns a bit-mask for mask properties.
+     *         Returns 0 if 'type' is out of bounds or if the Unicode version
+     *         does not have data for the property at all, or not for this code
+     *         point.
+     * @see UProperty
+     * @see #hasBinaryProperty
+     * @see #getIntPropertyMinValue
+     * @see #getIntPropertyMaxValue
+     * @see #getUnicodeVersion
+     * @stable ICU 2.4
+     */
+     // for BiDiBase.java
+    public static int getIntPropertyValue(int ch, int type) {
+        return UCharacterProperty.INSTANCE.getIntPropertyValue(ch, type);
+    }
+
+    // private constructor -----------------------------------------------
+
+    /**
+     * Private constructor to prevent instantiation
+     */
+    private UCharacter() { }
+
+      /*
+       * Copied from UCharacterEnums.java
+       */
+
+        /**
+         * Character type Mn
+         * @stable ICU 2.1
+         */
+        public static final byte NON_SPACING_MARK        = 6;
+        /**
+         * Character type Me
+         * @stable ICU 2.1
+         */
+        public static final byte ENCLOSING_MARK          = 7;
+        /**
+         * Character type Mc
+         * @stable ICU 2.1
+         */
+        public static final byte COMBINING_SPACING_MARK  = 8;
+        /**
+         * Character type count
+         * @stable ICU 2.1
+         */
+        public static final byte CHAR_CATEGORY_COUNT     = 30;
+
+        /**
+         * Directional type R
+         * @stable ICU 2.1
+         */
+        public static final int RIGHT_TO_LEFT              = 1;
+        /**
+         * Directional type AL
+         * @stable ICU 2.1
+         */
+        public static final int RIGHT_TO_LEFT_ARABIC       = 13;
+}
--- old/src/java.base/share/classes/sun/net/idn/UCharacterDirection.java	2020-01-10 15:57:50.000000000 -0800
+++ /dev/null	2020-01-10 15:57:50.000000000 -0800
@@ -1,112 +0,0 @@
-/*
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.  Oracle designates this
- * particular file as subject to the "Classpath" exception as provided
- * by Oracle in the LICENSE file that accompanied this code.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- */
-/*
-/**
-*******************************************************************************
-* Copyright (C) 1996-2004, International Business Machines Corporation and    *
-* others. All Rights Reserved.                                                *
-*******************************************************************************
-*/
-// CHANGELOG
-//      2005-05-19 Edward Wang
-//          - copy this file from icu4jsrc_3_2/src/com/ibm/icu/lang/UCharacterDirection.java
-//          - move from package com.ibm.icu.lang to package sun.net.idn
-//
-
-package sun.net.idn;
-
-/**
- * Enumerated Unicode character linguistic direction constants.
- * Used as return results from UCharacter
- * 

- * This class is not subclassable
- * 

- * @author Syn Wee Quek
- * @stable ICU 2.1
- */
-
-@SuppressWarnings("deprecation")
-final class UCharacterDirection implements UCharacterEnums.ECharacterDirection {
-
-    // private constructor =========================================
-    ///CLOVER:OFF
-    /**
-     * Private constructor to prevent initialisation
-     */
-    private UCharacterDirection()
-    {
-    }
-    ///CLOVER:ON
-
-    /**
-     * Gets the name of the argument direction
-     * @param dir direction type to retrieve name
-     * @return directional name
-     * @stable ICU 2.1
-     */
-    public static String toString(int dir) {
-        switch(dir)
-            {
-            case LEFT_TO_RIGHT :
-                return "Left-to-Right";
-            case RIGHT_TO_LEFT :
-                return "Right-to-Left";
-            case EUROPEAN_NUMBER :
-                return "European Number";
-            case EUROPEAN_NUMBER_SEPARATOR :
-                return "European Number Separator";
-            case EUROPEAN_NUMBER_TERMINATOR :
-                return "European Number Terminator";
-            case ARABIC_NUMBER :
-                return "Arabic Number";
-            case COMMON_NUMBER_SEPARATOR :
-                return "Common Number Separator";
-            case BLOCK_SEPARATOR :
-                return "Paragraph Separator";
-            case SEGMENT_SEPARATOR :
-                return "Segment Separator";
-            case WHITE_SPACE_NEUTRAL :
-                return "Whitespace";
-            case OTHER_NEUTRAL :
-                return "Other Neutrals";
-            case LEFT_TO_RIGHT_EMBEDDING :
-                return "Left-to-Right Embedding";
-            case LEFT_TO_RIGHT_OVERRIDE :
-                return "Left-to-Right Override";
-            case RIGHT_TO_LEFT_ARABIC :
-                return "Right-to-Left Arabic";
-            case RIGHT_TO_LEFT_EMBEDDING :
-                return "Right-to-Left Embedding";
-            case RIGHT_TO_LEFT_OVERRIDE :
-                return "Right-to-Left Override";
-            case POP_DIRECTIONAL_FORMAT :
-                return "Pop Directional Format";
-            case DIR_NON_SPACING_MARK :
-                return "Non-Spacing Mark";
-            case BOUNDARY_NEUTRAL :
-                return "Boundary Neutral";
-            }
-        return "Unassigned";
-    }
-}
--- /dev/null	2020-01-10 15:57:50.000000000 -0800
+++ new/src/java.base/share/classes/jdk/internal/icu/lang/UCharacterDirection.java	2020-01-10 15:57:49.000000000 -0800
@@ -0,0 +1,113 @@
+/*
+ * Copyright (c) 2005, 2020, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.  Oracle designates this
+ * particular file as subject to the "Classpath" exception as provided
+ * by Oracle in the LICENSE file that accompanied this code.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+/*
+/**
+*******************************************************************************
+* Copyright (C) 1996-2004, International Business Machines Corporation and    *
+* others. All Rights Reserved.                                                *
+*******************************************************************************
+*/
+// CHANGELOG
+//      2005-05-19 Edward Wang
+//          - copy this file from icu4jsrc_3_2/src/com/ibm/icu/lang/UCharacterDirection.java
+//          - move from package com.ibm.icu.lang to package sun.net.idn
+//
+
+package jdk.internal.icu.lang;
+
+/**
+ * Enumerated Unicode character linguistic direction constants.
+ * Used as return results from UCharacter
+ * 

+ * This class is not subclassable
+ * 

+ * @author Syn Wee Quek
+ * @stable ICU 2.1
+ */
+
+@SuppressWarnings("deprecation")
+public final class UCharacterDirection implements UCharacterEnums.ECharacterDirection {
+
+    // private constructor =========================================
+    ///CLOVER:OFF
+    /**
+     * Private constructor to prevent initialisation
+     */
+    private UCharacterDirection()
+    {
+    }
+    ///CLOVER:ON
+
+    /**
+     * Gets the name of the argument direction
+     * @param dir direction type to retrieve name
+     * @return directional name
+     * @stable ICU 2.1
+     */
+    public static String toString(int dir) {
+        switch(dir)
+            {
+            case LEFT_TO_RIGHT :
+                return "Left-to-Right";
+            case RIGHT_TO_LEFT :
+                return "Right-to-Left";
+            case EUROPEAN_NUMBER :
+                return "European Number";
+            case EUROPEAN_NUMBER_SEPARATOR :
+                return "European Number Separator";
+            case EUROPEAN_NUMBER_TERMINATOR :
+                return "European Number Terminator";
+            case ARABIC_NUMBER :
+                return "Arabic Number";
+            case COMMON_NUMBER_SEPARATOR :
+                return "Common Number Separator";
+            case BLOCK_SEPARATOR :
+                return "Paragraph Separator";
+            case SEGMENT_SEPARATOR :
+                return "Segment Separator";
+            case WHITE_SPACE_NEUTRAL :
+                return "Whitespace";
+            case OTHER_NEUTRAL :
+                return "Other Neutrals";
+            case LEFT_TO_RIGHT_EMBEDDING :
+                return "Left-to-Right Embedding";
+            case LEFT_TO_RIGHT_OVERRIDE :
+                return "Left-to-Right Override";
+            case RIGHT_TO_LEFT_ARABIC :
+                return "Right-to-Left Arabic";
+            case RIGHT_TO_LEFT_EMBEDDING :
+                return "Right-to-Left Embedding";
+            case RIGHT_TO_LEFT_OVERRIDE :
+                return "Right-to-Left Override";
+            case POP_DIRECTIONAL_FORMAT :
+                return "Pop Directional Format";
+            case DIR_NON_SPACING_MARK :
+                return "Non-Spacing Mark";
+            case BOUNDARY_NEUTRAL :
+                return "Boundary Neutral";
+            }
+        return "Unassigned";
+    }
+}
--- old/src/java.base/share/classes/sun/net/idn/UCharacterEnums.java	2020-01-10 15:57:51.000000000 -0800
+++ /dev/null	2020-01-10 15:57:51.000000000 -0800
@@ -1,587 +0,0 @@
-/*
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.  Oracle designates this
- * particular file as subject to the "Classpath" exception as provided
- * by Oracle in the LICENSE file that accompanied this code.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- */
-/*
-/**
- *******************************************************************************
- * Copyright (C) 2004, International Business Machines Corporation and         *
- * others. All Rights Reserved.                                                *
- *******************************************************************************
- */
-// CHANGELOG
-//      2005-05-19 Edward Wang
-//          - copy this file from icu4jsrc_3_2/src/com/ibm/icu/lang/UCharacterEnums.java
-//          - move from package com.ibm.icu.lang to package sun.net.idn
-//
-//      2011-09-06 Kurchi Subhra Hazra
-//          - Added @Deprecated tag to the following:
-//            - class UCharacterEnums
-//            - interfaces ECharacterCategory, ECharacterDirection
-//            - fields INITIAL_QUOTE_PUNCTUATION, FINAL_QUOTE_PUNCTUATION,
-//              DIRECTIONALITY_LEFT_TO_RIGHT, DIRECTIONALITY_RIGHT_TO_LEFT,
-//              DIRECTIONALITY_EUROPEAN_NUMBER, DIRECTIONALITY_EUROPEAN_NUMBER_SEPARATOR
-//              DIRECTIONALITY_EUROPEAN_NUMBER_TERMINATOR, DIRECTIONALITY_ARABIC_NUMBER,
-//              DIRECTIONALITY_COMMON_NUMBER_SEPARATOR, DIRECTIONALITY_PARAGRAPH_SEPARATOR,
-//              DIRECTIONALITY_SEGMENT_SEPARATOR, DIRECTIONALITY_WHITESPACE,
-//              DIRECTIONALITY_OTHER_NEUTRALS, DIRECTIONALITY_LEFT_TO_RIGHT_EMBEDDING,
-//              DIRECTIONALITY_LEFT_TO_RIGHT_OVERRIDE, DIRECTIONALITY_RIGHT_TO_LEFT_ARABIC,
-//              DIRECTIONALITY_RIGHT_TO_LEFT_EMBEDDING, DIRECTIONALITY_RIGHT_TO_LEFT_OVERRIDE,
-//              DIRECTIONALITY_POP_DIRECTIONAL_FORMAT, DIRECTIONALITY_NON_SPACING_MARK,
-//              DIRECTIONALITY_BOUNDARY_NEUTRAL, DIRECTIONALITY_UNDEFINED
-//
-
-package sun.net.idn;
-
-/**
- * A container for the different 'enumerated types' used by UCharacter.
- * @draft ICU 3.0
- * @deprecated This is a draft API and might change in a future release of ICU.
- */
-
-@Deprecated
-class UCharacterEnums {
-
-    /** This is just a namespace, it is not instantiatable. */
-    private UCharacterEnums() {};
-
-    /**
-     * 'Enum' for the CharacterCategory constants.  These constants are
-     * compatible in name but not in value with those defined in
-     * java.lang.Character.
-     * @see UCharacterCategory
-     * @draft ICU 3.0
-     * @deprecated This is a draft API and might change in a future release of ICU.
-     */
-    @Deprecated
-    public static interface ECharacterCategory {
-        /**
-         * Unassigned character type
-         * @stable ICU 2.1
-         */
-        public static final int UNASSIGNED              = 0;
-
-        /**
-         * Character type Cn
-         * Not Assigned (no characters in [UnicodeData.txt] have this property)
-         * @stable ICU 2.6
-         */
-        public static final int GENERAL_OTHER_TYPES     = 0;
-
-        /**
-         * Character type Lu
-         * @stable ICU 2.1
-         */
-        public static final int UPPERCASE_LETTER        = 1;
-
-        /**
-         * Character type Ll
-         * @stable ICU 2.1
-         */
-        public static final int LOWERCASE_LETTER        = 2;
-
-        /**
-         * Character type Lt
-         * @stable ICU 2.1
-         */
-
-        public static final int TITLECASE_LETTER        = 3;
-
-        /**
-         * Character type Lm
-         * @stable ICU 2.1
-         */
-        public static final int MODIFIER_LETTER         = 4;
-
-        /**
-         * Character type Lo
-         * @stable ICU 2.1
-         */
-        public static final int OTHER_LETTER            = 5;
-
-        /**
-         * Character type Mn
-         * @stable ICU 2.1
-         */
-        public static final int NON_SPACING_MARK        = 6;
-
-        /**
-         * Character type Me
-         * @stable ICU 2.1
-         */
-        public static final int ENCLOSING_MARK          = 7;
-
-        /**
-         * Character type Mc
-         * @stable ICU 2.1
-         */
-        public static final int COMBINING_SPACING_MARK  = 8;
-
-        /**
-         * Character type Nd
-         * @stable ICU 2.1
-         */
-        public static final int DECIMAL_DIGIT_NUMBER    = 9;
-
-        /**
-         * Character type Nl
-         * @stable ICU 2.1
-         */
-        public static final int LETTER_NUMBER           = 10;
-
-        /**
-         * Character type No
-         * @stable ICU 2.1
-         */
-        public static final int OTHER_NUMBER            = 11;
-
-        /**
-         * Character type Zs
-         * @stable ICU 2.1
-         */
-        public static final int SPACE_SEPARATOR         = 12;
-
-        /**
-         * Character type Zl
-         * @stable ICU 2.1
-         */
-        public static final int LINE_SEPARATOR          = 13;
-
-        /**
-         * Character type Zp
-         * @stable ICU 2.1
-         */
-        public static final int PARAGRAPH_SEPARATOR     = 14;
-
-        /**
-         * Character type Cc
-         * @stable ICU 2.1
-         */
-        public static final int CONTROL                 = 15;
-
-        /**
-         * Character type Cf
-         * @stable ICU 2.1
-         */
-        public static final int FORMAT                  = 16;
-
-        /**
-         * Character type Co
-         * @stable ICU 2.1
-         */
-        public static final int PRIVATE_USE             = 17;
-
-        /**
-         * Character type Cs
-         * @stable ICU 2.1
-         */
-        public static final int SURROGATE               = 18;
-
-        /**
-         * Character type Pd
-         * @stable ICU 2.1
-         */
-        public static final int DASH_PUNCTUATION        = 19;
-
-        /**
-         * Character type Ps
-         * @stable ICU 2.1
-         */
-        public static final int START_PUNCTUATION       = 20;
-
-        /**
-         * Character type Pe
-         * @stable ICU 2.1
-         */
-        public static final int END_PUNCTUATION         = 21;
-
-        /**
-         * Character type Pc
-         * @stable ICU 2.1
-         */
-        public static final int CONNECTOR_PUNCTUATION   = 22;
-
-        /**
-         * Character type Po
-         * @stable ICU 2.1
-         */
-        public static final int OTHER_PUNCTUATION       = 23;
-
-        /**
-         * Character type Sm
-         * @stable ICU 2.1
-         */
-        public static final int MATH_SYMBOL             = 24;
-
-        /**
-         * Character type Sc
-         * @stable ICU 2.1
-         */
-        public static final int CURRENCY_SYMBOL         = 25;
-
-        /**
-         * Character type Sk
-         * @stable ICU 2.1
-         */
-        public static final int MODIFIER_SYMBOL         = 26;
-
-        /**
-         * Character type So
-         * @stable ICU 2.1
-         */
-        public static final int OTHER_SYMBOL            = 27;
-
-        /**
-         * Character type Pi
-         * @see #INITIAL_QUOTE_PUNCTUATION
-         * @stable ICU 2.1
-         */
-        public static final int INITIAL_PUNCTUATION     = 28;
-
-        /**
-         * Character type Pi
-         * This name is compatible with java.lang.Character's name for this type.
-         * @see #INITIAL_PUNCTUATION
-         * @draft ICU 2.8
-     * @deprecated This is a draft API and might change in a future release of ICU.
-         */
-        @Deprecated
-        public static final int INITIAL_QUOTE_PUNCTUATION = 28;
-
-        /**
-         * Character type Pf
-         * @see #FINAL_QUOTE_PUNCTUATION
-         * @stable ICU 2.1
-         */
-        public static final int FINAL_PUNCTUATION       = 29;
-
-        /**
-         * Character type Pf
-         * This name is compatible with java.lang.Character's name for this type.
-         * @see #FINAL_PUNCTUATION
-         * @draft ICU 2.8
-     * @deprecated This is a draft API and might change in a future release of ICU.
-         */
-        @Deprecated
-        public static final int FINAL_QUOTE_PUNCTUATION   = 29;
-
-        /**
-         * Character type count
-         * @stable ICU 2.1
-         */
-        public static final int CHAR_CATEGORY_COUNT     = 30;
-    }
-
-    /**
-     * 'Enum' for the CharacterDirection constants.  There are two sets
-     * of names, those used in ICU, and those used in the JDK.  The
-     * JDK constants are compatible in name but not in value
-     * with those defined in java.lang.Character.
-     * @see UCharacterDirection
-     * @draft ICU 3.0
-     * @deprecated This is a draft API and might change in a future release of ICU.
-     */
-
-    @Deprecated
-    public static interface ECharacterDirection {
-        /**
-         * Directional type L
-         * @stable ICU 2.1
-         */
-        public static final int LEFT_TO_RIGHT              = 0;
-
-        /**
-         * JDK-compatible synonum for LEFT_TO_RIGHT.
-         * @draft ICU 3.0
-     * @deprecated This is a draft API and might change in a future release of ICU.
-         */
-        @Deprecated
-        public static final byte DIRECTIONALITY_LEFT_TO_RIGHT = (byte)LEFT_TO_RIGHT;
-
-        /**
-         * Directional type R
-         * @stable ICU 2.1
-         */
-        public static final int RIGHT_TO_LEFT              = 1;
-
-        /**
-         * JDK-compatible synonum for RIGHT_TO_LEFT.
-         * @draft ICU 3.0
-     * @deprecated This is a draft API and might change in a future release of ICU.
-         */
-        @Deprecated
-        public static final byte DIRECTIONALITY_RIGHT_TO_LEFT = (byte)RIGHT_TO_LEFT;
-
-        /**
-         * Directional type EN
-         * @stable ICU 2.1
-         */
-        public static final int EUROPEAN_NUMBER            = 2;
-
-        /**
-         * JDK-compatible synonum for EUROPEAN_NUMBER.
-         * @draft ICU 3.0
-     * @deprecated This is a draft API and might change in a future release of ICU.
-         */
-        @Deprecated
-        public static final byte DIRECTIONALITY_EUROPEAN_NUMBER = (byte)EUROPEAN_NUMBER;
-
-        /**
-         * Directional type ES
-         * @stable ICU 2.1
-         */
-        public static final int EUROPEAN_NUMBER_SEPARATOR  = 3;
-
-        /**
-         * JDK-compatible synonum for EUROPEAN_NUMBER_SEPARATOR.
-         * @draft ICU 3.0
-     * @deprecated This is a draft API and might change in a future release of ICU.
-         */
-        @Deprecated
-        public static final byte DIRECTIONALITY_EUROPEAN_NUMBER_SEPARATOR = (byte)EUROPEAN_NUMBER_SEPARATOR;
-
-        /**
-         * Directional type ET
-         * @stable ICU 2.1
-         */
-        public static final int EUROPEAN_NUMBER_TERMINATOR = 4;
-
-        /**
-         * JDK-compatible synonum for EUROPEAN_NUMBER_TERMINATOR.
-         * @draft ICU 3.0
-     * @deprecated This is a draft API and might change in a future release of ICU.
-         */
-        @Deprecated
-        public static final byte DIRECTIONALITY_EUROPEAN_NUMBER_TERMINATOR = (byte)EUROPEAN_NUMBER_TERMINATOR;
-
-        /**
-         * Directional type AN
-         * @stable ICU 2.1
-         */
-        public static final int ARABIC_NUMBER              = 5;
-
-        /**
-         * JDK-compatible synonum for ARABIC_NUMBER.
-         * @draft ICU 3.0
-     * @deprecated This is a draft API and might change in a future release of ICU.
-         */
-        @Deprecated
-        public static final byte DIRECTIONALITY_ARABIC_NUMBER = (byte)ARABIC_NUMBER;
-
-        /**
-         * Directional type CS
-         * @stable ICU 2.1
-         */
-        public static final int COMMON_NUMBER_SEPARATOR    = 6;
-
-        /**
-         * JDK-compatible synonum for COMMON_NUMBER_SEPARATOR.
-         * @draft ICU 3.0
-     * @deprecated This is a draft API and might change in a future release of ICU.
-         */
-        @Deprecated
-        public static final byte DIRECTIONALITY_COMMON_NUMBER_SEPARATOR = (byte)COMMON_NUMBER_SEPARATOR;
-
-        /**
-         * Directional type B
-         * @stable ICU 2.1
-         */
-        public static final int BLOCK_SEPARATOR            = 7;
-
-        /**
-         * JDK-compatible synonum for BLOCK_SEPARATOR.
-         * @draft ICU 3.0
-     * @deprecated This is a draft API and might change in a future release of ICU.
-         */
-        @Deprecated
-        public static final byte DIRECTIONALITY_PARAGRAPH_SEPARATOR = (byte)BLOCK_SEPARATOR;
-
-        /**
-         * Directional type S
-         * @stable ICU 2.1
-         */
-        public static final int SEGMENT_SEPARATOR          = 8;
-
-        /**
-         * JDK-compatible synonum for SEGMENT_SEPARATOR.
-         * @draft ICU 3.0
-     * @deprecated This is a draft API and might change in a future release of ICU.
-         */
-        @Deprecated
-        public static final byte DIRECTIONALITY_SEGMENT_SEPARATOR = (byte)SEGMENT_SEPARATOR;
-
-        /**
-         * Directional type WS
-         * @stable ICU 2.1
-         */
-        public static final int WHITE_SPACE_NEUTRAL        = 9;
-
-        /**
-         * JDK-compatible synonum for WHITE_SPACE_NEUTRAL.
-         * @draft ICU 3.0
-     * @deprecated This is a draft API and might change in a future release of ICU.
-         */
-        @Deprecated
-        public static final byte DIRECTIONALITY_WHITESPACE = (byte)WHITE_SPACE_NEUTRAL;
-
-        /**
-         * Directional type ON
-         * @stable ICU 2.1
-         */
-        public static final int OTHER_NEUTRAL              = 10;
-
-        /**
-         * JDK-compatible synonum for OTHER_NEUTRAL.
-         * @draft ICU 3.0
-     * @deprecated This is a draft API and might change in a future release of ICU.
-         */
-        @Deprecated
-        public static final byte DIRECTIONALITY_OTHER_NEUTRALS = (byte)OTHER_NEUTRAL;
-
-        /**
-         * Directional type LRE
-         * @stable ICU 2.1
-         */
-        public static final int LEFT_TO_RIGHT_EMBEDDING    = 11;
-
-        /**
-         * JDK-compatible synonum for LEFT_TO_RIGHT_EMBEDDING.
-         * @draft ICU 3.0
-     * @deprecated This is a draft API and might change in a future release of ICU.
-         */
-        @Deprecated
-        public static final byte DIRECTIONALITY_LEFT_TO_RIGHT_EMBEDDING = (byte)LEFT_TO_RIGHT_EMBEDDING;
-
-        /**
-         * Directional type LRO
-         * @stable ICU 2.1
-         */
-        public static final int LEFT_TO_RIGHT_OVERRIDE     = 12;
-
-        /**
-         * JDK-compatible synonum for LEFT_TO_RIGHT_OVERRIDE.
-         * @draft ICU 3.0
-     * @deprecated This is a draft API and might change in a future release of ICU.
-         */
-        @Deprecated
-        public static final byte DIRECTIONALITY_LEFT_TO_RIGHT_OVERRIDE = (byte)LEFT_TO_RIGHT_OVERRIDE;
-
-        /**
-         * Directional type AL
-         * @stable ICU 2.1
-         */
-        public static final int RIGHT_TO_LEFT_ARABIC       = 13;
-
-        /**
-         * JDK-compatible synonum for RIGHT_TO_LEFT_ARABIC.
-         * @draft ICU 3.0
-     * @deprecated This is a draft API and might change in a future release of ICU.
-         */
-        @Deprecated
-        public static final byte DIRECTIONALITY_RIGHT_TO_LEFT_ARABIC = (byte)RIGHT_TO_LEFT_ARABIC;
-
-        /**
-         * Directional type RLE
-         * @stable ICU 2.1
-         */
-        public static final int RIGHT_TO_LEFT_EMBEDDING    = 14;
-
-        /**
-         * JDK-compatible synonum for RIGHT_TO_LEFT_EMBEDDING.
-         * @draft ICU 3.0
-     * @deprecated This is a draft API and might change in a future release of ICU.
-         */
-        @Deprecated
-        public static final byte DIRECTIONALITY_RIGHT_TO_LEFT_EMBEDDING = (byte)RIGHT_TO_LEFT_EMBEDDING;
-
-        /**
-         * Directional type RLO
-         * @stable ICU 2.1
-         */
-        public static final int RIGHT_TO_LEFT_OVERRIDE     = 15;
-
-        /**
-         * JDK-compatible synonum for RIGHT_TO_LEFT_OVERRIDE.
-         * @draft ICU 3.0
-     * @deprecated This is a draft API and might change in a future release of ICU.
-         */
-        @Deprecated
-        public static final byte DIRECTIONALITY_RIGHT_TO_LEFT_OVERRIDE = (byte)RIGHT_TO_LEFT_OVERRIDE;
-
-        /**
-         * Directional type PDF
-         * @stable ICU 2.1
-         */
-        public static final int POP_DIRECTIONAL_FORMAT     = 16;
-
-        /**
-         * JDK-compatible synonum for POP_DIRECTIONAL_FORMAT.
-         * @draft ICU 3.0
-     * @deprecated This is a draft API and might change in a future release of ICU.
-         */
-        @Deprecated
-        public static final byte DIRECTIONALITY_POP_DIRECTIONAL_FORMAT = (byte)POP_DIRECTIONAL_FORMAT;
-
-        /**
-         * Directional type NSM
-         * @stable ICU 2.1
-         */
-        public static final int DIR_NON_SPACING_MARK       = 17;
-
-        /**
-         * JDK-compatible synonum for DIR_NON_SPACING_MARK.
-         * @draft ICU 3.0
-     * @deprecated This is a draft API and might change in a future release of ICU.
-         */
-        @Deprecated
-        public static final byte DIRECTIONALITY_NON_SPACING_MARK = (byte)DIR_NON_SPACING_MARK;
-
-        /**
-         * Directional type BN
-         * @stable ICU 2.1
-         */
-        public static final int BOUNDARY_NEUTRAL           = 18;
-
-        /**
-         * JDK-compatible synonum for BOUNDARY_NEUTRAL.
-         * @draft ICU 3.0
-     * @deprecated This is a draft API and might change in a future release of ICU.
-         */
-        @Deprecated
-        public static final byte DIRECTIONALITY_BOUNDARY_NEUTRAL = (byte)BOUNDARY_NEUTRAL;
-
-        /**
-         * Number of directional types
-         * @stable ICU 2.1
-         */
-        public static final int CHAR_DIRECTION_COUNT       = 19;
-
-        /**
-         * Undefined bidirectional character type. Undefined char
-         * values have undefined directionality in the Unicode specification.
-     * @draft ICU 3.0
-     * @deprecated This is a draft API and might change in a future release of ICU.
-         */
-        @Deprecated
-        public static final byte DIRECTIONALITY_UNDEFINED = -1;
-    }
-}
--- /dev/null	2020-01-10 15:57:51.000000000 -0800
+++ new/src/java.base/share/classes/jdk/internal/icu/lang/UCharacterEnums.java	2020-01-10 15:57:51.000000000 -0800
@@ -0,0 +1,588 @@
+/*
+ * Copyright (c) 2005, 2020, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.  Oracle designates this
+ * particular file as subject to the "Classpath" exception as provided
+ * by Oracle in the LICENSE file that accompanied this code.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+/*
+/**
+ *******************************************************************************
+ * Copyright (C) 2004, International Business Machines Corporation and         *
+ * others. All Rights Reserved.                                                *
+ *******************************************************************************
+ */
+// CHANGELOG
+//      2005-05-19 Edward Wang
+//          - copy this file from icu4jsrc_3_2/src/com/ibm/icu/lang/UCharacterEnums.java
+//          - move from package com.ibm.icu.lang to package sun.net.idn
+//
+//      2011-09-06 Kurchi Subhra Hazra
+//          - Added @Deprecated tag to the following:
+//            - class UCharacterEnums
+//            - interfaces ECharacterCategory, ECharacterDirection
+//            - fields INITIAL_QUOTE_PUNCTUATION, FINAL_QUOTE_PUNCTUATION,
+//              DIRECTIONALITY_LEFT_TO_RIGHT, DIRECTIONALITY_RIGHT_TO_LEFT,
+//              DIRECTIONALITY_EUROPEAN_NUMBER, DIRECTIONALITY_EUROPEAN_NUMBER_SEPARATOR
+//              DIRECTIONALITY_EUROPEAN_NUMBER_TERMINATOR, DIRECTIONALITY_ARABIC_NUMBER,
+//              DIRECTIONALITY_COMMON_NUMBER_SEPARATOR, DIRECTIONALITY_PARAGRAPH_SEPARATOR,
+//              DIRECTIONALITY_SEGMENT_SEPARATOR, DIRECTIONALITY_WHITESPACE,
+//              DIRECTIONALITY_OTHER_NEUTRALS, DIRECTIONALITY_LEFT_TO_RIGHT_EMBEDDING,
+//              DIRECTIONALITY_LEFT_TO_RIGHT_OVERRIDE, DIRECTIONALITY_RIGHT_TO_LEFT_ARABIC,
+//              DIRECTIONALITY_RIGHT_TO_LEFT_EMBEDDING, DIRECTIONALITY_RIGHT_TO_LEFT_OVERRIDE,
+//              DIRECTIONALITY_POP_DIRECTIONAL_FORMAT, DIRECTIONALITY_NON_SPACING_MARK,
+//              DIRECTIONALITY_BOUNDARY_NEUTRAL, DIRECTIONALITY_UNDEFINED
+//
+
+package jdk.internal.icu.lang;
+
+/**
+ * A container for the different 'enumerated types' used by UCharacter.
+ * @draft ICU 3.0
+ * @deprecated This is a draft API and might change in a future release of ICU.
+ */
+
+@Deprecated
+class UCharacterEnums {
+
+    /** This is just a namespace, it is not instantiatable. */
+    private UCharacterEnums() {};
+
+    /**
+     * 'Enum' for the CharacterCategory constants.  These constants are
+     * compatible in name but not in value with those defined in
+     * java.lang.Character.
+     * @see UCharacterCategory
+     * @draft ICU 3.0
+     * @deprecated This is a draft API and might change in a future release of ICU.
+     */
+    @Deprecated
+    public static interface ECharacterCategory {
+        /**
+         * Unassigned character type
+         * @stable ICU 2.1
+         */
+        public static final int UNASSIGNED              = 0;
+
+        /**
+         * Character type Cn
+         * Not Assigned (no characters in [UnicodeData.txt] have this property)
+         * @stable ICU 2.6
+         */
+        public static final int GENERAL_OTHER_TYPES     = 0;
+
+        /**
+         * Character type Lu
+         * @stable ICU 2.1
+         */
+        public static final int UPPERCASE_LETTER        = 1;
+
+        /**
+         * Character type Ll
+         * @stable ICU 2.1
+         */
+        public static final int LOWERCASE_LETTER        = 2;
+
+        /**
+         * Character type Lt
+         * @stable ICU 2.1
+         */
+
+        public static final int TITLECASE_LETTER        = 3;
+
+        /**
+         * Character type Lm
+         * @stable ICU 2.1
+         */
+        public static final int MODIFIER_LETTER         = 4;
+
+        /**
+         * Character type Lo
+         * @stable ICU 2.1
+         */
+        public static final int OTHER_LETTER            = 5;
+
+        /**
+         * Character type Mn
+         * @stable ICU 2.1
+         */
+        public static final int NON_SPACING_MARK        = 6;
+
+        /**
+         * Character type Me
+         * @stable ICU 2.1
+         */
+        public static final int ENCLOSING_MARK          = 7;
+
+        /**
+         * Character type Mc
+         * @stable ICU 2.1
+         */
+        public static final int COMBINING_SPACING_MARK  = 8;
+
+        /**
+         * Character type Nd
+         * @stable ICU 2.1
+         */
+        public static final int DECIMAL_DIGIT_NUMBER    = 9;
+
+        /**
+         * Character type Nl
+         * @stable ICU 2.1
+         */
+        public static final int LETTER_NUMBER           = 10;
+
+        /**
+         * Character type No
+         * @stable ICU 2.1
+         */
+        public static final int OTHER_NUMBER            = 11;
+
+        /**
+         * Character type Zs
+         * @stable ICU 2.1
+         */
+        public static final int SPACE_SEPARATOR         = 12;
+
+        /**
+         * Character type Zl
+         * @stable ICU 2.1
+         */
+        public static final int LINE_SEPARATOR          = 13;
+
+        /**
+         * Character type Zp
+         * @stable ICU 2.1
+         */
+        public static final int PARAGRAPH_SEPARATOR     = 14;
+
+        /**
+         * Character type Cc
+         * @stable ICU 2.1
+         */
+        public static final int CONTROL                 = 15;
+
+        /**
+         * Character type Cf
+         * @stable ICU 2.1
+         */
+        public static final int FORMAT                  = 16;
+
+        /**
+         * Character type Co
+         * @stable ICU 2.1
+         */
+        public static final int PRIVATE_USE             = 17;
+
+        /**
+         * Character type Cs
+         * @stable ICU 2.1
+         */
+        public static final int SURROGATE               = 18;
+
+        /**
+         * Character type Pd
+         * @stable ICU 2.1
+         */
+        public static final int DASH_PUNCTUATION        = 19;
+
+        /**
+         * Character type Ps
+         * @stable ICU 2.1
+         */
+        public static final int START_PUNCTUATION       = 20;
+
+        /**
+         * Character type Pe
+         * @stable ICU 2.1
+         */
+        public static final int END_PUNCTUATION         = 21;
+
+        /**
+         * Character type Pc
+         * @stable ICU 2.1
+         */
+        public static final int CONNECTOR_PUNCTUATION   = 22;
+
+        /**
+         * Character type Po
+         * @stable ICU 2.1
+         */
+        public static final int OTHER_PUNCTUATION       = 23;
+
+        /**
+         * Character type Sm
+         * @stable ICU 2.1
+         */
+        public static final int MATH_SYMBOL             = 24;
+
+        /**
+         * Character type Sc
+         * @stable ICU 2.1
+         */
+        public static final int CURRENCY_SYMBOL         = 25;
+
+        /**
+         * Character type Sk
+         * @stable ICU 2.1
+         */
+        public static final int MODIFIER_SYMBOL         = 26;
+
+        /**
+         * Character type So
+         * @stable ICU 2.1
+         */
+        public static final int OTHER_SYMBOL            = 27;
+
+        /**
+         * Character type Pi
+         * @see #INITIAL_QUOTE_PUNCTUATION
+         * @stable ICU 2.1
+         */
+        public static final int INITIAL_PUNCTUATION     = 28;
+
+        /**
+         * Character type Pi
+         * This name is compatible with java.lang.Character's name for this type.
+         * @see #INITIAL_PUNCTUATION
+         * @draft ICU 2.8
+     * @deprecated This is a draft API and might change in a future release of ICU.
+         */
+        @Deprecated
+        public static final int INITIAL_QUOTE_PUNCTUATION = 28;
+
+        /**
+         * Character type Pf
+         * @see #FINAL_QUOTE_PUNCTUATION
+         * @stable ICU 2.1
+         */
+        public static final int FINAL_PUNCTUATION       = 29;
+
+        /**
+         * Character type Pf
+         * This name is compatible with java.lang.Character's name for this type.
+         * @see #FINAL_PUNCTUATION
+         * @draft ICU 2.8
+     * @deprecated This is a draft API and might change in a future release of ICU.
+         */
+        @Deprecated
+        public static final int FINAL_QUOTE_PUNCTUATION   = 29;
+
+        /**
+         * Character type count
+         * @stable ICU 2.1
+         */
+        public static final int CHAR_CATEGORY_COUNT     = 30;
+    }
+
+    /**
+     * 'Enum' for the CharacterDirection constants.  There are two sets
+     * of names, those used in ICU, and those used in the JDK.  The
+     * JDK constants are compatible in name but not in value
+     * with those defined in java.lang.Character.
+     * @see UCharacterDirection
+     * @draft ICU 3.0
+     * @deprecated This is a draft API and might change in a future release of ICU.
+     */
+
+    @Deprecated
+    public static interface ECharacterDirection {
+        /**
+         * Directional type L
+         * @stable ICU 2.1
+         */
+        public static final int LEFT_TO_RIGHT              = 0;
+
+        /**
+         * JDK-compatible synonum for LEFT_TO_RIGHT.
+         * @draft ICU 3.0
+     * @deprecated This is a draft API and might change in a future release of ICU.
+         */
+        @Deprecated
+        public static final byte DIRECTIONALITY_LEFT_TO_RIGHT = (byte)LEFT_TO_RIGHT;
+
+        /**
+         * Directional type R
+         * @stable ICU 2.1
+         */
+        public static final int RIGHT_TO_LEFT              = 1;
+
+        /**
+         * JDK-compatible synonum for RIGHT_TO_LEFT.
+         * @draft ICU 3.0
+     * @deprecated This is a draft API and might change in a future release of ICU.
+         */
+        @Deprecated
+        public static final byte DIRECTIONALITY_RIGHT_TO_LEFT = (byte)RIGHT_TO_LEFT;
+
+        /**
+         * Directional type EN
+         * @stable ICU 2.1
+         */
+        public static final int EUROPEAN_NUMBER            = 2;
+
+        /**
+         * JDK-compatible synonum for EUROPEAN_NUMBER.
+         * @draft ICU 3.0
+     * @deprecated This is a draft API and might change in a future release of ICU.
+         */
+        @Deprecated
+        public static final byte DIRECTIONALITY_EUROPEAN_NUMBER = (byte)EUROPEAN_NUMBER;
+
+        /**
+         * Directional type ES
+         * @stable ICU 2.1
+         */
+        public static final int EUROPEAN_NUMBER_SEPARATOR  = 3;
+
+        /**
+         * JDK-compatible synonum for EUROPEAN_NUMBER_SEPARATOR.
+         * @draft ICU 3.0
+     * @deprecated This is a draft API and might change in a future release of ICU.
+         */
+        @Deprecated
+        public static final byte DIRECTIONALITY_EUROPEAN_NUMBER_SEPARATOR = (byte)EUROPEAN_NUMBER_SEPARATOR;
+
+        /**
+         * Directional type ET
+         * @stable ICU 2.1
+         */
+        public static final int EUROPEAN_NUMBER_TERMINATOR = 4;
+
+        /**
+         * JDK-compatible synonum for EUROPEAN_NUMBER_TERMINATOR.
+         * @draft ICU 3.0
+     * @deprecated This is a draft API and might change in a future release of ICU.
+         */
+        @Deprecated
+        public static final byte DIRECTIONALITY_EUROPEAN_NUMBER_TERMINATOR = (byte)EUROPEAN_NUMBER_TERMINATOR;
+
+        /**
+         * Directional type AN
+         * @stable ICU 2.1
+         */
+        public static final int ARABIC_NUMBER              = 5;
+
+        /**
+         * JDK-compatible synonum for ARABIC_NUMBER.
+         * @draft ICU 3.0
+     * @deprecated This is a draft API and might change in a future release of ICU.
+         */
+        @Deprecated
+        public static final byte DIRECTIONALITY_ARABIC_NUMBER = (byte)ARABIC_NUMBER;
+
+        /**
+         * Directional type CS
+         * @stable ICU 2.1
+         */
+        public static final int COMMON_NUMBER_SEPARATOR    = 6;
+
+        /**
+         * JDK-compatible synonum for COMMON_NUMBER_SEPARATOR.
+         * @draft ICU 3.0
+     * @deprecated This is a draft API and might change in a future release of ICU.
+         */
+        @Deprecated
+        public static final byte DIRECTIONALITY_COMMON_NUMBER_SEPARATOR = (byte)COMMON_NUMBER_SEPARATOR;
+
+        /**
+         * Directional type B
+         * @stable ICU 2.1
+         */
+        public static final int BLOCK_SEPARATOR            = 7;
+
+        /**
+         * JDK-compatible synonum for BLOCK_SEPARATOR.
+         * @draft ICU 3.0
+     * @deprecated This is a draft API and might change in a future release of ICU.
+         */
+        @Deprecated
+        public static final byte DIRECTIONALITY_PARAGRAPH_SEPARATOR = (byte)BLOCK_SEPARATOR;
+
+        /**
+         * Directional type S
+         * @stable ICU 2.1
+         */
+        public static final int SEGMENT_SEPARATOR          = 8;
+
+        /**
+         * JDK-compatible synonum for SEGMENT_SEPARATOR.
+         * @draft ICU 3.0
+     * @deprecated This is a draft API and might change in a future release of ICU.
+         */
+        @Deprecated
+        public static final byte DIRECTIONALITY_SEGMENT_SEPARATOR = (byte)SEGMENT_SEPARATOR;
+
+        /**
+         * Directional type WS
+         * @stable ICU 2.1
+         */
+        public static final int WHITE_SPACE_NEUTRAL        = 9;
+
+        /**
+         * JDK-compatible synonum for WHITE_SPACE_NEUTRAL.
+         * @draft ICU 3.0
+     * @deprecated This is a draft API and might change in a future release of ICU.
+         */
+        @Deprecated
+        public static final byte DIRECTIONALITY_WHITESPACE = (byte)WHITE_SPACE_NEUTRAL;
+
+        /**
+         * Directional type ON
+         * @stable ICU 2.1
+         */
+        public static final int OTHER_NEUTRAL              = 10;
+
+        /**
+         * JDK-compatible synonum for OTHER_NEUTRAL.
+         * @draft ICU 3.0
+     * @deprecated This is a draft API and might change in a future release of ICU.
+         */
+        @Deprecated
+        public static final byte DIRECTIONALITY_OTHER_NEUTRALS = (byte)OTHER_NEUTRAL;
+
+        /**
+         * Directional type LRE
+         * @stable ICU 2.1
+         */
+        public static final int LEFT_TO_RIGHT_EMBEDDING    = 11;
+
+        /**
+         * JDK-compatible synonum for LEFT_TO_RIGHT_EMBEDDING.
+         * @draft ICU 3.0
+     * @deprecated This is a draft API and might change in a future release of ICU.
+         */
+        @Deprecated
+        public static final byte DIRECTIONALITY_LEFT_TO_RIGHT_EMBEDDING = (byte)LEFT_TO_RIGHT_EMBEDDING;
+
+        /**
+         * Directional type LRO
+         * @stable ICU 2.1
+         */
+        public static final int LEFT_TO_RIGHT_OVERRIDE     = 12;
+
+        /**
+         * JDK-compatible synonum for LEFT_TO_RIGHT_OVERRIDE.
+         * @draft ICU 3.0
+     * @deprecated This is a draft API and might change in a future release of ICU.
+         */
+        @Deprecated
+        public static final byte DIRECTIONALITY_LEFT_TO_RIGHT_OVERRIDE = (byte)LEFT_TO_RIGHT_OVERRIDE;
+
+        /**
+         * Directional type AL
+         * @stable ICU 2.1
+         */
+        public static final int RIGHT_TO_LEFT_ARABIC       = 13;
+
+        /**
+         * JDK-compatible synonum for RIGHT_TO_LEFT_ARABIC.
+         * @draft ICU 3.0
+     * @deprecated This is a draft API and might change in a future release of ICU.
+         */
+        @Deprecated
+        public static final byte DIRECTIONALITY_RIGHT_TO_LEFT_ARABIC = (byte)RIGHT_TO_LEFT_ARABIC;
+
+        /**
+         * Directional type RLE
+         * @stable ICU 2.1
+         */
+        public static final int RIGHT_TO_LEFT_EMBEDDING    = 14;
+
+        /**
+         * JDK-compatible synonum for RIGHT_TO_LEFT_EMBEDDING.
+         * @draft ICU 3.0
+     * @deprecated This is a draft API and might change in a future release of ICU.
+         */
+        @Deprecated
+        public static final byte DIRECTIONALITY_RIGHT_TO_LEFT_EMBEDDING = (byte)RIGHT_TO_LEFT_EMBEDDING;
+
+        /**
+         * Directional type RLO
+         * @stable ICU 2.1
+         */
+        public static final int RIGHT_TO_LEFT_OVERRIDE     = 15;
+
+        /**
+         * JDK-compatible synonum for RIGHT_TO_LEFT_OVERRIDE.
+         * @draft ICU 3.0
+     * @deprecated This is a draft API and might change in a future release of ICU.
+         */
+        @Deprecated
+        public static final byte DIRECTIONALITY_RIGHT_TO_LEFT_OVERRIDE = (byte)RIGHT_TO_LEFT_OVERRIDE;
+
+        /**
+         * Directional type PDF
+         * @stable ICU 2.1
+         */
+        public static final int POP_DIRECTIONAL_FORMAT     = 16;
+
+        /**
+         * JDK-compatible synonum for POP_DIRECTIONAL_FORMAT.
+         * @draft ICU 3.0
+     * @deprecated This is a draft API and might change in a future release of ICU.
+         */
+        @Deprecated
+        public static final byte DIRECTIONALITY_POP_DIRECTIONAL_FORMAT = (byte)POP_DIRECTIONAL_FORMAT;
+
+        /**
+         * Directional type NSM
+         * @stable ICU 2.1
+         */
+        public static final int DIR_NON_SPACING_MARK       = 17;
+
+        /**
+         * JDK-compatible synonum for DIR_NON_SPACING_MARK.
+         * @draft ICU 3.0
+     * @deprecated This is a draft API and might change in a future release of ICU.
+         */
+        @Deprecated
+        public static final byte DIRECTIONALITY_NON_SPACING_MARK = (byte)DIR_NON_SPACING_MARK;
+
+        /**
+         * Directional type BN
+         * @stable ICU 2.1
+         */
+        public static final int BOUNDARY_NEUTRAL           = 18;
+
+        /**
+         * JDK-compatible synonum for BOUNDARY_NEUTRAL.
+         * @draft ICU 3.0
+     * @deprecated This is a draft API and might change in a future release of ICU.
+         */
+        @Deprecated
+        public static final byte DIRECTIONALITY_BOUNDARY_NEUTRAL = (byte)BOUNDARY_NEUTRAL;
+
+        /**
+         * Number of directional types
+         * @stable ICU 2.1
+         */
+        public static final int CHAR_DIRECTION_COUNT       = 19;
+
+        /**
+         * Undefined bidirectional character type. Undefined char
+         * values have undefined directionality in the Unicode specification.
+     * @draft ICU 3.0
+     * @deprecated This is a draft API and might change in a future release of ICU.
+         */
+        @Deprecated
+        public static final byte DIRECTIONALITY_UNDEFINED = -1;
+    }
+}
--- old/src/java.base/share/classes/sun/text/bidi/BidiBase.java	2020-01-10 15:57:52.000000000 -0800
+++ /dev/null	2020-01-10 15:57:52.000000000 -0800
@@ -1,4780 +0,0 @@
-/*
- * Copyright (c) 2009, 2018, Oracle and/or its affiliates. All rights reserved.
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.  Oracle designates this
- * particular file as subject to the "Classpath" exception as provided
- * by Oracle in the LICENSE file that accompanied this code.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- */
-
-/*
-*******************************************************************************
-*   Copyright (C) 2001-2014, International Business Machines
-*   Corporation and others.  All Rights Reserved.
-*******************************************************************************
-*/
-
-/* FOOD FOR THOUGHT: currently the reordering modes are a mixture of
- * algorithm for direct BiDi, algorithm for inverse Bidi and the bizarre
- * concept of RUNS_ONLY which is a double operation.
- * It could be advantageous to divide this into 3 concepts:
- * a) Operation: direct / inverse / RUNS_ONLY
- * b) Direct algorithm: default / NUMBERS_SPECIAL / GROUP_NUMBERS_WITH_L
- * c) Inverse algorithm: default / INVERSE_LIKE_DIRECT / NUMBERS_SPECIAL
- * This would allow combinations not possible today like RUNS_ONLY with
- * NUMBERS_SPECIAL.
- * Also allow to set INSERT_MARKS for the direct step of RUNS_ONLY and
- * REMOVE_CONTROLS for the inverse step.
- * Not all combinations would be supported, and probably not all do make sense.
- * This would need to document which ones are supported and what are the
- * fallbacks for unsupported combinations.
- */
-
-package sun.text.bidi;
-
-import java.lang.reflect.Array;
-import java.text.AttributedCharacterIterator;
-import java.text.Bidi;
-import java.util.Arrays;
-import jdk.internal.access.JavaAWTFontAccess;
-import jdk.internal.access.SharedSecrets;
-import sun.text.normalizer.UBiDiProps;
-import sun.text.normalizer.UCharacter;
-import sun.text.normalizer.UTF16;
-
-/**
- *
- * 
Bidi algorithm for ICU

- *
- * This is an implementation of the Unicode Bidirectional Algorithm. The
- * algorithm is defined in the Unicode Standard Annex #9.
- * 

- *
- * Note: Libraries that perform a bidirectional algorithm and reorder strings
- * accordingly are sometimes called "Storage Layout Engines". ICU's Bidi and
- * shaping (ArabicShaping) classes can be used at the core of such "Storage
- * Layout Engines".
- *
- * 
General remarks about the API:

- *
- * The "limit" of a sequence of characters is the position just after
- * their last character, i.e., one more than that position.
- * 

- *
- * Some of the API methods provide access to "runs". Such a
- * "run" is defined as a sequence of characters that are at the same
- * embedding level after performing the Bidi algorithm.
- *
- * 
Basic concept: paragraph

- * A piece of text can be divided into several paragraphs by characters
- * with the Bidi class Block Separator. For handling of
- * paragraphs, see:
- * 
    
- * 
  • {@link #countParagraphs}
- * 
  • {@link #getParaLevel}
- * 
  • {@link #getParagraph}
- * 
  • {@link #getParagraphByIndex}
- * 

- *
- * 
Basic concept: text direction

- * The direction of a piece of text may be:
- * 
    
- * 
  • {@link #LTR}
- * 
  • {@link #RTL}
- * 
  • {@link #MIXED}
- * 
  • {@link #NEUTRAL}
- * 

- *
- * 
Basic concept: levels

- *
- * Levels in this API represent embedding levels according to the Unicode
- * Bidirectional Algorithm.
- * Their low-order bit (even/odd value) indicates the visual direction.

- *
- * Levels can be abstract values when used for the
- * paraLevel and embeddingLevels
- * arguments of setPara(); there:
- * 
    
- * 
  • the high-order bit of an embeddingLevels[]
- * value indicates whether the using application is
- * specifying the level of a character to override whatever the
- * Bidi implementation would resolve it to.
    • 
- * 
  • paraLevel can be set to the
- * pseudo-level values LEVEL_DEFAULT_LTR
- * and LEVEL_DEFAULT_RTL.
    • 
- * 

- *
- * 
The related constants are not real, valid level values.
- * DEFAULT_XXX can be used to specify
- * a default for the paragraph level for
- * when the setPara() method
- * shall determine it but there is no
- * strongly typed character in the input.

- *
- * Note that the value for LEVEL_DEFAULT_LTR is even
- * and the one for LEVEL_DEFAULT_RTL is odd,
- * just like with normal LTR and RTL level values -
- * these special values are designed that way. Also, the implementation
- * assumes that MAX_EXPLICIT_LEVEL is odd.
- *
- * 
See Also:
- * 
    
- * 
  • {@link #LEVEL_DEFAULT_LTR}
- * 
  • {@link #LEVEL_DEFAULT_RTL}
- * 
  • {@link #LEVEL_OVERRIDE}
- * 
  • {@link #MAX_EXPLICIT_LEVEL}
- * 
  • {@link #setPara}
- * 

- *
- * 
Basic concept: Reordering Mode

- * Reordering mode values indicate which variant of the Bidi algorithm to
- * use.
- *
- * 
See Also:
- * 
    
- * 
  • {@link #setReorderingMode}
- * 
  • {@link #REORDER_DEFAULT}
- * 
  • {@link #REORDER_NUMBERS_SPECIAL}
- * 
  • {@link #REORDER_GROUP_NUMBERS_WITH_R}
- * 
  • {@link #REORDER_RUNS_ONLY}
- * 
  • {@link #REORDER_INVERSE_NUMBERS_AS_L}
- * 
  • {@link #REORDER_INVERSE_LIKE_DIRECT}
- * 
  • {@link #REORDER_INVERSE_FOR_NUMBERS_SPECIAL}
- * 

- *
- * 
Basic concept: Reordering Options

- * Reordering options can be applied during Bidi text transformations.
- *
- * 
See Also:
- * 
    
- * 
  • {@link #setReorderingOptions}
- * 
  • {@link #OPTION_DEFAULT}
- * 
  • {@link #OPTION_INSERT_MARKS}
- * 
  • {@link #OPTION_REMOVE_CONTROLS}
- * 
  • {@link #OPTION_STREAMING}
- * 

- *
- *
- * @author Simon Montagu, Matitiahu Allouche (ported from C code written by Markus W. Scherer)
- * @stable ICU 3.8
- *
- *
- * 
 Sample code for the ICU Bidi API 

- *
- * 
Rendering a paragraph with the ICU Bidi API

- *
- * This is (hypothetical) sample code that illustrates how the ICU Bidi API
- * could be used to render a paragraph of text. Rendering code depends highly on
- * the graphics system, therefore this sample code must make a lot of
- * assumptions, which may or may not match any existing graphics system's
- * properties.
- *
- * 

- * The basic assumptions are:
- * 

- * 
    
- * 
  • Rendering is done from left to right on a horizontal line.
    • 
- * 
  • A run of single-style, unidirectional text can be rendered at once.
- * 
    • 
- * 
  • Such a run of text is passed to the graphics system with characters
- * (code units) in logical order.
    • 
- * 
  • The line-breaking algorithm is very complicated and Locale-dependent -
- * and therefore its implementation omitted from this sample code.
    • 
- * 

- *
- * 
{@code
- *
- *  package com.ibm.icu.dev.test.bidi;
- *
- *  import com.ibm.icu.text.Bidi;
- *  import com.ibm.icu.text.BidiRun;
- *
- *  public class Sample {
- *
- *      static final int styleNormal = 0;
- *      static final int styleSelected = 1;
- *      static final int styleBold = 2;
- *      static final int styleItalics = 4;
- *      static final int styleSuper=8;
- *      static final int styleSub = 16;
- *
- *      static class StyleRun {
- *          int limit;
- *          int style;
- *
- *          public StyleRun(int limit, int style) {
- *              this.limit = limit;
- *              this.style = style;
- *          }
- *      }
- *
- *      static class Bounds {
- *          int start;
- *          int limit;
- *
- *          public Bounds(int start, int limit) {
- *              this.start = start;
- *              this.limit = limit;
- *          }
- *      }
- *
- *      static int getTextWidth(String text, int start, int limit,
- *                              StyleRun[] styleRuns, int styleRunCount) {
- *          // simplistic way to compute the width
- *          return limit - start;
- *      }
- *
- *      // set limit and StyleRun limit for a line
- *      // from text[start] and from styleRuns[styleRunStart]
- *      // using Bidi.getLogicalRun(...)
- *      // returns line width
- *      static int getLineBreak(String text, Bounds line, Bidi para,
- *                              StyleRun styleRuns[], Bounds styleRun) {
- *          // dummy return
- *          return 0;
- *      }
- *
- *      // render runs on a line sequentially, always from left to right
- *
- *      // prepare rendering a new line
- *      static void startLine(byte textDirection, int lineWidth) {
- *          System.out.println();
- *      }
- *
- *      // render a run of text and advance to the right by the run width
- *      // the text[start..limit-1] is always in logical order
- *      static void renderRun(String text, int start, int limit,
- *                            byte textDirection, int style) {
- *      }
- *
- *      // We could compute a cross-product
- *      // from the style runs with the directional runs
- *      // and then reorder it.
- *      // Instead, here we iterate over each run type
- *      // and render the intersections -
- *      // with shortcuts in simple (and common) cases.
- *      // renderParagraph() is the main function.
- *
- *      // render a directional run with
- *      // (possibly) multiple style runs intersecting with it
- *      static void renderDirectionalRun(String text, int start, int limit,
- *                                       byte direction, StyleRun styleRuns[],
- *                                       int styleRunCount) {
- *          int i;
- *
- *          // iterate over style runs
- *          if (direction == Bidi.LTR) {
- *              int styleLimit;
- *              for (i = 0; i < styleRunCount; ++i) {
- *                  styleLimit = styleRuns[i].limit;
- *                  if (start < styleLimit) {
- *                      if (styleLimit > limit) {
- *                          styleLimit = limit;
- *                      }
- *                      renderRun(text, start, styleLimit,
- *                                direction, styleRuns[i].style);
- *                      if (styleLimit == limit) {
- *                          break;
- *                      }
- *                      start = styleLimit;
- *                  }
- *              }
- *          } else {
- *              int styleStart;
- *
- *              for (i = styleRunCount-1; i >= 0; --i) {
- *                  if (i > 0) {
- *                      styleStart = styleRuns[i-1].limit;
- *                  } else {
- *                      styleStart = 0;
- *                  }
- *                  if (limit >= styleStart) {
- *                      if (styleStart < start) {
- *                          styleStart = start;
- *                      }
- *                      renderRun(text, styleStart, limit, direction,
- *                                styleRuns[i].style);
- *                      if (styleStart == start) {
- *                          break;
- *                      }
- *                      limit = styleStart;
- *                  }
- *              }
- *          }
- *      }
- *
- *      // the line object represents text[start..limit-1]
- *      static void renderLine(Bidi line, String text, int start, int limit,
- *                             StyleRun styleRuns[], int styleRunCount) {
- *          byte direction = line.getDirection();
- *          if (direction != Bidi.MIXED) {
- *              // unidirectional
- *              if (styleRunCount <= 1) {
- *                  renderRun(text, start, limit, direction, styleRuns[0].style);
- *              } else {
- *                  renderDirectionalRun(text, start, limit, direction,
- *                                       styleRuns, styleRunCount);
- *              }
- *          } else {
- *              // mixed-directional
- *              int count, i;
- *              BidiRun run;
- *
- *              try {
- *                  count = line.countRuns();
- *              } catch (IllegalStateException e) {
- *                  e.printStackTrace();
- *                  return;
- *              }
- *              if (styleRunCount <= 1) {
- *                  int style = styleRuns[0].style;
- *
- *                  // iterate over directional runs
- *                  for (i = 0; i < count; ++i) {
- *                      run = line.getVisualRun(i);
- *                      renderRun(text, run.getStart(), run.getLimit(),
- *                                run.getDirection(), style);
- *                  }
- *              } else {
- *                  // iterate over both directional and style runs
- *                  for (i = 0; i < count; ++i) {
- *                      run = line.getVisualRun(i);
- *                      renderDirectionalRun(text, run.getStart(),
- *                                           run.getLimit(), run.getDirection(),
- *                                           styleRuns, styleRunCount);
- *                  }
- *              }
- *          }
- *      }
- *
- *      static void renderParagraph(String text, byte textDirection,
- *                                  StyleRun styleRuns[], int styleRunCount,
- *                                  int lineWidth) {
- *          int length = text.length();
- *          Bidi para = new Bidi();
- *          try {
- *              para.setPara(text,
- *                           textDirection != 0 ? Bidi.LEVEL_DEFAULT_RTL
- *                                              : Bidi.LEVEL_DEFAULT_LTR,
- *                           null);
- *          } catch (Exception e) {
- *              e.printStackTrace();
- *              return;
- *          }
- *          byte paraLevel = (byte)(1 & para.getParaLevel());
- *          StyleRun styleRun = new StyleRun(length, styleNormal);
- *
- *          if (styleRuns == null || styleRunCount <= 0) {
- *              styleRuns = new StyleRun[1];
- *              styleRunCount = 1;
- *              styleRuns[0] = styleRun;
- *          }
- *          // assume styleRuns[styleRunCount-1].limit>=length
- *
- *          int width = getTextWidth(text, 0, length, styleRuns, styleRunCount);
- *          if (width <= lineWidth) {
- *              // everything fits onto one line
- *
- *              // prepare rendering a new line from either left or right
- *              startLine(paraLevel, width);
- *
- *              renderLine(para, text, 0, length, styleRuns, styleRunCount);
- *          } else {
- *              // we need to render several lines
- *              Bidi line = new Bidi(length, 0);
- *              int start = 0, limit;
- *              int styleRunStart = 0, styleRunLimit;
- *
- *              for (;;) {
- *                  limit = length;
- *                  styleRunLimit = styleRunCount;
- *                  width = getLineBreak(text, new Bounds(start, limit),
- *                                       para, styleRuns,
- *                                       new Bounds(styleRunStart, styleRunLimit));
- *                  try {
- *                      line = para.setLine(start, limit);
- *                  } catch (Exception e) {
- *                      e.printStackTrace();
- *                      return;
- *                  }
- *                  // prepare rendering a new line
- *                  // from either left or right
- *                  startLine(paraLevel, width);
- *
- *                  if (styleRunStart > 0) {
- *                      int newRunCount = styleRuns.length - styleRunStart;
- *                      StyleRun[] newRuns = new StyleRun[newRunCount];
- *                      System.arraycopy(styleRuns, styleRunStart, newRuns, 0,
- *                                       newRunCount);
- *                      renderLine(line, text, start, limit, newRuns,
- *                                 styleRunLimit - styleRunStart);
- *                  } else {
- *                      renderLine(line, text, start, limit, styleRuns,
- *                                 styleRunLimit - styleRunStart);
- *                  }
- *                  if (limit == length) {
- *                      break;
- *                  }
- *                  start = limit;
- *                  styleRunStart = styleRunLimit - 1;
- *                  if (start >= styleRuns[styleRunStart].limit) {
- *                      ++styleRunStart;
- *                  }
- *              }
- *          }
- *      }
- *
- *      public static void main(String[] args)
- *      {
- *          renderParagraph("Some Latin text...", Bidi.LTR, null, 0, 80);
- *          renderParagraph("Some Hebrew text...", Bidi.RTL, null, 0, 60);
- *      }
- *  }
- *
- * }

- */
-
-/*
- * General implementation notes:
- *
- * Throughout the implementation, there are comments like (W2) that refer to
- * rules of the BiDi algorithm, in this example to the second rule of the
- * resolution of weak types.
- *
- * For handling surrogate pairs, where two UChar's form one "abstract" (or UTF-32)
- * character according to UTF-16, the second UChar gets the directional property of
- * the entire character assigned, while the first one gets a BN, a boundary
- * neutral, type, which is ignored by most of the algorithm according to
- * rule (X9) and the implementation suggestions of the BiDi algorithm.
- *
- * Later, adjustWSLevels() will set the level for each BN to that of the
- * following character (UChar), which results in surrogate pairs getting the
- * same level on each of their surrogates.
- *
- * In a UTF-8 implementation, the same thing could be done: the last byte of
- * a multi-byte sequence would get the "real" property, while all previous
- * bytes of that sequence would get BN.
- *
- * It is not possible to assign all those parts of a character the same real
- * property because this would fail in the resolution of weak types with rules
- * that look at immediately surrounding types.
- *
- * As a related topic, this implementation does not remove Boundary Neutral
- * types from the input, but ignores them wherever this is relevant.
- * For example, the loop for the resolution of the weak types reads
- * types until it finds a non-BN.
- * Also, explicit embedding codes are neither changed into BN nor removed.
- * They are only treated the same way real BNs are.
- * As stated before, adjustWSLevels() takes care of them at the end.
- * For the purpose of conformance, the levels of all these codes
- * do not matter.
- *
- * Note that this implementation modifies the dirProps
- * after the initial setup, when applying X5c (replace FSI by LRI or RLI),
- * X6, N0 (replace paired brackets by L or R).
- *
- * In this implementation, the resolution of weak types (W1 to W6),
- * neutrals (N1 and N2), and the assignment of the resolved level (In)
- * are all done in one single loop, in resolveImplicitLevels().
- * Changes of dirProp values are done on the fly, without writing
- * them back to the dirProps array.
- *
- *
- * This implementation contains code that allows to bypass steps of the
- * algorithm that are not needed on the specific paragraph
- * in order to speed up the most common cases considerably,
- * like text that is entirely LTR, or RTL text without numbers.
- *
- * Most of this is done by setting a bit for each directional property
- * in a flags variable and later checking for whether there are
- * any LTR characters or any RTL characters, or both, whether
- * there are any explicit embedding codes, etc.
- *
- * If the (Xn) steps are performed, then the flags are re-evaluated,
- * because they will then not contain the embedding codes any more
- * and will be adjusted for override codes, so that subsequently
- * more bypassing may be possible than what the initial flags suggested.
- *
- * If the text is not mixed-directional, then the
- * algorithm steps for the weak type resolution are not performed,
- * and all levels are set to the paragraph level.
- *
- * If there are no explicit embedding codes, then the (Xn) steps
- * are not performed.
- *
- * If embedding levels are supplied as a parameter, then all
- * explicit embedding codes are ignored, and the (Xn) steps
- * are not performed.
- *
- * White Space types could get the level of the run they belong to,
- * and are checked with a test of (flags&MASK_EMBEDDING) to
- * consider if the paragraph direction should be considered in
- * the flags variable.
- *
- * If there are no White Space types in the paragraph, then
- * (L1) is not necessary in adjustWSLevels().
- */
-
-public class BidiBase {
-
-    static class Point {
-        int pos;    /* position in text */
-        int flag;   /* flag for LRM/RLM, before/after */
-    }
-
-    static class InsertPoints {
-        int size;
-        int confirmed;
-        Point[] points = new Point[0];
-    }
-
-    static class Opening {
-        int   position;                 /* position of opening bracket */
-        int   match;                    /* matching char or -position of closing bracket */
-        int   contextPos;               /* position of last strong char found before opening */
-        short flags;                    /* bits for L or R/AL found within the pair */
-        byte  contextDir;               /* L or R according to last strong char before opening */
-    }
-
-    static class IsoRun {
-        int   contextPos;               /* position of char determining context */
-        short start;                    /* index of first opening entry for this run */
-        short limit;                    /* index after last opening entry for this run */
-        byte  level;                    /* level of this run */
-        byte  lastStrong;               /* bidi class of last strong char found in this run */
-        byte  lastBase;                 /* bidi class of last base char found in this run */
-        byte  contextDir;               /* L or R to use as context for following openings */
-    }
-
-    static class BracketData {
-        Opening[] openings = new Opening[SIMPLE_PARAS_COUNT];
-        int   isoRunLast;               /* index of last used entry */
-        /* array of nested isolated sequence entries; can never excess UBIDI_MAX_EXPLICIT_LEVEL
-           + 1 for index 0, + 1 for before the first isolated sequence */
-        IsoRun[]  isoRuns = new IsoRun[MAX_EXPLICIT_LEVEL+2];
-        boolean   isNumbersSpecial;     /*reordering mode for NUMBERS_SPECIAL */
-    }
-
-    static class Isolate {
-        int   startON;
-        int   start1;
-        short stateImp;
-        short state;
-    }
-
-    /** Paragraph level setting

-     *
-     * Constant indicating that the base direction depends on the first strong
-     * directional character in the text according to the Unicode Bidirectional
-     * Algorithm. If no strong directional character is present,
-     * then set the paragraph level to 0 (left-to-right).

-     *
-     * If this value is used in conjunction with reordering modes
-     * REORDER_INVERSE_LIKE_DIRECT or
-     * REORDER_INVERSE_FOR_NUMBERS_SPECIAL, the text to reorder
-     * is assumed to be visual LTR, and the text after reordering is required
-     * to be the corresponding logical string with appropriate contextual
-     * direction. The direction of the result string will be RTL if either
-     * the rightmost or leftmost strong character of the source text is RTL
-     * or Arabic Letter, the direction will be LTR otherwise.

-     *
-     * If reordering option OPTION_INSERT_MARKS is set, an RLM may
-     * be added at the beginning of the result string to ensure round trip
-     * (that the result string, when reordered back to visual, will produce
-     * the original source text).
-     * @see #REORDER_INVERSE_LIKE_DIRECT
-     * @see #REORDER_INVERSE_FOR_NUMBERS_SPECIAL
-     * @stable ICU 3.8
-     */
-    public static final byte LEVEL_DEFAULT_LTR = (byte)0x7e;
-
-    /** Paragraph level setting

-     *
-     * Constant indicating that the base direction depends on the first strong
-     * directional character in the text according to the Unicode Bidirectional
-     * Algorithm. If no strong directional character is present,
-     * then set the paragraph level to 1 (right-to-left).

-     *
-     * If this value is used in conjunction with reordering modes
-     * REORDER_INVERSE_LIKE_DIRECT or
-     * REORDER_INVERSE_FOR_NUMBERS_SPECIAL, the text to reorder
-     * is assumed to be visual LTR, and the text after reordering is required
-     * to be the corresponding logical string with appropriate contextual
-     * direction. The direction of the result string will be RTL if either
-     * the rightmost or leftmost strong character of the source text is RTL
-     * or Arabic Letter, or if the text contains no strong character;
-     * the direction will be LTR otherwise.

-     *
-     * If reordering option OPTION_INSERT_MARKS is set, an RLM may
-     * be added at the beginning of the result string to ensure round trip
-     * (that the result string, when reordered back to visual, will produce
-     * the original source text).
-     * @see #REORDER_INVERSE_LIKE_DIRECT
-     * @see #REORDER_INVERSE_FOR_NUMBERS_SPECIAL
-     * @stable ICU 3.8
-     */
-    public static final byte LEVEL_DEFAULT_RTL = (byte)0x7f;
-
-    /**
-     * Maximum explicit embedding level.
-     * (The maximum resolved level can be up to MAX_EXPLICIT_LEVEL+1).
-     * @stable ICU 3.8
-     */
-    public static final byte MAX_EXPLICIT_LEVEL = 125;
-
-    /**
-     * Bit flag for level input.
-     * Overrides directional properties.
-     * @stable ICU 3.8
-     */
-    public static final byte LEVEL_OVERRIDE = (byte)0x80;
-
-    /**
-     * Special value which can be returned by the mapping methods when a
-     * logical index has no corresponding visual index or vice-versa. This may
-     * happen for the logical-to-visual mapping of a Bidi control when option
-     * OPTION_REMOVE_CONTROLS is
-     * specified. This can also happen for the visual-to-logical mapping of a
-     * Bidi mark (LRM or RLM) inserted by option
-     * OPTION_INSERT_MARKS.
-     * @see #getVisualIndex
-     * @see #getVisualMap
-     * @see #getLogicalIndex
-     * @see #getLogicalMap
-     * @see #OPTION_INSERT_MARKS
-     * @see #OPTION_REMOVE_CONTROLS
-     * @stable ICU 3.8
-     */
-    public static final int MAP_NOWHERE = -1;
-
-    /**
-     * Left-to-right text.
-     * 
    
-     * 
  • As return value for getDirection(), it means
-     *     that the source string contains no right-to-left characters, or
-     *     that the source string is empty and the paragraph level is even.
-     * 
  • As return value for getBaseDirection(), it
-     *     means that the first strong character of the source string has
-     *     a left-to-right direction.
-     * 

-     * @stable ICU 3.8
-     */
-    public static final byte LTR = 0;
-
-    /**
-     * Right-to-left text.
-     * 
    
-     * 
  • As return value for getDirection(), it means
-     *     that the source string contains no left-to-right characters, or
-     *     that the source string is empty and the paragraph level is odd.
-     * 
  • As return value for getBaseDirection(), it
-     *     means that the first strong character of the source string has
-     *     a right-to-left direction.
-     * 

-     * @stable ICU 3.8
-     */
-    public static final byte RTL = 1;
-
-    /**
-     * Mixed-directional text.
-     * 
As return value for getDirection(), it means
-     *    that the source string contains both left-to-right and
-     *    right-to-left characters.
-     * @stable ICU 3.8
-     */
-    public static final byte MIXED = 2;
-
-    /**
-     * option bit for writeReordered():
-     * keep combining characters after their base characters in RTL runs
-     *
-     * @see #writeReordered
-     * @stable ICU 3.8
-     */
-    public static final short KEEP_BASE_COMBINING = 1;
-
-    /**
-     * option bit for writeReordered():
-     * replace characters with the "mirrored" property in RTL runs
-     * by their mirror-image mappings
-     *
-     * @see #writeReordered
-     * @stable ICU 3.8
-     */
-    public static final short DO_MIRRORING = 2;
-
-    /**
-     * option bit for writeReordered():
-     * surround the run with LRMs if necessary;
-     * this is part of the approximate "inverse Bidi" algorithm
-     *
-     * 
This option does not imply corresponding adjustment of the index
-     * mappings.

-     *
-     * @see #setInverse
-     * @see #writeReordered
-     * @stable ICU 3.8
-     */
-    public static final short INSERT_LRM_FOR_NUMERIC = 4;
-
-    /**
-     * option bit for writeReordered():
-     * remove Bidi control characters
-     * (this does not affect INSERT_LRM_FOR_NUMERIC)
-     *
-     * 
This option does not imply corresponding adjustment of the index
-     * mappings.

-     *
-     * @see #writeReordered
-     * @see #INSERT_LRM_FOR_NUMERIC
-     * @stable ICU 3.8
-     */
-    public static final short REMOVE_BIDI_CONTROLS = 8;
-
-    /**
-     * option bit for writeReordered():
-     * write the output in reverse order
-     *
-     * 
This has the same effect as calling writeReordered()
-     * first without this option, and then calling
-     * writeReverse() without mirroring.
-     * Doing this in the same step is faster and avoids a temporary buffer.
-     * An example for using this option is output to a character terminal that
-     * is designed for RTL scripts and stores text in reverse order.

-     *
-     * @see #writeReordered
-     * @stable ICU 3.8
-     */
-    public static final short OUTPUT_REVERSE = 16;
-
-    /** Reordering mode: Regular Logical to Visual Bidi algorithm according to Unicode.
-     * @see #setReorderingMode
-     * @stable ICU 3.8
-     */
-    private static final short REORDER_DEFAULT = 0;
-
-    /** Reordering mode: Logical to Visual algorithm which handles numbers in
-     * a way which mimicks the behavior of Windows XP.
-     * @see #setReorderingMode
-     * @stable ICU 3.8
-     */
-    private static final short REORDER_NUMBERS_SPECIAL = 1;
-
-    /** Reordering mode: Logical to Visual algorithm grouping numbers with
-     * adjacent R characters (reversible algorithm).
-     * @see #setReorderingMode
-     * @stable ICU 3.8
-     */
-    private static final short REORDER_GROUP_NUMBERS_WITH_R = 2;
-
-    /** Reordering mode: Reorder runs only to transform a Logical LTR string
-     * to the logical RTL string with the same display, or vice-versa.

-     * If this mode is set together with option
-     * OPTION_INSERT_MARKS, some Bidi controls in the source
-     * text may be removed and other controls may be added to produce the
-     * minimum combination which has the required display.
-     * @see #OPTION_INSERT_MARKS
-     * @see #setReorderingMode
-     * @stable ICU 3.8
-     */
-    static final short REORDER_RUNS_ONLY = 3;
-
-    /** Reordering mode: Visual to Logical algorithm which handles numbers
-     * like L (same algorithm as selected by setInverse(true).
-     * @see #setInverse
-     * @see #setReorderingMode
-     * @stable ICU 3.8
-     */
-    static final short REORDER_INVERSE_NUMBERS_AS_L = 4;
-
-    /** Reordering mode: Visual to Logical algorithm equivalent to the regular
-     * Logical to Visual algorithm.
-     * @see #setReorderingMode
-     * @stable ICU 3.8
-     */
-    static final short REORDER_INVERSE_LIKE_DIRECT = 5;
-
-    /** Reordering mode: Inverse Bidi (Visual to Logical) algorithm for the
-     * REORDER_NUMBERS_SPECIAL Bidi algorithm.
-     * @see #setReorderingMode
-     * @stable ICU 3.8
-     */
-    static final short REORDER_INVERSE_FOR_NUMBERS_SPECIAL = 6;
-
-    /* Reordering mode values must be ordered so that all the regular logical to
-     * visual modes come first, and all inverse Bidi modes come last.
-     */
-    private static final short REORDER_LAST_LOGICAL_TO_VISUAL =
-            REORDER_NUMBERS_SPECIAL;
-
-    /**
-     * Option bit for setReorderingOptions:
-     * insert Bidi marks (LRM or RLM) when needed to ensure correct result of
-     * a reordering to a Logical order
-     *
-     * 
This option must be set or reset before calling
-     * setPara.

-     *
-     * 
This option is significant only with reordering modes which generate
-     * a result with Logical order, specifically.

-     * 
    
-     *   
  • REORDER_RUNS_ONLY
    • 
-     *   
  • REORDER_INVERSE_NUMBERS_AS_L
    • 
-     *   
  • REORDER_INVERSE_LIKE_DIRECT
    • 
-     *   
  • REORDER_INVERSE_FOR_NUMBERS_SPECIAL
    • 
-     * 

-     *
-     * 
If this option is set in conjunction with reordering mode
-     * REORDER_INVERSE_NUMBERS_AS_L or with calling
-     * setInverse(true), it implies option
-     * INSERT_LRM_FOR_NUMERIC in calls to method
-     * writeReordered().

-     *
-     * 
For other reordering modes, a minimum number of LRM or RLM characters
-     * will be added to the source text after reordering it so as to ensure
-     * round trip, i.e. when applying the inverse reordering mode on the
-     * resulting logical text with removal of Bidi marks
-     * (option OPTION_REMOVE_CONTROLS set before calling
-     * setPara() or option
-     * REMOVE_BIDI_CONTROLS in
-     * writeReordered), the result will be identical to the
-     * source text in the first transformation.
-     *
-     * 
This option will be ignored if specified together with option
-     * OPTION_REMOVE_CONTROLS. It inhibits option
-     * REMOVE_BIDI_CONTROLS in calls to method
-     * writeReordered() and it implies option
-     * INSERT_LRM_FOR_NUMERIC in calls to method
-     * writeReordered() if the reordering mode is
-     * REORDER_INVERSE_NUMBERS_AS_L.

-     *
-     * @see #setReorderingMode
-     * @see #setReorderingOptions
-     * @see #INSERT_LRM_FOR_NUMERIC
-     * @see #REMOVE_BIDI_CONTROLS
-     * @see #OPTION_REMOVE_CONTROLS
-     * @see #REORDER_RUNS_ONLY
-     * @see #REORDER_INVERSE_NUMBERS_AS_L
-     * @see #REORDER_INVERSE_LIKE_DIRECT
-     * @see #REORDER_INVERSE_FOR_NUMBERS_SPECIAL
-     * @stable ICU 3.8
-     */
-    static final int OPTION_INSERT_MARKS = 1;
-
-    /**
-     * Option bit for setReorderingOptions:
-     * remove Bidi control characters
-     *
-     * 
This option must be set or reset before calling
-     * setPara.

-     *
-     * 
This option nullifies option
-     * OPTION_INSERT_MARKS. It inhibits option
-     * INSERT_LRM_FOR_NUMERIC in calls to method
-     * writeReordered() and it implies option
-     * REMOVE_BIDI_CONTROLS in calls to that method.

-     *
-     * @see #setReorderingMode
-     * @see #setReorderingOptions
-     * @see #OPTION_INSERT_MARKS
-     * @see #INSERT_LRM_FOR_NUMERIC
-     * @see #REMOVE_BIDI_CONTROLS
-     * @stable ICU 3.8
-     */
-    static final int OPTION_REMOVE_CONTROLS = 2;
-
-    /**
-     * Option bit for setReorderingOptions:
-     * process the output as part of a stream to be continued
-     *
-     * 
This option must be set or reset before calling
-     * setPara.

-     *
-     * 
This option specifies that the caller is interested in processing
-     * large text object in parts. The results of the successive calls are
-     * expected to be concatenated by the caller. Only the call for the last
-     * part will have this option bit off.

-     *
-     * 
When this option bit is on, setPara() may process
-     * less than the full source text in order to truncate the text at a
-     * meaningful boundary. The caller should call
-     * getProcessedLength() immediately after calling
-     * setPara() in order to determine how much of the source
-     * text has been processed. Source text beyond that length should be
-     * resubmitted in following calls to setPara. The
-     * processed length may be less than the length of the source text if a
-     * character preceding the last character of the source text constitutes a
-     * reasonable boundary (like a block separator) for text to be continued.

-     * If the last character of the source text constitutes a reasonable
-     * boundary, the whole text will be processed at once.

-     * If nowhere in the source text there exists
-     * such a reasonable boundary, the processed length will be zero.

-     * The caller should check for such an occurrence and do one of the following:
-     * 
  • submit a larger amount of text with a better chance to include
-     *         a reasonable boundary.
    • 
-     *     
  • resubmit the same text after turning off option
-     *         OPTION_STREAMING.

-     * In all cases, this option should be turned off before processing the last
-     * part of the text.

-     *
-     * 
When the OPTION_STREAMING option is used, it is
-     * recommended to call orderParagraphsLTR(true) before calling
-     * setPara() so that later paragraphs may be concatenated to
-     * previous paragraphs on the right.
-     * 

-     *
-     * @see #setReorderingMode
-     * @see #setReorderingOptions
-     * @see #getProcessedLength
-     * @stable ICU 3.8
-     */
-    private static final int OPTION_STREAMING = 4;
-
-    /*
-     *   Comparing the description of the Bidi algorithm with this implementation
-     *   is easier with the same names for the Bidi types in the code as there.
-     *   See UCharacterDirection
-     */
-    /* private */ static final byte L   = 0;
-    private static final byte R   = 1;
-    private static final byte EN  = 2;
-    private static final byte ES  = 3;
-    private static final byte ET  = 4;
-    private static final byte AN  = 5;
-    private static final byte CS  = 6;
-    static final byte B   = 7;
-    private static final byte S   = 8;
-    private static final byte WS  = 9;
-    private static final byte ON  = 10;
-    private static final byte LRE = 11;
-    private static final byte LRO = 12;
-    private static final byte AL  = 13;
-    private static final byte RLE = 14;
-    private static final byte RLO = 15;
-    private static final byte PDF = 16;
-    private static final byte NSM = 17;
-    private static final byte BN  = 18;
-    private static final byte FSI = 19;
-    private static final byte LRI = 20;
-    private static final byte RLI = 21;
-    private static final byte PDI = 22;
-    private static final byte ENL = PDI + 1;    /* EN after W7 */
-    private static final byte ENR = ENL + 1;    /* EN not subject to W7 */
-
-    // Number of directional types
-    private static final int CHAR_DIRECTION_COUNT       = 23;
-
-    /**
-     * Enumerated property Bidi_Paired_Bracket_Type (new in Unicode 6.3).
-     * Used in UAX #9: Unicode Bidirectional Algorithm
-     * (http://www.unicode.org/reports/tr9/)
-     * Returns UCharacter.BidiPairedBracketType values.
-     * @stable ICU 52
-     */
-    public static final int BIDI_PAIRED_BRACKET_TYPE = 0x1015;
-
-    /**
-     * Bidi Paired Bracket Type constants.
-     *
-     * @see UProperty#BIDI_PAIRED_BRACKET_TYPE
-     * @stable ICU 52
-     */
-    public static interface BidiPairedBracketType {
-        /**
-         * Not a paired bracket.
-         * @stable ICU 52
-         */
-        public static final int NONE = 0;
-        /**
-         * Open paired bracket.
-         * @stable ICU 52
-         */
-        public static final int OPEN = 1;
-        /**
-         * Close paired bracket.
-         * @stable ICU 52
-         */
-        public static final int CLOSE = 2;
-        /**
-         * @stable ICU 52
-         */
-        public static final int COUNT = 3;
-    }
-
-    /* number of paras entries allocated initially */
-    static final int SIMPLE_PARAS_COUNT = 10;
-
-    private static final char CR = '\r';
-    private static final char LF = '\n';
-
-    static final int LRM_BEFORE = 1;
-    static final int LRM_AFTER = 2;
-    static final int RLM_BEFORE = 4;
-    static final int RLM_AFTER = 8;
-
-    /* flags for Opening.flags */
-    static final byte FOUND_L = (byte)DirPropFlag(L);
-    static final byte FOUND_R = (byte)DirPropFlag(R);
-
-    /*
-     * The following bit is used for the directional isolate status.
-     * Stack entries corresponding to isolate sequences are greater than ISOLATE.
-     */
-    static final int ISOLATE = 0x0100;
-
-    /*
-     * reference to parent paragraph object (reference to self if this object is
-     * a paragraph object); set to null in a newly opened object; set to a
-     * real value after a successful execution of setPara or setLine
-     */
-    BidiBase            paraBidi;
-
-    final UBiDiProps    bdp;
-
-    /* character array representing the current text */
-    char[]              text;
-
-    /* length of the current text */
-    int                 originalLength;
-
-    /* if the option OPTION_STREAMING is set, this is the length of
-     * text actually processed by setPara, which may be shorter
-     * than the original length. Otherwise, it is identical to the original
-     * length.
-     */
-    public int                 length;
-
-    /* if option OPTION_REMOVE_CONTROLS is set, and/or Bidi
-     * marks are allowed to be inserted in one of the reordering modes, the
-     * length of the result string may be different from the processed length.
-     */
-    int                 resultLength;
-
-    /* indicators for whether memory may be allocated after construction */
-    boolean             mayAllocateText;
-    boolean             mayAllocateRuns;
-
-    /* arrays with one value per text-character */
-    byte[]              dirPropsMemory = new byte[1];
-    byte[]              levelsMemory = new byte[1];
-    byte[]              dirProps;
-    byte[]              levels;
-
-    /* are we performing an approximation of the "inverse Bidi" algorithm? */
-    boolean             isInverse;
-
-    /* are we using the basic algorithm or its variation? */
-    int                 reorderingMode;
-
-    /* bitmask for reordering options */
-    int                 reorderingOptions;
-
-    /* must block separators receive level 0? */
-    boolean             orderParagraphsLTR;
-
-    /* the paragraph level */
-    byte                paraLevel;
-
-    /* original paraLevel when contextual */
-    /* must be one of DEFAULT_xxx or 0 if not contextual */
-    byte                defaultParaLevel;
-
-    /* the following is set in setPara, used in processPropertySeq */
-
-    ImpTabPair          impTabPair;  /* reference to levels state table pair */
-
-    /* the overall paragraph or line directionality*/
-    byte                direction;
-
-    /* flags is a bit set for which directional properties are in the text */
-    int                 flags;
-
-    /* lastArabicPos is index to the last AL in the text, -1 if none */
-    int                 lastArabicPos;
-
-    /* characters after trailingWSStart are WS and are */
-    /* implicitly at the paraLevel (rule (L1)) - levels may not reflect that */
-    int                 trailingWSStart;
-
-    /* fields for paragraph handling, set in getDirProps() */
-    int                 paraCount;
-    int[]               paras_limit = new int[SIMPLE_PARAS_COUNT];
-    byte[]              paras_level = new byte[SIMPLE_PARAS_COUNT];
-
-    /* fields for line reordering */
-    int                 runCount;     /* ==-1: runs not set up yet */
-    BidiRun[]           runsMemory = new BidiRun[0];
-    BidiRun[]           runs;
-
-    /* for non-mixed text, we only need a tiny array of runs (no allocation) */
-    BidiRun[]           simpleRuns = {new BidiRun()};
-
-    /* fields for managing isolate sequences */
-    Isolate[]           isolates;
-
-    /* maximum or current nesting depth of isolate sequences */
-    /* Within resolveExplicitLevels() and checkExplicitLevels(), this is the maximal
-       nesting encountered.
-       Within resolveImplicitLevels(), this is the index of the current isolates
-       stack entry. */
-    int                 isolateCount;
-
-    /* mapping of runs in logical order to visual order */
-    int[]               logicalToVisualRunsMap;
-    /* flag to indicate that the map has been updated */
-    boolean             isGoodLogicalToVisualRunsMap;
-
-    /* for inverse Bidi with insertion of directional marks */
-    InsertPoints        insertPoints = new InsertPoints();
-
-    /* for option OPTION_REMOVE_CONTROLS */
-    int                 controlCount;
-
-    /*
-     * Sometimes, bit values are more appropriate
-     * to deal with directionality properties.
-     * Abbreviations in these method names refer to names
-     * used in the Bidi algorithm.
-     */
-    static int DirPropFlag(byte dir) {
-        return (1 << dir);
-    }
-
-    boolean testDirPropFlagAt(int flag, int index) {
-        return ((DirPropFlag(dirProps[index]) & flag) != 0);
-    }
-
-    static final int DirPropFlagMultiRuns = DirPropFlag((byte)31);
-
-    /* to avoid some conditional statements, use tiny constant arrays */
-    static final int DirPropFlagLR[] = { DirPropFlag(L), DirPropFlag(R) };
-    static final int DirPropFlagE[] = { DirPropFlag(LRE), DirPropFlag(RLE) };
-    static final int DirPropFlagO[] = { DirPropFlag(LRO), DirPropFlag(RLO) };
-
-    static final int DirPropFlagLR(byte level) { return DirPropFlagLR[level & 1]; }
-    static final int DirPropFlagE(byte level)  { return DirPropFlagE[level & 1]; }
-    static final int DirPropFlagO(byte level)  { return DirPropFlagO[level & 1]; }
-    static final byte DirFromStrong(byte strong) { return strong == L ? L : R; }
-    static final byte NoOverride(byte level) { return (byte)(level & ~LEVEL_OVERRIDE); }
-
-    /*  are there any characters that are LTR or RTL? */
-    static final int MASK_LTR =
-        DirPropFlag(L)|DirPropFlag(EN)|DirPropFlag(ENL)|DirPropFlag(ENR)|DirPropFlag(AN)|DirPropFlag(LRE)|DirPropFlag(LRO)|DirPropFlag(LRI);
-    static final int MASK_RTL = DirPropFlag(R)|DirPropFlag(AL)|DirPropFlag(RLE)|DirPropFlag(RLO)|DirPropFlag(RLI);
-
-    static final int MASK_R_AL = DirPropFlag(R)|DirPropFlag(AL);
-
-    /* explicit embedding codes */
-    private static final int MASK_EXPLICIT = DirPropFlag(LRE)|DirPropFlag(LRO)|DirPropFlag(RLE)|DirPropFlag(RLO)|DirPropFlag(PDF);
-    private static final int MASK_BN_EXPLICIT = DirPropFlag(BN)|MASK_EXPLICIT;
-
-    /* explicit isolate codes */
-    private static final int MASK_ISO = DirPropFlag(LRI)|DirPropFlag(RLI)|DirPropFlag(FSI)|DirPropFlag(PDI);
-
-    /* paragraph and segment separators */
-    private static final int MASK_B_S = DirPropFlag(B)|DirPropFlag(S);
-
-    /* all types that are counted as White Space or Neutral in some steps */
-    static final int MASK_WS = MASK_B_S|DirPropFlag(WS)|MASK_BN_EXPLICIT|MASK_ISO;
-
-    /* types that are neutrals or could becomes neutrals in (Wn) */
-    private static final int MASK_POSSIBLE_N = DirPropFlag(ON)|DirPropFlag(CS)|DirPropFlag(ES)|DirPropFlag(ET)|MASK_WS;
-
-    /*
-     * These types may be changed to "e",
-     * the embedding type (L or R) of the run,
-     * in the Bidi algorithm (N2)
-     */
-    private static final int MASK_EMBEDDING = DirPropFlag(NSM)|MASK_POSSIBLE_N;
-
-    /*
-     *  the dirProp's L and R are defined to 0 and 1 values in UCharacterDirection.java
-     */
-    private static byte GetLRFromLevel(byte level)
-    {
-        return (byte)(level & 1);
-    }
-
-    private static boolean IsDefaultLevel(byte level)
-    {
-        return ((level & LEVEL_DEFAULT_LTR) == LEVEL_DEFAULT_LTR);
-    }
-
-    static boolean IsBidiControlChar(int c)
-    {
-        /* check for range 0x200c to 0x200f (ZWNJ, ZWJ, LRM, RLM) or
-                           0x202a to 0x202e (LRE, RLE, PDF, LRO, RLO) */
-        return (((c & 0xfffffffc) == 0x200c) || ((c >= 0x202a) && (c <= 0x202e))
-                                             || ((c >= 0x2066) && (c <= 0x2069)));
-    }
-
-    void verifyValidPara()
-    {
-        if (!(this == this.paraBidi)) {
-            throw new IllegalStateException();
-        }
-    }
-
-    void verifyValidParaOrLine()
-    {
-        BidiBase para = this.paraBidi;
-        /* verify Para */
-        if (this == para) {
-            return;
-        }
-        /* verify Line */
-        if ((para == null) || (para != para.paraBidi)) {
-            throw new IllegalStateException();
-        }
-    }
-
-    void verifyRange(int index, int start, int limit)
-    {
-        if (index < start || index >= limit) {
-            throw new IllegalArgumentException("Value " + index +
-                      " is out of range " + start + " to " + limit);
-        }
-    }
-
-    /**
-     * Allocate a Bidi object with preallocated memory
-     * for internal structures.
-     * This method provides a Bidi object like the default constructor
-     * but it also preallocates memory for internal structures
-     * according to the sizings supplied by the caller.

-     * The preallocation can be limited to some of the internal memory
-     * by setting some values to 0 here. That means that if, e.g.,
-     * maxRunCount cannot be reasonably predetermined and should not
-     * be set to maxLength (the only failproof value) to avoid
-     * wasting  memory, then maxRunCount could be set to 0 here
-     * and the internal structures that are associated with it will be allocated
-     * on demand, just like with the default constructor.
-     *
-     * @param maxLength is the maximum text or line length that internal memory
-     *        will be preallocated for. An attempt to associate this object with a
-     *        longer text will fail, unless this value is 0, which leaves the allocation
-     *        up to the implementation.
-     *
-     * @param maxRunCount is the maximum anticipated number of same-level runs
-     *        that internal memory will be preallocated for. An attempt to access
-     *        visual runs on an object that was not preallocated for as many runs
-     *        as the text was actually resolved to will fail,
-     *        unless this value is 0, which leaves the allocation up to the implementation.


-     *        The number of runs depends on the actual text and maybe anywhere between
-     *        1 and maxLength. It is typically small.
-     *
-     * @throws IllegalArgumentException if maxLength or maxRunCount is less than 0
-     * @stable ICU 3.8
-     */
-    public BidiBase(int maxLength, int maxRunCount)
-    {
-        /* check the argument values */
-        if (maxLength < 0 || maxRunCount < 0) {
-            throw new IllegalArgumentException();
-        }
-
-        /* reset the object, all reference variables null, all flags false,
-           all sizes 0.
-           In fact, we don't need to do anything, since class members are
-           initialized as zero when an instance is created.
-         */
-        /*
-        mayAllocateText = false;
-        mayAllocateRuns = false;
-        orderParagraphsLTR = false;
-        paraCount = 0;
-        runCount = 0;
-        trailingWSStart = 0;
-        flags = 0;
-        paraLevel = 0;
-        defaultParaLevel = 0;
-        direction = 0;
-        */
-        /* get Bidi properties */
-        bdp = UBiDiProps.INSTANCE;
-
-        /* allocate memory for arrays as requested */
-        if (maxLength > 0) {
-            getInitialDirPropsMemory(maxLength);
-            getInitialLevelsMemory(maxLength);
-        } else {
-            mayAllocateText = true;
-        }
-
-        if (maxRunCount > 0) {
-            // if maxRunCount == 1, use simpleRuns[]
-            if (maxRunCount > 1) {
-                getInitialRunsMemory(maxRunCount);
-            }
-        } else {
-            mayAllocateRuns = true;
-        }
-    }
-
-    /*
-     * We are allowed to allocate memory if object==null or
-     * mayAllocate==true for each array that we need.
-     *
-     * Assume sizeNeeded>0.
-     * If object != null, then assume size > 0.
-     */
-    private Object getMemory(String label, Object array, Class arrayClass,
-            boolean mayAllocate, int sizeNeeded)
-    {
-        int len = Array.getLength(array);
-
-        /* we have at least enough memory and must not allocate */
-        if (sizeNeeded == len) {
-            return array;
-        }
-        if (!mayAllocate) {
-            /* we must not allocate */
-            if (sizeNeeded <= len) {
-                return array;
-            }
-            throw new OutOfMemoryError("Failed to allocate memory for "
-                                       + label);
-        }
-        /* we may try to grow or shrink */
-        /* FOOD FOR THOUGHT: when shrinking it should be possible to avoid
-           the allocation altogether and rely on this.length */
-        try {
-            return Array.newInstance(arrayClass, sizeNeeded);
-        } catch (Exception e) {
-            throw new OutOfMemoryError("Failed to allocate memory for "
-                                       + label);
-        }
-    }
-
-    /* helper methods for each allocated array */
-    private void getDirPropsMemory(boolean mayAllocate, int len)
-    {
-        Object array = getMemory("DirProps", dirPropsMemory, Byte.TYPE, mayAllocate, len);
-        dirPropsMemory = (byte[]) array;
-    }
-
-    void getDirPropsMemory(int len)
-    {
-        getDirPropsMemory(mayAllocateText, len);
-    }
-
-    private void getLevelsMemory(boolean mayAllocate, int len)
-    {
-        Object array = getMemory("Levels", levelsMemory, Byte.TYPE, mayAllocate, len);
-        levelsMemory = (byte[]) array;
-    }
-
-    void getLevelsMemory(int len)
-    {
-        getLevelsMemory(mayAllocateText, len);
-    }
-
-    private void getRunsMemory(boolean mayAllocate, int len)
-    {
-        Object array = getMemory("Runs", runsMemory, BidiRun.class, mayAllocate, len);
-        runsMemory = (BidiRun[]) array;
-    }
-
-    void getRunsMemory(int len)
-    {
-        getRunsMemory(mayAllocateRuns, len);
-    }
-
-    /* additional methods used by constructor - always allow allocation */
-    private void getInitialDirPropsMemory(int len)
-    {
-        getDirPropsMemory(true, len);
-    }
-
-    private void getInitialLevelsMemory(int len)
-    {
-        getLevelsMemory(true, len);
-    }
-
-    private void getInitialRunsMemory(int len)
-    {
-        getRunsMemory(true, len);
-    }
-
-    /**
-     * Is this Bidi object set to perform the inverse Bidi
-     * algorithm?
-     * 
Note: calling this method after setting the reordering mode with
-     * setReorderingMode will return true if the
-     * reordering mode was set to
-     * REORDER_INVERSE_NUMBERS_AS_L, false
-     * for all other values.

-     *
-     * @return true if the Bidi object is set to
-     * perform the inverse Bidi algorithm by handling numbers as L.
-     *
-     * @see #setInverse
-     * @see #setReorderingMode
-     * @see #REORDER_INVERSE_NUMBERS_AS_L
-     * @stable ICU 3.8
-     */
-    public boolean isInverse() {
-        return isInverse;
-    }
-
-    /* perform (P2)..(P3) ------------------------------------------------------- */
-
-    /*
-     * Check that there are enough entries in the arrays paras_limit and paras_level
-     */
-    private void checkParaCount() {
-        int[] saveLimits;
-        byte[] saveLevels;
-        int count = paraCount;
-        if (count <= paras_level.length)
-            return;
-        int oldLength = paras_level.length;
-        saveLimits = paras_limit;
-        saveLevels = paras_level;
-        try {
-            paras_limit = new int[count * 2];
-            paras_level = new byte[count * 2];
-        } catch (Exception e) {
-            throw new OutOfMemoryError("Failed to allocate memory for paras");
-        }
-        System.arraycopy(saveLimits, 0, paras_limit, 0, oldLength);
-        System.arraycopy(saveLevels, 0, paras_level, 0, oldLength);
-    }
-
-    /*
-     * Get the directional properties for the text, calculate the flags bit-set, and
-     * determine the paragraph level if necessary (in paras_level[i]).
-     * FSI initiators are also resolved and their dirProp replaced with LRI or RLI.
-     * When encountering an FSI, it is initially replaced with an LRI, which is the
-     * default. Only if a strong R or AL is found within its scope will the LRI be
-     * replaced by an RLI.
-     */
-    static final int NOT_SEEKING_STRONG = 0;        /* 0: not contextual paraLevel, not after FSI */
-    static final int SEEKING_STRONG_FOR_PARA = 1;   /* 1: looking for first strong char in para */
-    static final int SEEKING_STRONG_FOR_FSI = 2;    /* 2: looking for first strong after FSI */
-    static final int LOOKING_FOR_PDI = 3;           /* 3: found strong after FSI, looking for PDI */
-
-    private void getDirProps()
-    {
-        int i = 0, i0, i1;
-        flags = 0;          /* collect all directionalities in the text */
-        int uchar;
-        byte dirProp;
-        byte defaultParaLevel = 0;   /* initialize to avoid compiler warnings */
-        boolean isDefaultLevel = IsDefaultLevel(paraLevel);
-        /* for inverse Bidi, the default para level is set to RTL if there is a
-           strong R or AL character at either end of the text                */
-        boolean isDefaultLevelInverse=isDefaultLevel &&
-                (reorderingMode == REORDER_INVERSE_LIKE_DIRECT ||
-                 reorderingMode == REORDER_INVERSE_FOR_NUMBERS_SPECIAL);
-        lastArabicPos = -1;
-        int controlCount = 0;
-        boolean removeBidiControls = (reorderingOptions & OPTION_REMOVE_CONTROLS) != 0;
-
-        byte state;
-        byte lastStrong = ON;           /* for default level & inverse Bidi */
-    /* The following stacks are used to manage isolate sequences. Those
-       sequences may be nested, but obviously never more deeply than the
-       maximum explicit embedding level.
-       lastStack is the index of the last used entry in the stack. A value of -1
-       means that there is no open isolate sequence.
-       lastStack is reset to -1 on paragraph boundaries. */
-    /* The following stack contains the position of the initiator of
-       each open isolate sequence */
-        int[] isolateStartStack= new int[MAX_EXPLICIT_LEVEL+1];
-    /* The following stack contains the last known state before
-       encountering the initiator of an isolate sequence */
-        byte[] previousStateStack = new byte[MAX_EXPLICIT_LEVEL+1];
-        int  stackLast=-1;
-
-        if ((reorderingOptions & OPTION_STREAMING) != 0)
-            length = 0;
-        defaultParaLevel = (byte)(paraLevel & 1);
-
-        if (isDefaultLevel) {
-            paras_level[0] = defaultParaLevel;
-            lastStrong = defaultParaLevel;
-            state = SEEKING_STRONG_FOR_PARA;
-        } else {
-            paras_level[0] = paraLevel;
-            state = NOT_SEEKING_STRONG;
-        }
-        /* count paragraphs and determine the paragraph level (P2..P3) */
-        /*
-         * see comment on constant fields:
-         * the LEVEL_DEFAULT_XXX values are designed so that
-         * their low-order bit alone yields the intended default
-         */
-
-        for (i = 0; i < originalLength; /* i is incremented in the loop */) {
-            i0 = i;                     /* index of first code unit */
-            uchar = UTF16.charAt(text, 0, originalLength, i);
-            i += UTF16.getCharCount(uchar);
-            i1 = i - 1; /* index of last code unit, gets the directional property */
-
-            dirProp = (byte)getCustomizedClass(uchar);
-            flags |= DirPropFlag(dirProp);
-            dirProps[i1] = dirProp;
-            if (i1 > i0) {     /* set previous code units' properties to BN */
-                flags |= DirPropFlag(BN);
-                do {
-                    dirProps[--i1] = BN;
-                } while (i1 > i0);
-            }
-            if (removeBidiControls && IsBidiControlChar(uchar)) {
-                controlCount++;
-            }
-            if (dirProp == L) {
-                if (state == SEEKING_STRONG_FOR_PARA) {
-                    paras_level[paraCount - 1] = 0;
-                    state = NOT_SEEKING_STRONG;
-                }
-                else if (state == SEEKING_STRONG_FOR_FSI) {
-                    if (stackLast <= MAX_EXPLICIT_LEVEL) {
-                        /* no need for next statement, already set by default */
-                        /* dirProps[isolateStartStack[stackLast]] = LRI; */
-                        flags |= DirPropFlag(LRI);
-                    }
-                    state = LOOKING_FOR_PDI;
-                }
-                lastStrong = L;
-                continue;
-            }
-            if (dirProp == R || dirProp == AL) {
-                if (state == SEEKING_STRONG_FOR_PARA) {
-                    paras_level[paraCount - 1] = 1;
-                    state = NOT_SEEKING_STRONG;
-                }
-                else if (state == SEEKING_STRONG_FOR_FSI) {
-                    if (stackLast <= MAX_EXPLICIT_LEVEL) {
-                        dirProps[isolateStartStack[stackLast]] = RLI;
-                        flags |= DirPropFlag(RLI);
-                    }
-                    state = LOOKING_FOR_PDI;
-                }
-                lastStrong = R;
-                if (dirProp == AL)
-                    lastArabicPos = i - 1;
-                continue;
-            }
-            if (dirProp >= FSI && dirProp <= RLI) { /* FSI, LRI or RLI */
-                stackLast++;
-                if (stackLast <= MAX_EXPLICIT_LEVEL) {
-                    isolateStartStack[stackLast] = i - 1;
-                    previousStateStack[stackLast] = state;
-                }
-                if (dirProp == FSI) {
-                    dirProps[i-1] = LRI;    /* default if no strong char */
-                    state = SEEKING_STRONG_FOR_FSI;
-                }
-                else
-                    state = LOOKING_FOR_PDI;
-                continue;
-            }
-            if (dirProp == PDI) {
-                if (state == SEEKING_STRONG_FOR_FSI) {
-                    if (stackLast <= MAX_EXPLICIT_LEVEL) {
-                        /* no need for next statement, already set by default */
-                        /* dirProps[isolateStartStack[stackLast]] = LRI; */
-                        flags |= DirPropFlag(LRI);
-                    }
-                }
-                if (stackLast >= 0) {
-                    if (stackLast <= MAX_EXPLICIT_LEVEL)
-                        state = previousStateStack[stackLast];
-                    stackLast--;
-                }
-                continue;
-            }
-            if (dirProp == B) {
-                if (i < originalLength && uchar == CR && text[i] == LF) /* do nothing on the CR */
-                    continue;
-                paras_limit[paraCount - 1] = i;
-                if (isDefaultLevelInverse && lastStrong == R)
-                    paras_level[paraCount - 1] = 1;
-                if ((reorderingOptions & OPTION_STREAMING) != 0) {
-                /* When streaming, we only process whole paragraphs
-                   thus some updates are only done on paragraph boundaries */
-                   length = i;          /* i is index to next character */
-                   this.controlCount = controlCount;
-                }
-                if (i < originalLength) {       /* B not last char in text */
-                    paraCount++;
-                    checkParaCount();   /* check that there is enough memory for a new para entry */
-                    if (isDefaultLevel) {
-                        paras_level[paraCount - 1] = defaultParaLevel;
-                        state = SEEKING_STRONG_FOR_PARA;
-                        lastStrong = defaultParaLevel;
-                    } else {
-                        paras_level[paraCount - 1] = paraLevel;
-                        state = NOT_SEEKING_STRONG;
-                    }
-                    stackLast = -1;
-                }
-                continue;
-            }
-        }
-        /* +Ignore still open isolate sequences with overflow */
-      if (stackLast > MAX_EXPLICIT_LEVEL) {
-            stackLast = MAX_EXPLICIT_LEVEL;
-            state=SEEKING_STRONG_FOR_FSI;   /* to be on the safe side */
-        }
-        /* Resolve direction of still unresolved open FSI sequences */
-        while (stackLast >= 0) {
-            if (state == SEEKING_STRONG_FOR_FSI) {
-                /* no need for next statement, already set by default */
-                /* dirProps[isolateStartStack[stackLast]] = LRI; */
-                flags |= DirPropFlag(LRI);
-                break;
-            }
-            state = previousStateStack[stackLast];
-            stackLast--;
-        }
-        /* When streaming, ignore text after the last paragraph separator */
-        if ((reorderingOptions & OPTION_STREAMING) != 0) {
-            if (length < originalLength)
-                paraCount--;
-        } else {
-            paras_limit[paraCount - 1] = originalLength;
-            this.controlCount = controlCount;
-        }
-        /* For inverse bidi, default para direction is RTL if there is
-           a strong R or AL at either end of the paragraph */
-        if (isDefaultLevelInverse && lastStrong == R) {
-            paras_level[paraCount - 1] = 1;
-        }
-        if (isDefaultLevel) {
-            paraLevel = paras_level[0];
-        }
-        /* The following is needed to resolve the text direction for default level
-           paragraphs containing no strong character */
-        for (i = 0; i < paraCount; i++)
-            flags |= DirPropFlagLR(paras_level[i]);
-
-        if (orderParagraphsLTR && (flags & DirPropFlag(B)) != 0) {
-            flags |= DirPropFlag(L);
-        }
-    }
-
-    /* determine the paragraph level at position index */
-    byte GetParaLevelAt(int pindex)
-    {
-        if (defaultParaLevel == 0 || pindex < paras_limit[0])
-            return paraLevel;
-        int i;
-        for (i = 1; i < paraCount; i++)
-            if (pindex < paras_limit[i])
-                break;
-        if (i >= paraCount)
-            i = paraCount - 1;
-        return paras_level[i];
-    }
-
-    /* Functions for handling paired brackets ----------------------------------- */
-
-    /* In the isoRuns array, the first entry is used for text outside of any
-       isolate sequence.  Higher entries are used for each more deeply nested
-       isolate sequence. isoRunLast is the index of the last used entry.  The
-       openings array is used to note the data of opening brackets not yet
-       matched by a closing bracket, or matched but still susceptible to change
-       level.
-       Each isoRun entry contains the index of the first and
-       one-after-last openings entries for pending opening brackets it
-       contains.  The next openings entry to use is the one-after-last of the
-       most deeply nested isoRun entry.
-       isoRun entries also contain their current embedding level and the last
-       encountered strong character, since these will be needed to resolve
-       the level of paired brackets.  */
-
-    private void bracketInit(BracketData bd) {
-        bd.isoRunLast = 0;
-        bd.isoRuns[0] = new IsoRun();
-        bd.isoRuns[0].start = 0;
-        bd.isoRuns[0].limit = 0;
-        bd.isoRuns[0].level = GetParaLevelAt(0);
-        bd.isoRuns[0].lastStrong = bd.isoRuns[0].lastBase = bd.isoRuns[0].contextDir = (byte)(GetParaLevelAt(0) & 1);
-        bd.isoRuns[0].contextPos = 0;
-        bd.openings = new Opening[SIMPLE_PARAS_COUNT];
-        bd.isNumbersSpecial = reorderingMode == REORDER_NUMBERS_SPECIAL ||
-                              reorderingMode == REORDER_INVERSE_FOR_NUMBERS_SPECIAL;
-    }
-
-    /* paragraph boundary */
-    private void bracketProcessB(BracketData bd, byte level) {
-        bd.isoRunLast = 0;
-        bd.isoRuns[0].limit = 0;
-        bd.isoRuns[0].level = level;
-        bd.isoRuns[0].lastStrong = bd.isoRuns[0].lastBase = bd.isoRuns[0].contextDir = (byte)(level & 1);
-        bd.isoRuns[0].contextPos = 0;
-    }
-
-    /* LRE, LRO, RLE, RLO, PDF */
-    private void bracketProcessBoundary(BracketData bd, int lastCcPos,
-                                        byte contextLevel, byte embeddingLevel) {
-        IsoRun pLastIsoRun = bd.isoRuns[bd.isoRunLast];
-        if ((DirPropFlag(dirProps[lastCcPos]) & MASK_ISO) != 0) /* after an isolate */
-            return;
-        if (NoOverride(embeddingLevel) > NoOverride(contextLevel))  /* not a PDF */
-            contextLevel = embeddingLevel;
-        pLastIsoRun.limit = pLastIsoRun.start;
-        pLastIsoRun.level = embeddingLevel;
-        pLastIsoRun.lastStrong = pLastIsoRun.lastBase = pLastIsoRun.contextDir = (byte)(contextLevel & 1);
-        pLastIsoRun.contextPos = lastCcPos;
-    }
-
-    /* LRI or RLI */
-    private void bracketProcessLRI_RLI(BracketData bd, byte level) {
-        IsoRun pLastIsoRun = bd.isoRuns[bd.isoRunLast];
-        short lastLimit;
-        pLastIsoRun.lastBase = ON;
-        lastLimit = pLastIsoRun.limit;
-        bd.isoRunLast++;
-        pLastIsoRun = bd.isoRuns[bd.isoRunLast];
-        if (pLastIsoRun == null)
-            pLastIsoRun = bd.isoRuns[bd.isoRunLast] = new IsoRun();
-        pLastIsoRun.start = pLastIsoRun.limit = lastLimit;
-        pLastIsoRun.level = level;
-        pLastIsoRun.lastStrong = pLastIsoRun.lastBase = pLastIsoRun.contextDir = (byte)(level & 1);
-        pLastIsoRun.contextPos = 0;
-    }
-
-    /* PDI */
-    private void bracketProcessPDI(BracketData bd) {
-        IsoRun pLastIsoRun;
-        bd.isoRunLast--;
-        pLastIsoRun = bd.isoRuns[bd.isoRunLast];
-        pLastIsoRun.lastBase = ON;
-    }
-
-    /* newly found opening bracket: create an openings entry */
-    private void bracketAddOpening(BracketData bd, char match, int position) {
-        IsoRun pLastIsoRun = bd.isoRuns[bd.isoRunLast];
-        Opening pOpening;
-        if (pLastIsoRun.limit >= bd.openings.length) {  /* no available new entry */
-            Opening[] saveOpenings = bd.openings;
-            int count;
-            try {
-                count = bd.openings.length;
-                bd.openings = new Opening[count * 2];
-            } catch (Exception e) {
-                throw new OutOfMemoryError("Failed to allocate memory for openings");
-            }
-            System.arraycopy(saveOpenings, 0, bd.openings, 0, count);
-        }
-        pOpening = bd.openings[pLastIsoRun.limit];
-        if (pOpening == null)
-            pOpening = bd.openings[pLastIsoRun.limit]= new Opening();
-        pOpening.position = position;
-        pOpening.match = match;
-        pOpening.contextDir = pLastIsoRun.contextDir;
-        pOpening.contextPos = pLastIsoRun.contextPos;
-        pOpening.flags = 0;
-        pLastIsoRun.limit++;
-    }
-
-    /* change N0c1 to N0c2 when a preceding bracket is assigned the embedding level */
-    private void fixN0c(BracketData bd, int openingIndex, int newPropPosition, byte newProp) {
-        /* This function calls itself recursively */
-        IsoRun pLastIsoRun = bd.isoRuns[bd.isoRunLast];
-        Opening qOpening;
-        int k, openingPosition, closingPosition;
-        for (k = openingIndex+1; k < pLastIsoRun.limit; k++) {
-            qOpening = bd.openings[k];
-            if (qOpening.match >= 0)    /* not an N0c match */
-                continue;
-            if (newPropPosition < qOpening.contextPos)
-                break;
-            if (newPropPosition >= qOpening.position)
-                continue;
-            if (newProp == qOpening.contextDir)
-                break;
-            openingPosition = qOpening.position;
-            dirProps[openingPosition] = newProp;
-            closingPosition = -(qOpening.match);
-            dirProps[closingPosition] = newProp;
-            qOpening.match = 0;                                 /* prevent further changes */
-            fixN0c(bd, k, openingPosition, newProp);
-            fixN0c(bd, k, closingPosition, newProp);
-        }
-    }
-
-    /* process closing bracket; return L or R if N0b or N0c, ON if N0d */
-    private byte bracketProcessClosing(BracketData bd, int openIdx, int position) {
-        IsoRun pLastIsoRun = bd.isoRuns[bd.isoRunLast];
-        Opening pOpening, qOpening;
-        byte direction;
-        boolean stable;
-        byte newProp;
-        pOpening = bd.openings[openIdx];
-        direction = (byte)(pLastIsoRun.level & 1);
-        stable = true;          /* assume stable until proved otherwise */
-
-        /* The stable flag is set when brackets are paired and their
-           level is resolved and cannot be changed by what will be
-           found later in the source string.
-           An unstable match can occur only when applying N0c, where
-           the resolved level depends on the preceding context, and
-           this context may be affected by text occurring later.
-           Example: RTL paragraph containing:  abc[(latin) HEBREW]
-           When the closing parenthesis is encountered, it appears
-           that N0c1 must be applied since 'abc' sets an opposite
-           direction context and both parentheses receive level 2.
-           However, when the closing square bracket is processed,
-           N0b applies because of 'HEBREW' being included within the
-           brackets, thus the square brackets are treated like R and
-           receive level 1. However, this changes the preceding
-           context of the opening parenthesis, and it now appears
-           that N0c2 must be applied to the parentheses rather than
-           N0c1. */
-
-            if ((direction == 0 && (pOpening.flags & FOUND_L) > 0) ||
-                (direction == 1 && (pOpening.flags & FOUND_R) > 0)) {   /* N0b */
-                newProp = direction;
-            }
-            else if ((pOpening.flags & (FOUND_L | FOUND_R)) != 0) {     /* N0c */
-                    /* it is stable if there is no preceding text or in
-                       conditions too complicated and not worth checking */
-                    stable = (openIdx == pLastIsoRun.start);
-                if (direction != pOpening.contextDir)
-                    newProp = pOpening.contextDir;                      /* N0c1 */
-                else
-                    newProp = direction;                                /* N0c2 */
-            } else {
-            /* forget this and any brackets nested within this pair */
-            pLastIsoRun.limit = (short)openIdx;
-            return ON;                                                  /* N0d */
-        }
-        dirProps[pOpening.position] = newProp;
-        dirProps[position] = newProp;
-        /* Update nested N0c pairs that may be affected */
-        fixN0c(bd, openIdx, pOpening.position, newProp);
-        if (stable) {
-            pLastIsoRun.limit = (short)openIdx; /* forget any brackets nested within this pair */
-            /* remove lower located synonyms if any */
-            while (pLastIsoRun.limit > pLastIsoRun.start &&
-                   bd.openings[pLastIsoRun.limit - 1].position == pOpening.position)
-                pLastIsoRun.limit--;
-        } else {
-            int k;
-            pOpening.match = -position;
-            /* neutralize lower located synonyms if any */
-            k = openIdx - 1;
-            while (k >= pLastIsoRun.start &&
-                   bd.openings[k].position == pOpening.position)
-                bd.openings[k--].match = 0;
-            /* neutralize any unmatched opening between the current pair;
-               this will also neutralize higher located synonyms if any */
-            for (k = openIdx + 1; k < pLastIsoRun.limit; k++) {
-                qOpening =bd.openings[k];
-                if (qOpening.position >= position)
-                    break;
-                if (qOpening.match > 0)
-                    qOpening.match = 0;
-            }
-        }
-        return newProp;
-    }
-
-    /* handle strong characters, digits and candidates for closing brackets */
-    private void bracketProcessChar(BracketData bd, int position) {
-        IsoRun pLastIsoRun = bd.isoRuns[bd.isoRunLast];
-        byte dirProp, newProp;
-        byte level;
-        dirProp = dirProps[position];
-        if (dirProp == ON) {
-            char c, match;
-            int idx;
-            /* First see if it is a matching closing bracket. Hopefully, this is
-               more efficient than checking if it is a closing bracket at all */
-            c = text[position];
-            for (idx = pLastIsoRun.limit - 1; idx >= pLastIsoRun.start; idx--) {
-                if (bd.openings[idx].match != c)
-                    continue;
-                /* We have a match */
-                newProp = bracketProcessClosing(bd, idx, position);
-                if(newProp == ON) {         /* N0d */
-                    c = 0;          /* prevent handling as an opening */
-                    break;
-                }
-                pLastIsoRun.lastBase = ON;
-                pLastIsoRun.contextDir = newProp;
-                pLastIsoRun.contextPos = position;
-                level = levels[position];
-                if ((level & LEVEL_OVERRIDE) != 0) {    /* X4, X5 */
-                    short flag;
-                    int i;
-                    newProp = (byte)(level & 1);
-                    pLastIsoRun.lastStrong = newProp;
-                    flag = (short)DirPropFlag(newProp);
-                    for (i = pLastIsoRun.start; i < idx; i++)
-                        bd.openings[i].flags |= flag;
-                    /* matching brackets are not overridden by LRO/RLO */
-                    levels[position] &= ~LEVEL_OVERRIDE;
-                }
-                /* matching brackets are not overridden by LRO/RLO */
-                levels[bd.openings[idx].position] &= ~LEVEL_OVERRIDE;
-                return;
-            }
-            /* We get here only if the ON character is not a matching closing
-               bracket or it is a case of N0d */
-            /* Now see if it is an opening bracket */
-            if (c != 0) {
-                match = (char)UCharacter.getBidiPairedBracket(c); /* get the matching char */
-            } else {
-                match = 0;
-            }
-            if (match != c &&               /* has a matching char */
-                UCharacter.getIntPropertyValue(c, BIDI_PAIRED_BRACKET_TYPE) ==
-                    /* opening bracket */         BidiPairedBracketType.OPEN) {
-                /* special case: process synonyms
-                   create an opening entry for each synonym */
-                if (match == 0x232A) {      /* RIGHT-POINTING ANGLE BRACKET */
-                    bracketAddOpening(bd, (char)0x3009, position);
-                }
-                else if (match == 0x3009) { /* RIGHT ANGLE BRACKET */
-                    bracketAddOpening(bd, (char)0x232A, position);
-                }
-                bracketAddOpening(bd, match, position);
-            }
-        }
-        level = levels[position];
-        if ((level & LEVEL_OVERRIDE) != 0) {    /* X4, X5 */
-            newProp = (byte)(level & 1);
-            if (dirProp != S && dirProp != WS && dirProp != ON)
-                dirProps[position] = newProp;
-            pLastIsoRun.lastBase = newProp;
-            pLastIsoRun.lastStrong = newProp;
-            pLastIsoRun.contextDir = newProp;
-            pLastIsoRun.contextPos = position;
-        }
-        else if (dirProp <= R || dirProp == AL) {
-            newProp = DirFromStrong(dirProp);
-            pLastIsoRun.lastBase = dirProp;
-            pLastIsoRun.lastStrong = dirProp;
-            pLastIsoRun.contextDir = newProp;
-            pLastIsoRun.contextPos = position;
-        }
-        else if(dirProp == EN) {
-            pLastIsoRun.lastBase = EN;
-            if (pLastIsoRun.lastStrong == L) {
-                newProp = L;                    /* W7 */
-                if (!bd.isNumbersSpecial)
-                    dirProps[position] = ENL;
-                pLastIsoRun.contextDir = L;
-                pLastIsoRun.contextPos = position;
-            }
-            else {
-                newProp = R;                    /* N0 */
-                if (pLastIsoRun.lastStrong == AL)
-                    dirProps[position] = AN;    /* W2 */
-                else
-                    dirProps[position] = ENR;
-                pLastIsoRun.contextDir = R;
-                pLastIsoRun.contextPos = position;
-            }
-        }
-        else if (dirProp == AN) {
-            newProp = R;                        /* N0 */
-            pLastIsoRun.lastBase = AN;
-            pLastIsoRun.contextDir = R;
-            pLastIsoRun.contextPos = position;
-        }
-        else if (dirProp == NSM) {
-            /* if the last real char was ON, change NSM to ON so that it
-               will stay ON even if the last real char is a bracket which
-               may be changed to L or R */
-            newProp = pLastIsoRun.lastBase;
-            if (newProp == ON)
-                dirProps[position] = newProp;
-        }
-        else {
-            newProp = dirProp;
-            pLastIsoRun.lastBase = dirProp;
-        }
-        if (newProp <= R || newProp == AL) {
-            int i;
-            short flag = (short)DirPropFlag(DirFromStrong(newProp));
-            for (i = pLastIsoRun.start; i < pLastIsoRun.limit; i++)
-                if (position > bd.openings[i].position)
-                    bd.openings[i].flags |= flag;
-        }
-    }
-
-    /* perform (X1)..(X9) ------------------------------------------------------- */
-
-    /* determine if the text is mixed-directional or single-directional */
-    private byte directionFromFlags() {
-
-        /* if the text contains AN and neutrals, then some neutrals may become RTL */
-        if (!((flags & MASK_RTL) != 0 ||
-              ((flags & DirPropFlag(AN)) != 0 &&
-               (flags & MASK_POSSIBLE_N) != 0))) {
-            return LTR;
-        } else if ((flags & MASK_LTR) == 0) {
-            return RTL;
-        } else {
-            return MIXED;
-        }
-    }
-
-    /*
-     * Resolve the explicit levels as specified by explicit embedding codes.
-     * Recalculate the flags to have them reflect the real properties
-     * after taking the explicit embeddings into account.
-     *
-     * The BiDi algorithm is designed to result in the same behavior whether embedding
-     * levels are externally specified (from "styled text", supposedly the preferred
-     * method) or set by explicit embedding codes (LRx, RLx, PDF, FSI, PDI) in the plain text.
-     * That is why (X9) instructs to remove all not-isolate explicit codes (and BN).
-     * However, in a real implementation, the removal of these codes and their index
-     * positions in the plain text is undesirable since it would result in
-     * reallocated, reindexed text.
-     * Instead, this implementation leaves the codes in there and just ignores them
-     * in the subsequent processing.
-     * In order to get the same reordering behavior, positions with a BN or a not-isolate
-     * explicit embedding code just get the same level assigned as the last "real"
-     * character.
-     *
-     * Some implementations, not this one, then overwrite some of these
-     * directionality properties at "real" same-level-run boundaries by
-     * L or R codes so that the resolution of weak types can be performed on the
-     * entire paragraph at once instead of having to parse it once more and
-     * perform that resolution on same-level-runs.
-     * This limits the scope of the implicit rules in effectively
-     * the same way as the run limits.
-     *
-     * Instead, this implementation does not modify these codes, except for
-     * paired brackets whose properties (ON) may be replaced by L or R.
-     * On one hand, the paragraph has to be scanned for same-level-runs, but
-     * on the other hand, this saves another loop to reset these codes,
-     * or saves making and modifying a copy of dirProps[].
-     *
-     *
-     * Note that (Pn) and (Xn) changed significantly from version 4 of the BiDi algorithm.
-     *
-     *
-     * Handling the stack of explicit levels (Xn):
-     *
-     * With the BiDi stack of explicit levels, as pushed with each
-     * LRE, RLE, LRO, RLO, LRI, RLI and FSI and popped with each PDF and PDI,
-     * the explicit level must never exceed MAX_EXPLICIT_LEVEL.
-     *
-     * In order to have a correct push-pop semantics even in the case of overflows,
-     * overflow counters and a valid isolate counter are used as described in UAX#9
-     * section 3.3.2 "Explicit Levels and Directions".
-     *
-     * This implementation assumes that MAX_EXPLICIT_LEVEL is odd.
-     *
-     * Returns the direction
-     *
-     */
-    private byte resolveExplicitLevels() {
-        int i = 0;
-        byte dirProp;
-        byte level = GetParaLevelAt(0);
-        byte dirct;
-        isolateCount = 0;
-
-        /* determine if the text is mixed-directional or single-directional */
-        dirct = directionFromFlags();
-
-        /* we may not need to resolve any explicit levels */
-        if (dirct != MIXED) {
-            /* not mixed directionality: levels don't matter - trailingWSStart will be 0 */
-            return dirct;
-        }
-
-        if (reorderingMode > REORDER_LAST_LOGICAL_TO_VISUAL) {
-            /* inverse BiDi: mixed, but all characters are at the same embedding level */
-            /* set all levels to the paragraph level */
-            int paraIndex, start, limit;
-            for (paraIndex = 0; paraIndex < paraCount; paraIndex++) {
-                if (paraIndex == 0)
-                    start = 0;
-                else
-                    start = paras_limit[paraIndex - 1];
-                limit = paras_limit[paraIndex];
-                level = paras_level[paraIndex];
-                for (i = start; i < limit; i++)
-                    levels[i] =level;
-            }
-            return dirct;               /* no bracket matching for inverse BiDi */
-        }
-        if ((flags & (MASK_EXPLICIT | MASK_ISO)) == 0) {
-            /* no embeddings, set all levels to the paragraph level */
-            /* we still have to perform bracket matching */
-            int paraIndex, start, limit;
-            BracketData bracketData = new BracketData();
-            bracketInit(bracketData);
-            for (paraIndex = 0; paraIndex < paraCount; paraIndex++) {
-                if (paraIndex == 0)
-                    start = 0;
-                else
-                    start = paras_limit[paraIndex-1];
-                limit = paras_limit[paraIndex];
-                level = paras_level[paraIndex];
-                for (i = start; i < limit; i++) {
-                    levels[i] = level;
-                    dirProp = dirProps[i];
-                    if (dirProp == BN)
-                        continue;
-                    if (dirProp == B) {
-                        if ((i + 1) < length) {
-                            if (text[i] == CR && text[i + 1] == LF)
-                                continue;   /* skip CR when followed by LF */
-                            bracketProcessB(bracketData, level);
-                        }
-                        continue;
-                    }
-                    bracketProcessChar(bracketData, i);
-                }
-            }
-            return dirct;
-        }
-        /* continue to perform (Xn) */
-
-        /* (X1) level is set for all codes, embeddingLevel keeps track of the push/pop operations */
-        /* both variables may carry the LEVEL_OVERRIDE flag to indicate the override status */
-        byte embeddingLevel = level, newLevel;
-        byte previousLevel = level; /* previous level for regular (not CC) characters */
-        int lastCcPos = 0;          /* index of last effective LRx,RLx, PDx */
-
-        /* The following stack remembers the embedding level and the ISOLATE flag of level runs.
-           stackLast points to its current entry. */
-        short[] stack = new short[MAX_EXPLICIT_LEVEL + 2];  /* we never push anything >= MAX_EXPLICIT_LEVEL
-                                                               but we need one more entry as base */
-        int stackLast = 0;
-        int overflowIsolateCount = 0;
-        int overflowEmbeddingCount = 0;
-        int validIsolateCount = 0;
-        BracketData bracketData = new BracketData();
-        bracketInit(bracketData);
-        stack[0] = level;       /* initialize base entry to para level, no override, no isolate */
-
-        /* recalculate the flags */
-        flags = 0;
-
-        for (i = 0; i < length; i++) {
-            dirProp = dirProps[i];
-            switch (dirProp) {
-            case LRE:
-            case RLE:
-            case LRO:
-            case RLO:
-                /* (X2, X3, X4, X5) */
-                flags |= DirPropFlag(BN);
-                levels[i] = previousLevel;
-                if (dirProp == LRE || dirProp == LRO) {
-                    /* least greater even level */
-                    newLevel = (byte)((embeddingLevel+2) & ~(LEVEL_OVERRIDE | 1));
-                } else {
-                    /* least greater odd level */
-                    newLevel = (byte)((NoOverride(embeddingLevel) + 1) | 1);
-                }
-                if (newLevel <= MAX_EXPLICIT_LEVEL && overflowIsolateCount == 0 &&
-                                                      overflowEmbeddingCount == 0) {
-                    lastCcPos = i;
-                    embeddingLevel = newLevel;
-                    if (dirProp == LRO || dirProp == RLO)
-                        embeddingLevel |= LEVEL_OVERRIDE;
-                    stackLast++;
-                    stack[stackLast] = embeddingLevel;
-                    /* we don't need to set LEVEL_OVERRIDE off for LRE and RLE
-                       since this has already been done for newLevel which is
-                       the source for embeddingLevel.
-                     */
-                } else {
-                    if (overflowIsolateCount == 0)
-                        overflowEmbeddingCount++;
-                }
-                break;
-            case PDF:
-                /* (X7) */
-                flags |= DirPropFlag(BN);
-                levels[i] = previousLevel;
-                /* handle all the overflow cases first */
-                if (overflowIsolateCount > 0) {
-                    break;
-                }
-                if (overflowEmbeddingCount > 0) {
-                    overflowEmbeddingCount--;
-                    break;
-                }
-                if (stackLast > 0 && stack[stackLast] < ISOLATE) {   /* not an isolate entry */
-                    lastCcPos = i;
-                    stackLast--;
-                    embeddingLevel = (byte)stack[stackLast];
-                }
-                break;
-            case LRI:
-            case RLI:
-                flags |= DirPropFlag(ON) | DirPropFlagLR(embeddingLevel);
-                levels[i] = NoOverride(embeddingLevel);
-                if (NoOverride(embeddingLevel) != NoOverride(previousLevel)) {
-                    bracketProcessBoundary(bracketData, lastCcPos,
-                                           previousLevel, embeddingLevel);
-                    flags |= DirPropFlagMultiRuns;
-                }
-                previousLevel = embeddingLevel;
-                /* (X5a, X5b) */
-                if (dirProp == LRI)
-                    /* least greater even level */
-                    newLevel=(byte)((embeddingLevel+2)&~(LEVEL_OVERRIDE|1));
-                else
-                    /* least greater odd level */
-                    newLevel=(byte)((NoOverride(embeddingLevel)+1)|1);
-                if (newLevel <= MAX_EXPLICIT_LEVEL && overflowIsolateCount == 0
-                                                   && overflowEmbeddingCount == 0) {
-                    flags |= DirPropFlag(dirProp);
-                    lastCcPos = i;
-                    validIsolateCount++;
-                    if (validIsolateCount > isolateCount)
-                        isolateCount = validIsolateCount;
-                    embeddingLevel = newLevel;
-                    /* we can increment stackLast without checking because newLevel
-                       will exceed UBIDI_MAX_EXPLICIT_LEVEL before stackLast overflows */
-                    stackLast++;
-                    stack[stackLast] = (short)(embeddingLevel + ISOLATE);
-                    bracketProcessLRI_RLI(bracketData, embeddingLevel);
-                } else {
-                    /* make it WS so that it is handled by adjustWSLevels() */
-                    dirProps[i] = WS;
-                    overflowIsolateCount++;
-                }
-                break;
-            case PDI:
-                if (NoOverride(embeddingLevel) != NoOverride(previousLevel)) {
-                    bracketProcessBoundary(bracketData, lastCcPos,
-                                           previousLevel, embeddingLevel);
-                    flags |= DirPropFlagMultiRuns;
-                }
-                /* (X6a) */
-                if (overflowIsolateCount > 0) {
-                    overflowIsolateCount--;
-                    /* make it WS so that it is handled by adjustWSLevels() */
-                    dirProps[i] = WS;
-                }
-                else if (validIsolateCount > 0) {
-                    flags |= DirPropFlag(PDI);
-                    lastCcPos = i;
-                    overflowEmbeddingCount = 0;
-                    while (stack[stackLast] < ISOLATE)  /* pop embedding entries */
-                        stackLast--;                    /* until the last isolate entry */
-                    stackLast--;                        /* pop also the last isolate entry */
-                    validIsolateCount--;
-                    bracketProcessPDI(bracketData);
-                } else
-                    /* make it WS so that it is handled by adjustWSLevels() */
-                    dirProps[i] = WS;
-                embeddingLevel = (byte)(stack[stackLast] & ~ISOLATE);
-                flags |= DirPropFlag(ON) | DirPropFlagLR(embeddingLevel);
-                previousLevel = embeddingLevel;
-                levels[i] = NoOverride(embeddingLevel);
-                break;
-            case B:
-                flags |= DirPropFlag(B);
-                levels[i] = GetParaLevelAt(i);
-                if ((i + 1) < length) {
-                    if (text[i] == CR && text[i + 1] == LF)
-                        break;          /* skip CR when followed by LF */
-                    overflowEmbeddingCount = overflowIsolateCount = 0;
-                    validIsolateCount = 0;
-                    stackLast = 0;
-                    previousLevel = embeddingLevel = GetParaLevelAt(i + 1);
-                    stack[0] = embeddingLevel;   /* initialize base entry to para level, no override, no isolate */
-                    bracketProcessB(bracketData, embeddingLevel);
-                }
-                break;
-            case BN:
-                /* BN, LRE, RLE, and PDF are supposed to be removed (X9) */
-                /* they will get their levels set correctly in adjustWSLevels() */
-                levels[i] = previousLevel;
-                flags |= DirPropFlag(BN);
-                break;
-            default:
-                /* all other types are normal characters and get the "real" level */
-                if (NoOverride(embeddingLevel) != NoOverride(previousLevel)) {
-                    bracketProcessBoundary(bracketData, lastCcPos,
-                                           previousLevel, embeddingLevel);
-                    flags |= DirPropFlagMultiRuns;
-                    if ((embeddingLevel & LEVEL_OVERRIDE) != 0)
-                        flags |= DirPropFlagO(embeddingLevel);
-                    else
-                        flags |= DirPropFlagE(embeddingLevel);
-                }
-                previousLevel = embeddingLevel;
-                levels[i] = embeddingLevel;
-                bracketProcessChar(bracketData, i);
-                /* the dirProp may have been changed in bracketProcessChar() */
-                flags |= DirPropFlag(dirProps[i]);
-                break;
-            }
-        }
-        if ((flags & MASK_EMBEDDING) != 0) {
-            flags |= DirPropFlagLR(paraLevel);
-        }
-        if (orderParagraphsLTR && (flags & DirPropFlag(B)) != 0) {
-            flags |= DirPropFlag(L);
-        }
-        /* again, determine if the text is mixed-directional or single-directional */
-        dirct = directionFromFlags();
-
-        return dirct;
-    }
-
-    /*
-     * Use a pre-specified embedding levels array:
-     *
-     * Adjust the directional properties for overrides (->LEVEL_OVERRIDE),
-     * ignore all explicit codes (X9),
-     * and check all the preset levels.
-     *
-     * Recalculate the flags to have them reflect the real properties
-     * after taking the explicit embeddings into account.
-     */
-    private byte checkExplicitLevels() {
-        byte dirProp;
-        int i;
-        int isolateCount = 0;
-
-        this.flags = 0;     /* collect all directionalities in the text */
-        byte level;
-        this.isolateCount = 0;
-
-        for (i = 0; i < length; ++i) {
-            if (levels[i] == 0) {
-               levels[i] = paraLevel;
-            }
-
-            // for backward compatibility
-            if (MAX_EXPLICIT_LEVEL < (levels[i]&0x7f)) {
-                if ((levels[i] & LEVEL_OVERRIDE) != 0) {
-                    levels[i] =  (byte)(paraLevel|LEVEL_OVERRIDE);
-                } else {
-                    levels[i] = paraLevel;
-                }
-            }
-
-            level = levels[i];
-            dirProp = dirProps[i];
-            if (dirProp == LRI || dirProp == RLI) {
-                isolateCount++;
-                if (isolateCount > this.isolateCount)
-                    this.isolateCount = isolateCount;
-            }
-            else if (dirProp == PDI) {
-                isolateCount--;
-            } else if (dirProp == B) {
-                isolateCount = 0;
-            }
-            if ((level & LEVEL_OVERRIDE) != 0) {
-                /* keep the override flag in levels[i] but adjust the flags */
-                level &= ~LEVEL_OVERRIDE;     /* make the range check below simpler */
-                flags |= DirPropFlagO(level);
-            } else {
-                /* set the flags */
-                flags |= DirPropFlagE(level) | DirPropFlag(dirProp);
-            }
-            if ((level < GetParaLevelAt(i) &&
-                    !((0 == level) && (dirProp == B))) ||
-                    (MAX_EXPLICIT_LEVEL < level)) {
-                /* level out of bounds */
-                throw new IllegalArgumentException("level " + level +
-                                                   " out of bounds at " + i);
-            }
-        }
-        if ((flags & MASK_EMBEDDING) != 0) {
-            flags |= DirPropFlagLR(paraLevel);
-        }
-        /* determine if the text is mixed-directional or single-directional */
-        return directionFromFlags();
-    }
-
-    /*********************************************************************/
-    /* The Properties state machine table                                */
-    /*********************************************************************/
-    /*                                                                   */
-    /* All table cells are 8 bits:                                       */
-    /*      bits 0..4:  next state                                       */
-    /*      bits 5..7:  action to perform (if > 0)                       */
-    /*                                                                   */
-    /* Cells may be of format "n" where n represents the next state      */
-    /* (except for the rightmost column).                                */
-    /* Cells may also be of format "_(x,y)" where x represents an action */
-    /* to perform and y represents the next state.                       */
-    /*                                                                   */
-    /*********************************************************************/
-    /* Definitions and type for properties state tables                  */
-    /*********************************************************************/
-    private static final int IMPTABPROPS_COLUMNS = 16;
-    private static final int IMPTABPROPS_RES = IMPTABPROPS_COLUMNS - 1;
-    private static short GetStateProps(short cell) {
-        return (short)(cell & 0x1f);
-    }
-    private static short GetActionProps(short cell) {
-        return (short)(cell >> 5);
-    }
-
-    private static final short groupProp[] =          /* dirProp regrouped */
-    {
-        /*  L   R   EN  ES  ET  AN  CS  B   S   WS  ON  LRE LRO AL  RLE RLO PDF NSM BN  FSI LRI RLI PDI ENL ENR */
-            0,  1,  2,  7,  8,  3,  9,  6,  5,  4,  4,  10, 10, 12, 10, 10, 10, 11, 10, 4,  4,  4,  4,  13, 14
-    };
-    private static final short _L  = 0;
-    private static final short _R  = 1;
-    private static final short _EN = 2;
-    private static final short _AN = 3;
-    private static final short _ON = 4;
-    private static final short _S  = 5;
-    private static final short _B  = 6; /* reduced dirProp */
-
-    /*********************************************************************/
-    /*                                                                   */
-    /*      PROPERTIES  STATE  TABLE                                     */
-    /*                                                                   */
-    /* In table impTabProps,                                             */
-    /*      - the ON column regroups ON and WS, FSI, RLI, LRI and PDI    */
-    /*      - the BN column regroups BN, LRE, RLE, LRO, RLO, PDF         */
-    /*      - the Res column is the reduced property assigned to a run   */
-    /*                                                                   */
-    /* Action 1: process current run1, init new run1                     */
-    /*        2: init new run2                                           */
-    /*        3: process run1, process run2, init new run1               */
-    /*        4: process run1, set run1=run2, init new run2              */
-    /*                                                                   */
-    /* Notes:                                                            */
-    /*  1) This table is used in resolveImplicitLevels().                */
-    /*  2) This table triggers actions when there is a change in the Bidi*/
-    /*     property of incoming characters (action 1).                   */
-    /*  3) Most such property sequences are processed immediately (in    */
-    /*     fact, passed to processPropertySeq().                         */
-    /*  4) However, numbers are assembled as one sequence. This means    */
-    /*     that undefined situations (like CS following digits, until    */
-    /*     it is known if the next char will be a digit) are held until  */
-    /*     following chars define them.                                  */
-    /*     Example: digits followed by CS, then comes another CS or ON;  */
-    /*              the digits will be processed, then the CS assigned   */
-    /*              as the start of an ON sequence (action 3).           */
-    /*  5) There are cases where more than one sequence must be          */
-    /*     processed, for instance digits followed by CS followed by L:  */
-    /*     the digits must be processed as one sequence, and the CS      */
-    /*     must be processed as an ON sequence, all this before starting */
-    /*     assembling chars for the opening L sequence.                  */
-    /*                                                                   */
-    /*                                                                   */
-    private static final short impTabProps[][] =
-    {
-/*                        L,     R,    EN,    AN,    ON,     S,     B,    ES,    ET,    CS,    BN,   NSM,    AL,   ENL,   ENR,   Res */
-/* 0 Init        */ {     1,     2,     4,     5,     7,    15,    17,     7,     9,     7,     0,     7,     3,    18,    21,   _ON },
-/* 1 L           */ {     1,  32+2,  32+4,  32+5,  32+7, 32+15, 32+17,  32+7,  32+9,  32+7,     1,     1,  32+3, 32+18, 32+21,    _L },
-/* 2 R           */ {  32+1,     2,  32+4,  32+5,  32+7, 32+15, 32+17,  32+7,  32+9,  32+7,     2,     2,  32+3, 32+18, 32+21,    _R },
-/* 3 AL          */ {  32+1,  32+2,  32+6,  32+6,  32+8, 32+16, 32+17,  32+8,  32+8,  32+8,     3,     3,     3, 32+18, 32+21,    _R },
-/* 4 EN          */ {  32+1,  32+2,     4,  32+5,  32+7, 32+15, 32+17, 64+10,    11, 64+10,     4,     4,  32+3,    18,    21,   _EN },
-/* 5 AN          */ {  32+1,  32+2,  32+4,     5,  32+7, 32+15, 32+17,  32+7,  32+9, 64+12,     5,     5,  32+3, 32+18, 32+21,   _AN },
-/* 6 AL:EN/AN    */ {  32+1,  32+2,     6,     6,  32+8, 32+16, 32+17,  32+8,  32+8, 64+13,     6,     6,  32+3,    18,    21,   _AN },
-/* 7 ON          */ {  32+1,  32+2,  32+4,  32+5,     7, 32+15, 32+17,     7, 64+14,     7,     7,     7,  32+3, 32+18, 32+21,   _ON },
-/* 8 AL:ON       */ {  32+1,  32+2,  32+6,  32+6,     8, 32+16, 32+17,     8,     8,     8,     8,     8,  32+3, 32+18, 32+21,   _ON },
-/* 9 ET          */ {  32+1,  32+2,     4,  32+5,     7, 32+15, 32+17,     7,     9,     7,     9,     9,  32+3,    18,    21,   _ON },
-/*10 EN+ES/CS    */ {  96+1,  96+2,     4,  96+5, 128+7, 96+15, 96+17, 128+7,128+14, 128+7,    10, 128+7,  96+3,    18,    21,   _EN },
-/*11 EN+ET       */ {  32+1,  32+2,     4,  32+5,  32+7, 32+15, 32+17,  32+7,    11,  32+7,    11,    11,  32+3,    18,    21,   _EN },
-/*12 AN+CS       */ {  96+1,  96+2,  96+4,     5, 128+7, 96+15, 96+17, 128+7,128+14, 128+7,    12, 128+7,  96+3, 96+18, 96+21,   _AN },
-/*13 AL:EN/AN+CS */ {  96+1,  96+2,     6,     6, 128+8, 96+16, 96+17, 128+8, 128+8, 128+8,    13, 128+8,  96+3,    18,    21,   _AN },
-/*14 ON+ET       */ {  32+1,  32+2, 128+4,  32+5,     7, 32+15, 32+17,     7,    14,     7,    14,    14,  32+3,128+18,128+21,   _ON },
-/*15 S           */ {  32+1,  32+2,  32+4,  32+5,  32+7,    15, 32+17,  32+7,  32+9,  32+7,    15,  32+7,  32+3, 32+18, 32+21,    _S },
-/*16 AL:S        */ {  32+1,  32+2,  32+6,  32+6,  32+8,    16, 32+17,  32+8,  32+8,  32+8,    16,  32+8,  32+3, 32+18, 32+21,    _S },
-/*17 B           */ {  32+1,  32+2,  32+4,  32+5,  32+7, 32+15,    17,  32+7,  32+9,  32+7,    17,  32+7,  32+3, 32+18, 32+21,    _B },
-/*18 ENL         */ {  32+1,  32+2,    18,  32+5,  32+7, 32+15, 32+17, 64+19,    20, 64+19,    18,    18,  32+3,    18,    21,    _L },
-/*19 ENL+ES/CS   */ {  96+1,  96+2,    18,  96+5, 128+7, 96+15, 96+17, 128+7,128+14, 128+7,    19, 128+7,  96+3,    18,    21,    _L },
-/*20 ENL+ET      */ {  32+1,  32+2,    18,  32+5,  32+7, 32+15, 32+17,  32+7,    20,  32+7,    20,    20,  32+3,    18,    21,    _L },
-/*21 ENR         */ {  32+1,  32+2,    21,  32+5,  32+7, 32+15, 32+17, 64+22,    23, 64+22,    21,    21,  32+3,    18,    21,   _AN },
-/*22 ENR+ES/CS   */ {  96+1,  96+2,    21,  96+5, 128+7, 96+15, 96+17, 128+7,128+14, 128+7,    22, 128+7,  96+3,    18,    21,   _AN },
-/*23 ENR+ET      */ {  32+1,  32+2,    21,  32+5,  32+7, 32+15, 32+17,  32+7,    23,  32+7,    23,    23,  32+3,    18,    21,   _AN }
-    };
-
-    /*********************************************************************/
-    /* The levels state machine tables                                   */
-    /*********************************************************************/
-    /*                                                                   */
-    /* All table cells are 8 bits:                                       */
-    /*      bits 0..3:  next state                                       */
-    /*      bits 4..7:  action to perform (if > 0)                       */
-    /*                                                                   */
-    /* Cells may be of format "n" where n represents the next state      */
-    /* (except for the rightmost column).                                */
-    /* Cells may also be of format "_(x,y)" where x represents an action */
-    /* to perform and y represents the next state.                       */
-    /*                                                                   */
-    /* This format limits each table to 16 states each and to 15 actions.*/
-    /*                                                                   */
-    /*********************************************************************/
-    /* Definitions and type for levels state tables                      */
-    /*********************************************************************/
-    private static final int IMPTABLEVELS_COLUMNS = _B + 2;
-    private static final int IMPTABLEVELS_RES = IMPTABLEVELS_COLUMNS - 1;
-    private static short GetState(byte cell) { return (short)(cell & 0x0f); }
-    private static short GetAction(byte cell) { return (short)(cell >> 4); }
-
-    private static class ImpTabPair {
-        byte[][][] imptab;
-        short[][] impact;
-
-        ImpTabPair(byte[][] table1, byte[][] table2,
-                   short[] act1, short[] act2) {
-            imptab = new byte[][][] {table1, table2};
-            impact = new short[][] {act1, act2};
-        }
-    }
-
-    /*********************************************************************/
-    /*                                                                   */
-    /*      LEVELS  STATE  TABLES                                        */
-    /*                                                                   */
-    /* In all levels state tables,                                       */
-    /*      - state 0 is the initial state                               */
-    /*      - the Res column is the increment to add to the text level   */
-    /*        for this property sequence.                                */
-    /*                                                                   */
-    /* The impact arrays for each table of a pair map the local action   */
-    /* numbers of the table to the total list of actions. For instance,  */
-    /* action 2 in a given table corresponds to the action number which  */
-    /* appears in entry [2] of the impact array for that table.          */
-    /* The first entry of all impact arrays must be 0.                   */
-    /*                                                                   */
-    /* Action 1: init conditional sequence                               */
-    /*        2: prepend conditional sequence to current sequence        */
-    /*        3: set ON sequence to new level - 1                        */
-    /*        4: init EN/AN/ON sequence                                  */
-    /*        5: fix EN/AN/ON sequence followed by R                     */
-    /*        6: set previous level sequence to level 2                  */
-    /*                                                                   */
-    /* Notes:                                                            */
-    /*  1) These tables are used in processPropertySeq(). The input      */
-    /*     is property sequences as determined by resolveImplicitLevels. */
-    /*  2) Most such property sequences are processed immediately        */
-    /*     (levels are assigned).                                        */
-    /*  3) However, some sequences cannot be assigned a final level till */
-    /*     one or more following sequences are received. For instance,   */
-    /*     ON following an R sequence within an even-level paragraph.    */
-    /*     If the following sequence is R, the ON sequence will be       */
-    /*     assigned basic run level+1, and so will the R sequence.       */
-    /*  4) S is generally handled like ON, since its level will be fixed */
-    /*     to paragraph level in adjustWSLevels().                       */
-    /*                                                                   */
-
-    private static final byte impTabL_DEFAULT[][] = /* Even paragraph level */
-        /*  In this table, conditional sequences receive the lower possible level
-            until proven otherwise.
-        */
-    {
-        /*                         L,     R,    EN,    AN,    ON,     S,     B, Res */
-        /* 0 : init       */ {     0,     1,     0,     2,     0,     0,     0,  0 },
-        /* 1 : R          */ {     0,     1,     3,     3,  0x14,  0x14,     0,  1 },
-        /* 2 : AN         */ {     0,     1,     0,     2,  0x15,  0x15,     0,  2 },
-        /* 3 : R+EN/AN    */ {     0,     1,     3,     3,  0x14,  0x14,     0,  2 },
-        /* 4 : R+ON       */ {     0,  0x21,  0x33,  0x33,     4,     4,     0,  0 },
-        /* 5 : AN+ON      */ {     0,  0x21,     0,  0x32,     5,     5,     0,  0 }
-    };
-
-    private static final byte impTabR_DEFAULT[][] = /* Odd  paragraph level */
-        /*  In this table, conditional sequences receive the lower possible level
-            until proven otherwise.
-        */
-    {
-        /*                         L,     R,    EN,    AN,    ON,     S,     B, Res */
-        /* 0 : init       */ {     1,     0,     2,     2,     0,     0,     0,  0 },
-        /* 1 : L          */ {     1,     0,     1,     3,  0x14,  0x14,     0,  1 },
-        /* 2 : EN/AN      */ {     1,     0,     2,     2,     0,     0,     0,  1 },
-        /* 3 : L+AN       */ {     1,     0,     1,     3,     5,     5,     0,  1 },
-        /* 4 : L+ON       */ {  0x21,     0,  0x21,     3,     4,     4,     0,  0 },
-        /* 5 : L+AN+ON    */ {     1,     0,     1,     3,     5,     5,     0,  0 }
-    };
-
-    private static final short[] impAct0 = {0,1,2,3,4};
-
-    private static final ImpTabPair impTab_DEFAULT = new ImpTabPair(
-            impTabL_DEFAULT, impTabR_DEFAULT, impAct0, impAct0);
-
-    private static final byte impTabL_NUMBERS_SPECIAL[][] = { /* Even paragraph level */
-        /* In this table, conditional sequences receive the lower possible
-           level until proven otherwise.
-        */
-        /*                         L,     R,    EN,    AN,    ON,     S,     B, Res */
-        /* 0 : init       */ {     0,     2,  0x11,  0x11,     0,     0,     0,  0 },
-        /* 1 : L+EN/AN    */ {     0,  0x42,     1,     1,     0,     0,     0,  0 },
-        /* 2 : R          */ {     0,     2,     4,     4,  0x13,  0x13,     0,  1 },
-        /* 3 : R+ON       */ {     0,  0x22,  0x34,  0x34,     3,     3,     0,  0 },
-        /* 4 : R+EN/AN    */ {     0,     2,     4,     4,  0x13,  0x13,     0,  2 }
-    };
-    private static final ImpTabPair impTab_NUMBERS_SPECIAL = new ImpTabPair(
-            impTabL_NUMBERS_SPECIAL, impTabR_DEFAULT, impAct0, impAct0);
-
-    private static final byte impTabL_GROUP_NUMBERS_WITH_R[][] = {
-        /* In this table, EN/AN+ON sequences receive levels as if associated with R
-           until proven that there is L or sor/eor on both sides. AN is handled like EN.
-        */
-        /*                         L,     R,    EN,    AN,    ON,     S,     B, Res */
-        /* 0 init         */ {     0,     3,  0x11,  0x11,     0,     0,     0,  0 },
-        /* 1 EN/AN        */ {  0x20,     3,     1,     1,     2,  0x20,  0x20,  2 },
-        /* 2 EN/AN+ON     */ {  0x20,     3,     1,     1,     2,  0x20,  0x20,  1 },
-        /* 3 R            */ {     0,     3,     5,     5,  0x14,     0,     0,  1 },
-        /* 4 R+ON         */ {  0x20,     3,     5,     5,     4,  0x20,  0x20,  1 },
-        /* 5 R+EN/AN      */ {     0,     3,     5,     5,  0x14,     0,     0,  2 }
-    };
-    private static final byte impTabR_GROUP_NUMBERS_WITH_R[][] = {
-        /*  In this table, EN/AN+ON sequences receive levels as if associated with R
-            until proven that there is L on both sides. AN is handled like EN.
-        */
-        /*                         L,     R,    EN,    AN,    ON,     S,     B, Res */
-        /* 0 init         */ {     2,     0,     1,     1,     0,     0,     0,  0 },
-        /* 1 EN/AN        */ {     2,     0,     1,     1,     0,     0,     0,  1 },
-        /* 2 L            */ {     2,     0,  0x14,  0x14,  0x13,     0,     0,  1 },
-        /* 3 L+ON         */ {  0x22,     0,     4,     4,     3,     0,     0,  0 },
-        /* 4 L+EN/AN      */ {  0x22,     0,     4,     4,     3,     0,     0,  1 }
-    };
-    private static final ImpTabPair impTab_GROUP_NUMBERS_WITH_R = new
-            ImpTabPair(impTabL_GROUP_NUMBERS_WITH_R,
-                       impTabR_GROUP_NUMBERS_WITH_R, impAct0, impAct0);
-
-    private static final byte impTabL_INVERSE_NUMBERS_AS_L[][] = {
-        /* This table is identical to the Default LTR table except that EN and AN
-           are handled like L.
-        */
-        /*                         L,     R,    EN,    AN,    ON,     S,     B, Res */
-        /* 0 : init       */ {     0,     1,     0,     0,     0,     0,     0,  0 },
-        /* 1 : R          */ {     0,     1,     0,     0,  0x14,  0x14,     0,  1 },
-        /* 2 : AN         */ {     0,     1,     0,     0,  0x15,  0x15,     0,  2 },
-        /* 3 : R+EN/AN    */ {     0,     1,     0,     0,  0x14,  0x14,     0,  2 },
-        /* 4 : R+ON       */ {  0x20,     1,  0x20,  0x20,     4,     4,  0x20,  1 },
-        /* 5 : AN+ON      */ {  0x20,     1,  0x20,  0x20,     5,     5,  0x20,  1 }
-    };
-    private static final byte impTabR_INVERSE_NUMBERS_AS_L[][] = {
-        /* This table is identical to the Default RTL table except that EN and AN
-           are handled like L.
-        */
-        /*                         L,     R,    EN,    AN,    ON,     S,     B, Res */
-        /* 0 : init       */ {     1,     0,     1,     1,     0,     0,     0,  0 },
-        /* 1 : L          */ {     1,     0,     1,     1,  0x14,  0x14,     0,  1 },
-        /* 2 : EN/AN      */ {     1,     0,     1,     1,     0,     0,     0,  1 },
-        /* 3 : L+AN       */ {     1,     0,     1,     1,     5,     5,     0,  1 },
-        /* 4 : L+ON       */ {  0x21,     0,  0x21,  0x21,     4,     4,     0,  0 },
-        /* 5 : L+AN+ON    */ {     1,     0,     1,     1,     5,     5,     0,  0 }
-    };
-    private static final ImpTabPair impTab_INVERSE_NUMBERS_AS_L = new ImpTabPair
-            (impTabL_INVERSE_NUMBERS_AS_L, impTabR_INVERSE_NUMBERS_AS_L,
-             impAct0, impAct0);
-
-    private static final byte impTabR_INVERSE_LIKE_DIRECT[][] = {  /* Odd  paragraph level */
-        /*  In this table, conditional sequences receive the lower possible level
-            until proven otherwise.
-        */
-        /*                         L,     R,    EN,    AN,    ON,     S,     B, Res */
-        /* 0 : init       */ {     1,     0,     2,     2,     0,     0,     0,  0 },
-        /* 1 : L          */ {     1,     0,     1,     2,  0x13,  0x13,     0,  1 },
-        /* 2 : EN/AN      */ {     1,     0,     2,     2,     0,     0,     0,  1 },
-        /* 3 : L+ON       */ {  0x21,  0x30,     6,     4,     3,     3,  0x30,  0 },
-        /* 4 : L+ON+AN    */ {  0x21,  0x30,     6,     4,     5,     5,  0x30,  3 },
-        /* 5 : L+AN+ON    */ {  0x21,  0x30,     6,     4,     5,     5,  0x30,  2 },
-        /* 6 : L+ON+EN    */ {  0x21,  0x30,     6,     4,     3,     3,  0x30,  1 }
-    };
-    private static final short[] impAct1 = {0,1,13,14};
-    private static final ImpTabPair impTab_INVERSE_LIKE_DIRECT = new ImpTabPair(
-            impTabL_DEFAULT, impTabR_INVERSE_LIKE_DIRECT, impAct0, impAct1);
-
-    private static final byte impTabL_INVERSE_LIKE_DIRECT_WITH_MARKS[][] = {
-        /* The case handled in this table is (visually):  R EN L
-         */
-        /*                         L,     R,    EN,    AN,    ON,     S,     B, Res */
-        /* 0 : init       */ {     0,  0x63,     0,     1,     0,     0,     0,  0 },
-        /* 1 : L+AN       */ {     0,  0x63,     0,     1,  0x12,  0x30,     0,  4 },
-        /* 2 : L+AN+ON    */ {  0x20,  0x63,  0x20,     1,     2,  0x30,  0x20,  3 },
-        /* 3 : R          */ {     0,  0x63,  0x55,  0x56,  0x14,  0x30,     0,  3 },
-        /* 4 : R+ON       */ {  0x30,  0x43,  0x55,  0x56,     4,  0x30,  0x30,  3 },
-        /* 5 : R+EN       */ {  0x30,  0x43,     5,  0x56,  0x14,  0x30,  0x30,  4 },
-        /* 6 : R+AN       */ {  0x30,  0x43,  0x55,     6,  0x14,  0x30,  0x30,  4 }
-    };
-    private static final byte impTabR_INVERSE_LIKE_DIRECT_WITH_MARKS[][] = {
-        /* The cases handled in this table are (visually):  R EN L
-                                                            R L AN L
-        */
-        /*                         L,     R,    EN,    AN,    ON,     S,     B, Res */
-        /* 0 : init       */ {  0x13,     0,     1,     1,     0,     0,     0,  0 },
-        /* 1 : R+EN/AN    */ {  0x23,     0,     1,     1,     2,  0x40,     0,  1 },
-        /* 2 : R+EN/AN+ON */ {  0x23,     0,     1,     1,     2,  0x40,     0,  0 },
-        /* 3 : L          */ {     3,     0,     3,  0x36,  0x14,  0x40,     0,  1 },
-        /* 4 : L+ON       */ {  0x53,  0x40,     5,  0x36,     4,  0x40,  0x40,  0 },
-        /* 5 : L+ON+EN    */ {  0x53,  0x40,     5,  0x36,     4,  0x40,  0x40,  1 },
-        /* 6 : L+AN       */ {  0x53,  0x40,     6,     6,     4,  0x40,  0x40,  3 }
-    };
-    private static final short[] impAct2 = {0,1,2,5,6,7,8};
-    private static final short[] impAct3 = {0,1,9,10,11,12};
-    private static final ImpTabPair impTab_INVERSE_LIKE_DIRECT_WITH_MARKS =
-            new ImpTabPair(impTabL_INVERSE_LIKE_DIRECT_WITH_MARKS,
-                           impTabR_INVERSE_LIKE_DIRECT_WITH_MARKS, impAct2, impAct3);
-
-    private static final ImpTabPair impTab_INVERSE_FOR_NUMBERS_SPECIAL = new ImpTabPair(
-            impTabL_NUMBERS_SPECIAL, impTabR_INVERSE_LIKE_DIRECT, impAct0, impAct1);
-
-    private static final byte impTabL_INVERSE_FOR_NUMBERS_SPECIAL_WITH_MARKS[][] = {
-        /*  The case handled in this table is (visually):  R EN L
-        */
-        /*                         L,     R,    EN,    AN,    ON,     S,     B, Res */
-        /* 0 : init       */ {     0,  0x62,     1,     1,     0,     0,     0,  0 },
-        /* 1 : L+EN/AN    */ {     0,  0x62,     1,     1,     0,  0x30,     0,  4 },
-        /* 2 : R          */ {     0,  0x62,  0x54,  0x54,  0x13,  0x30,     0,  3 },
-        /* 3 : R+ON       */ {  0x30,  0x42,  0x54,  0x54,     3,  0x30,  0x30,  3 },
-        /* 4 : R+EN/AN    */ {  0x30,  0x42,     4,     4,  0x13,  0x30,  0x30,  4 }
-    };
-    private static final ImpTabPair impTab_INVERSE_FOR_NUMBERS_SPECIAL_WITH_MARKS = new
-            ImpTabPair(impTabL_INVERSE_FOR_NUMBERS_SPECIAL_WITH_MARKS,
-                       impTabR_INVERSE_LIKE_DIRECT_WITH_MARKS, impAct2, impAct3);
-
-    private static class LevState {
-        byte[][] impTab;                /* level table pointer          */
-        short[] impAct;                 /* action map array             */
-        int startON;                    /* start of ON sequence         */
-        int startL2EN;                  /* start of level 2 sequence    */
-        int lastStrongRTL;              /* index of last found R or AL  */
-        int runStart;                   /* start position of the run    */
-        short state;                    /* current state                */
-        byte runLevel;                  /* run level before implicit solving */
-    }
-
-    /*------------------------------------------------------------------------*/
-
-    static final int FIRSTALLOC = 10;
-    /*
-     *  param pos:     position where to insert
-     *  param flag:    one of LRM_BEFORE, LRM_AFTER, RLM_BEFORE, RLM_AFTER
-     */
-    private void addPoint(int pos, int flag)
-    {
-        Point point = new Point();
-
-        int len = insertPoints.points.length;
-        if (len == 0) {
-            insertPoints.points = new Point[FIRSTALLOC];
-            len = FIRSTALLOC;
-        }
-        if (insertPoints.size >= len) { /* no room for new point */
-            Point[] savePoints = insertPoints.points;
-            insertPoints.points = new Point[len * 2];
-            System.arraycopy(savePoints, 0, insertPoints.points, 0, len);
-        }
-        point.pos = pos;
-        point.flag = flag;
-        insertPoints.points[insertPoints.size] = point;
-        insertPoints.size++;
-    }
-
-    private void setLevelsOutsideIsolates(int start, int limit, byte level)
-    {
-        byte dirProp;
-        int  isolateCount = 0, k;
-        for (k = start; k < limit; k++) {
-            dirProp = dirProps[k];
-            if (dirProp == PDI)
-                isolateCount--;
-            if (isolateCount == 0) {
-                levels[k] = level;
-            }
-            if (dirProp == LRI || dirProp == RLI)
-                isolateCount++;
-        }
-    }
-
-    /* perform rules (Wn), (Nn), and (In) on a run of the text ------------------ */
-
-    /*
-     * This implementation of the (Wn) rules applies all rules in one pass.
-     * In order to do so, it needs a look-ahead of typically 1 character
-     * (except for W5: sequences of ET) and keeps track of changes
-     * in a rule Wp that affect a later Wq (p= 0) {
-                    addPoint(levState.startL2EN, LRM_BEFORE);
-                }
-                levState.startL2EN = -1;  /* not within previous if since could also be -2 */
-                /* check if we had any relevant EN/AN after R/AL */
-                if ((insertPoints.points.length == 0) ||
-                        (insertPoints.size <= insertPoints.confirmed)) {
-                    /* nothing, just clean up */
-                    levState.lastStrongRTL = -1;
-                    /* check if we have a pending conditional segment */
-                    level = impTab[oldStateSeq][IMPTABLEVELS_RES];
-                    if ((level & 1) != 0 && levState.startON > 0) { /* after ON */
-                        start = levState.startON;   /* reset to basic run level */
-                    }
-                    if (_prop == _S) {              /* add LRM before S */
-                        addPoint(start0, LRM_BEFORE);
-                        insertPoints.confirmed = insertPoints.size;
-                    }
-                    break;
-                }
-                /* reset previous RTL cont to level for LTR text */
-                for (k = levState.lastStrongRTL + 1; k < start0; k++) {
-                    /* reset odd level, leave runLevel+2 as is */
-                    levels[k] = (byte)((levels[k] - 2) & ~1);
-                }
-                /* mark insert points as confirmed */
-                insertPoints.confirmed = insertPoints.size;
-                levState.lastStrongRTL = -1;
-                if (_prop == _S) {           /* add LRM before S */
-                    addPoint(start0, LRM_BEFORE);
-                    insertPoints.confirmed = insertPoints.size;
-                }
-                break;
-
-            case 6:                     /* R/AL after possible relevant EN/AN */
-                /* just clean up */
-                if (insertPoints.points.length > 0)
-                    /* remove all non confirmed insert points */
-                    insertPoints.size = insertPoints.confirmed;
-                levState.startON = -1;
-                levState.startL2EN = -1;
-                levState.lastStrongRTL = limit - 1;
-                break;
-
-            case 7:                     /* EN/AN after R/AL + possible cont */
-                /* check for real AN */
-
-                if ((_prop == _AN) && (dirProps[start0] == AN) &&
-                (reorderingMode != REORDER_INVERSE_FOR_NUMBERS_SPECIAL))
-                {
-                    /* real AN */
-                    if (levState.startL2EN == -1) { /* if no relevant EN already found */
-                        /* just note the rightmost digit as a strong RTL */
-                        levState.lastStrongRTL = limit - 1;
-                        break;
-                    }
-                    if (levState.startL2EN >= 0)  { /* after EN, no AN */
-                        addPoint(levState.startL2EN, LRM_BEFORE);
-                        levState.startL2EN = -2;
-                    }
-                    /* note AN */
-                    addPoint(start0, LRM_BEFORE);
-                    break;
-                }
-                /* if first EN/AN after R/AL */
-                if (levState.startL2EN == -1) {
-                    levState.startL2EN = start0;
-                }
-                break;
-
-            case 8:                     /* note location of latest R/AL */
-                levState.lastStrongRTL = limit - 1;
-                levState.startON = -1;
-                break;
-
-            case 9:                     /* L after R+ON/EN/AN */
-                /* include possible adjacent number on the left */
-                for (k = start0-1; k >= 0 && ((levels[k] & 1) == 0); k--) {
-                }
-                if (k >= 0) {
-                    addPoint(k, RLM_BEFORE);    /* add RLM before */
-                    insertPoints.confirmed = insertPoints.size; /* confirm it */
-                }
-                levState.startON = start0;
-                break;
-
-            case 10:                    /* AN after L */
-                /* AN numbers between L text on both sides may be trouble. */
-                /* tentatively bracket with LRMs; will be confirmed if followed by L */
-                addPoint(start0, LRM_BEFORE);   /* add LRM before */
-                addPoint(start0, LRM_AFTER);    /* add LRM after  */
-                break;
-
-            case 11:                    /* R after L+ON/EN/AN */
-                /* false alert, infirm LRMs around previous AN */
-                insertPoints.size=insertPoints.confirmed;
-                if (_prop == _S) {          /* add RLM before S */
-                    addPoint(start0, RLM_BEFORE);
-                    insertPoints.confirmed = insertPoints.size;
-                }
-                break;
-
-            case 12:                    /* L after L+ON/AN */
-                level = (byte)(levState.runLevel + addLevel);
-                for (k=levState.startON; k < start0; k++) {
-                    if (levels[k] < level) {
-                        levels[k] = level;
-                    }
-                }
-                insertPoints.confirmed = insertPoints.size;   /* confirm inserts */
-                levState.startON = start0;
-                break;
-
-            case 13:                    /* L after L+ON+EN/AN/ON */
-                level = levState.runLevel;
-                for (k = start0-1; k >= levState.startON; k--) {
-                    if (levels[k] == level+3) {
-                        while (levels[k] == level+3) {
-                            levels[k--] -= 2;
-                        }
-                        while (levels[k] == level) {
-                            k--;
-                        }
-                    }
-                    if (levels[k] == level+2) {
-                        levels[k] = level;
-                        continue;
-                    }
-                    levels[k] = (byte)(level+1);
-                }
-                break;
-
-            case 14:                    /* R after L+ON+EN/AN/ON */
-                level = (byte)(levState.runLevel+1);
-                for (k = start0-1; k >= levState.startON; k--) {
-                    if (levels[k] > level) {
-                        levels[k] -= 2;
-                    }
-                }
-                break;
-
-            default:                        /* we should never get here */
-                throw new IllegalStateException("Internal ICU error in processPropertySeq");
-            }
-        }
-        if ((addLevel) != 0 || (start < start0)) {
-            level = (byte)(levState.runLevel + addLevel);
-            if (start >= levState.runStart) {
-                for (k = start; k < limit; k++) {
-                    levels[k] = level;
-                }
-            } else {
-                setLevelsOutsideIsolates(start, limit, level);
-            }
-        }
-    }
-
-    private void resolveImplicitLevels(int start, int limit, short sor, short eor)
-    {
-        byte dirProp;
-        LevState levState = new LevState();
-        int i, start1, start2;
-        short oldStateImp, stateImp, actionImp;
-        short gprop, resProp, cell;
-        boolean inverseRTL;
-        short nextStrongProp = R;
-        int nextStrongPos = -1;
-
-        /* check for RTL inverse Bidi mode */
-        /* FOOD FOR THOUGHT: in case of RTL inverse Bidi, it would make sense to
-         * loop on the text characters from end to start.
-         * This would need a different properties state table (at least different
-         * actions) and different levels state tables (maybe very similar to the
-         * LTR corresponding ones.
-         */
-        inverseRTL=((start0) &&
-                    (reorderingMode == REORDER_INVERSE_LIKE_DIRECT  ||
-                     reorderingMode == REORDER_INVERSE_FOR_NUMBERS_SPECIAL));
-        /* initialize for property and levels state table */
-        levState.startL2EN = -1;        /* used for INVERSE_LIKE_DIRECT_WITH_MARKS */
-        levState.lastStrongRTL = -1;    /* used for INVERSE_LIKE_DIRECT_WITH_MARKS */
-        levState.runStart = start;
-        levState.runLevel = levels[start];
-        levState.impTab = impTabPair.imptab[levState.runLevel & 1];
-        levState.impAct = impTabPair.impact[levState.runLevel & 1];
-
-        /* The isolates[] entries contain enough information to
-           resume the bidi algorithm in the same state as it was
-           when it was interrupted by an isolate sequence. */
-        if (dirProps[start] == PDI) {
-            levState.startON = isolates[isolateCount].startON;
-            start1 = isolates[isolateCount].start1;
-            stateImp = isolates[isolateCount].stateImp;
-            levState.state = isolates[isolateCount].state;
-            isolateCount--;
-        } else {
-            levState.startON = -1;
-            start1 = start;
-            if (dirProps[start] == NSM)
-              stateImp = (short)(1 + sor);
-            else
-                stateImp = 0;
-            levState.state = 0;
-            processPropertySeq(levState, sor, start, start);
-        }
-        start2 = start;                 /* to make the Java compiler happy */
-
-        for (i = start; i <= limit; i++) {
-            if (i >= limit) {
-                int k;
-                for (k = limit - 1;
-                     k > start &&
-                         (DirPropFlag(dirProps[k]) & MASK_BN_EXPLICIT) != 0;
-                     k--);
-                dirProp = dirProps[k];
-                if (dirProp == LRI || dirProp == RLI)
-                    break;  /* no forced closing for sequence ending with LRI/RLI */
-                gprop = eor;
-            } else {
-                byte prop, prop1;
-                prop = dirProps[i];
-                if (prop == B)
-                    isolateCount = -1;  /* current isolates stack entry == none */
-                if (inverseRTL) {
-                    if (prop == AL) {
-                        /* AL before EN does not make it AN */
-                        prop = R;
-                    } else if (prop == EN) {
-                        if (nextStrongPos <= i) {
-                            /* look for next strong char (L/R/AL) */
-                            int j;
-                            nextStrongProp = R;     /* set default */
-                            nextStrongPos = limit;
-                            for (j = i+1; j < limit; j++) {
-                                prop1 = dirProps[j];
-                                if (prop1 == L || prop1 == R || prop1 == AL) {
-                                    nextStrongProp = prop1;
-                                    nextStrongPos = j;
-                                    break;
-                                }
-                            }
-                        }
-                        if (nextStrongProp == AL) {
-                            prop = AN;
-                        }
-                    }
-                }
-                gprop = groupProp[prop];
-            }
-            oldStateImp = stateImp;
-            cell = impTabProps[oldStateImp][gprop];
-            stateImp = GetStateProps(cell);     /* isolate the new state */
-            actionImp = GetActionProps(cell);   /* isolate the action */
-            if ((i == limit) && (actionImp == 0)) {
-                /* there is an unprocessed sequence if its property == eor   */
-                actionImp = 1;                  /* process the last sequence */
-            }
-            if (actionImp != 0) {
-                resProp = impTabProps[oldStateImp][IMPTABPROPS_RES];
-                switch (actionImp) {
-                case 1:             /* process current seq1, init new seq1 */
-                    processPropertySeq(levState, resProp, start1, i);
-                    start1 = i;
-                    break;
-                case 2:             /* init new seq2 */
-                    start2 = i;
-                    break;
-                case 3:             /* process seq1, process seq2, init new seq1 */
-                    processPropertySeq(levState, resProp, start1, start2);
-                    processPropertySeq(levState, _ON, start2, i);
-                    start1 = i;
-                    break;
-                case 4:             /* process seq1, set seq1=seq2, init new seq2 */
-                    processPropertySeq(levState, resProp, start1, start2);
-                    start1 = start2;
-                    start2 = i;
-                    break;
-                default:            /* we should never get here */
-                    throw new IllegalStateException("Internal ICU error in resolveImplicitLevels");
-                }
-            }
-        }
-
-        /* look for the last char not a BN or LRE/RLE/LRO/RLO/PDF */
-        for (i = limit - 1;
-             i > start &&
-                 (DirPropFlag(dirProps[i]) & MASK_BN_EXPLICIT) != 0;
-             i--);
-        dirProp = dirProps[i];
-        if ((dirProp == LRI || dirProp == RLI) && limit < length) {
-            isolateCount++;
-            if (isolates[isolateCount] == null)
-                isolates[isolateCount] = new Isolate();
-            isolates[isolateCount].stateImp = stateImp;
-            isolates[isolateCount].state = levState.state;
-            isolates[isolateCount].start1 = start1;
-            isolates[isolateCount].startON = levState.startON;
-        }
-        else
-            processPropertySeq(levState, eor, limit, limit);
-    }
-
-    /* perform (L1) and (X9) ---------------------------------------------------- */
-
-    /*
-     * Reset the embedding levels for some non-graphic characters (L1).
-     * This method also sets appropriate levels for BN, and
-     * explicit embedding types that are supposed to have been removed
-     * from the paragraph in (X9).
-     */
-    private void adjustWSLevels() {
-        int i;
-
-        if ((flags & MASK_WS) != 0) {
-            int flag;
-            i = trailingWSStart;
-            while (i > 0) {
-                /* reset a sequence of WS/BN before eop and B/S to the paragraph paraLevel */
-                while (i > 0 && ((flag = DirPropFlag(dirProps[--i])) & MASK_WS) != 0) {
-                    if (orderParagraphsLTR && (flag & DirPropFlag(B)) != 0) {
-                        levels[i] = 0;
-                    } else {
-                        levels[i] = GetParaLevelAt(i);
-                    }
-                }
-
-                /* reset BN to the next character's paraLevel until B/S, which restarts above loop */
-                /* here, i+1 is guaranteed to be  0) {
-                    flag = DirPropFlag(dirProps[--i]);
-                    if ((flag & MASK_BN_EXPLICIT) != 0) {
-                        levels[i] = levels[i + 1];
-                    } else if (orderParagraphsLTR && (flag & DirPropFlag(B)) != 0) {
-                        levels[i] = 0;
-                        break;
-                    } else if ((flag & MASK_B_S) != 0){
-                        levels[i] = GetParaLevelAt(i);
-                        break;
-                    }
-                }
-            }
-        }
-    }
-
-    private void setParaSuccess() {
-        paraBidi = this;                /* mark successful setPara */
-    }
-
-    private int Bidi_Min(int x, int y) {
-        return x < y ? x : y;
-    }
-
-    private int Bidi_Abs(int x) {
-        return x >= 0 ? x : -x;
-    }
-
-    void setParaRunsOnly(char[] parmText, byte parmParaLevel) {
-        int[] visualMap;
-        String visualText;
-        int saveLength, saveTrailingWSStart;
-        byte[] saveLevels;
-        byte saveDirection;
-        int i, j, visualStart, logicalStart,
-            oldRunCount, runLength, addedRuns, insertRemove,
-            start, limit, step, indexOddBit, logicalPos,
-            index, index1;
-        int saveOptions;
-
-        reorderingMode = REORDER_DEFAULT;
-        int parmLength = parmText.length;
-        if (parmLength == 0) {
-            setPara(parmText, parmParaLevel, null);
-            reorderingMode = REORDER_RUNS_ONLY;
-            return;
-        }
-        /* obtain memory for mapping table and visual text */
-        saveOptions = reorderingOptions;
-        if ((saveOptions & OPTION_INSERT_MARKS) > 0) {
-            reorderingOptions &= ~OPTION_INSERT_MARKS;
-            reorderingOptions |= OPTION_REMOVE_CONTROLS;
-        }
-        parmParaLevel &= 1;             /* accept only 0 or 1 */
-        setPara(parmText, parmParaLevel, null);
-        /* we cannot access directly levels since it is not yet set if
-         * direction is not MIXED
-         */
-        saveLevels = new byte[this.length];
-        System.arraycopy(getLevels(), 0, saveLevels, 0, this.length);
-        saveTrailingWSStart = trailingWSStart;
-
-        /* FOOD FOR THOUGHT: instead of writing the visual text, we could use
-         * the visual map and the dirProps array to drive the second call
-         * to setPara (but must make provision for possible removal of
-         * Bidi controls.  Alternatively, only use the dirProps array via
-         * customized classifier callback.
-         */
-        visualText = writeReordered(DO_MIRRORING);
-        visualMap = getVisualMap();
-        this.reorderingOptions = saveOptions;
-        saveLength = this.length;
-        saveDirection=this.direction;
-
-        this.reorderingMode = REORDER_INVERSE_LIKE_DIRECT;
-        parmParaLevel ^= 1;
-        setPara(visualText, parmParaLevel, null);
-        BidiLine.getRuns(this);
-        /* check if some runs must be split, count how many splits */
-        addedRuns = 0;
-        oldRunCount = this.runCount;
-        visualStart = 0;
-        for (i = 0; i < oldRunCount; i++, visualStart += runLength) {
-            runLength = runs[i].limit - visualStart;
-            if (runLength < 2) {
-                continue;
-            }
-            logicalStart = runs[i].start;
-            for (j = logicalStart+1; j < logicalStart+runLength; j++) {
-                index = visualMap[j];
-                index1 = visualMap[j-1];
-                if ((Bidi_Abs(index-index1)!=1) || (saveLevels[index]!=saveLevels[index1])) {
-                    addedRuns++;
-                }
-            }
-        }
-        if (addedRuns > 0) {
-            getRunsMemory(oldRunCount + addedRuns);
-            if (runCount == 1) {
-                /* because we switch from UBiDi.simpleRuns to UBiDi.runs */
-                runsMemory[0] = runs[0];
-            } else {
-                System.arraycopy(runs, 0, runsMemory, 0, runCount);
-            }
-            runs = runsMemory;
-            runCount += addedRuns;
-            for (i = oldRunCount; i < runCount; i++) {
-                if (runs[i] == null) {
-                    runs[i] = new BidiRun(0, 0, (byte)0);
-                }
-            }
-        }
-        /* split runs which are not consecutive in source text */
-        int newI;
-        for (i = oldRunCount-1; i >= 0; i--) {
-            newI = i + addedRuns;
-            runLength = i==0 ? runs[0].limit :
-                               runs[i].limit - runs[i-1].limit;
-            logicalStart = runs[i].start;
-            indexOddBit = runs[i].level & 1;
-            if (runLength < 2) {
-                if (addedRuns > 0) {
-                    runs[newI].copyFrom(runs[i]);
-                }
-                logicalPos = visualMap[logicalStart];
-                runs[newI].start = logicalPos;
-                runs[newI].level = (byte)(saveLevels[logicalPos] ^ indexOddBit);
-                continue;
-            }
-            if (indexOddBit > 0) {
-                start = logicalStart;
-                limit = logicalStart + runLength - 1;
-                step = 1;
-            } else {
-                start = logicalStart + runLength - 1;
-                limit = logicalStart;
-                step = -1;
-            }
-            for (j = start; j != limit; j += step) {
-                index = visualMap[j];
-                index1 = visualMap[j+step];
-                if ((Bidi_Abs(index-index1)!=1) || (saveLevels[index]!=saveLevels[index1])) {
-                    logicalPos = Bidi_Min(visualMap[start], index);
-                    runs[newI].start = logicalPos;
-                    runs[newI].level = (byte)(saveLevels[logicalPos] ^ indexOddBit);
-                    runs[newI].limit = runs[i].limit;
-                    runs[i].limit -= Bidi_Abs(j - start) + 1;
-                    insertRemove = runs[i].insertRemove & (LRM_AFTER|RLM_AFTER);
-                    runs[newI].insertRemove = insertRemove;
-                    runs[i].insertRemove &= ~insertRemove;
-                    start = j + step;
-                    addedRuns--;
-                    newI--;
-                }
-            }
-            if (addedRuns > 0) {
-                runs[newI].copyFrom(runs[i]);
-            }
-            logicalPos = Bidi_Min(visualMap[start], visualMap[limit]);
-            runs[newI].start = logicalPos;
-            runs[newI].level = (byte)(saveLevels[logicalPos] ^ indexOddBit);
-        }
-
-    cleanup1:
-        /* restore initial paraLevel */
-        this.paraLevel ^= 1;
-    cleanup2:
-        /* restore real text */
-        this.text = parmText;
-        this.length = saveLength;
-        this.originalLength = parmLength;
-        this.direction=saveDirection;
-        this.levels = saveLevels;
-        this.trailingWSStart = saveTrailingWSStart;
-        if (runCount > 1) {
-            this.direction = MIXED;
-        }
-    cleanup3:
-        this.reorderingMode = REORDER_RUNS_ONLY;
-    }
-
-    /**
-     * Perform the Unicode Bidi algorithm. It is defined in the
-     * Unicode Standard Annex #9,
-     * version 13,
-     * also described in The Unicode Standard, Version 4.0 .

-     *
-     * This method takes a piece of plain text containing one or more paragraphs,
-     * with or without externally specified embedding levels from styled
-     * text and computes the left-right-directionality of each character.

-     *
-     * If the entire text is all of the same directionality, then
-     * the method may not perform all the steps described by the algorithm,
-     * i.e., some levels may not be the same as if all steps were performed.
-     * This is not relevant for unidirectional text.

-     * For example, in pure LTR text with numbers the numbers would get
-     * a resolved level of 2 higher than the surrounding text according to
-     * the algorithm. This implementation may set all resolved levels to
-     * the same value in such a case.

-     *
-     * The text can be composed of multiple paragraphs. Occurrence of a block
-     * separator in the text terminates a paragraph, and whatever comes next starts
-     * a new paragraph. The exception to this rule is when a Carriage Return (CR)
-     * is followed by a Line Feed (LF). Both CR and LF are block separators, but
-     * in that case, the pair of characters is considered as terminating the
-     * preceding paragraph, and a new paragraph will be started by a character
-     * coming after the LF.
-     *
-     * Although the text is passed here as a String, it is
-     * stored internally as an array of characters. Therefore the
-     * documentation will refer to indexes of the characters in the text.
-     *
-     * @param text contains the text that the Bidi algorithm will be performed
-     *        on. This text can be retrieved with getText() or
-     *        getTextAsString.

-     *
-     * @param paraLevel specifies the default level for the text;
-     *        it is typically 0 (LTR) or 1 (RTL).
-     *        If the method shall determine the paragraph level from the text,
-     *        then paraLevel can be set to
-     *        either LEVEL_DEFAULT_LTR
-     *        or LEVEL_DEFAULT_RTL; if the text contains multiple
-     *        paragraphs, the paragraph level shall be determined separately for
-     *        each paragraph; if a paragraph does not include any strongly typed
-     *        character, then the desired default is used (0 for LTR or 1 for RTL).
-     *        Any other value between 0 and MAX_EXPLICIT_LEVEL
-     *        is also valid, with odd levels indicating RTL.
-     *
-     * @param embeddingLevels (in) may be used to preset the embedding and override levels,
-     *        ignoring characters like LRE and PDF in the text.
-     *        A level overrides the directional property of its corresponding
-     *        (same index) character if the level has the
-     *        LEVEL_OVERRIDE bit set.


-     *        Except for that bit, it must be
-     *        paraLevel<=embeddingLevels[]<=MAX_EXPLICIT_LEVEL,
-     *        with one exception: a level of zero may be specified for a
-     *        paragraph separator even if paraLevel>0 when multiple
-     *        paragraphs are submitted in the same call to setPara().


-     *        Caution: A reference to this array, not a copy
-     *        of the levels, will be stored in the Bidi object;
-     *        the embeddingLevels
-     *        should not be modified to avoid unexpected results on subsequent
-     *        Bidi operations. However, the setPara() and
-     *        setLine() methods may modify some or all of the
-     *        levels.


-     *        Note: the embeddingLevels array must
-     *        have one entry for each character in text.
-     *
-     * @throws IllegalArgumentException if the values in embeddingLevels are
-     *         not within the allowed range
-     *
-     * @see #LEVEL_DEFAULT_LTR
-     * @see #LEVEL_DEFAULT_RTL
-     * @see #LEVEL_OVERRIDE
-     * @see #MAX_EXPLICIT_LEVEL
-     * @stable ICU 3.8
-     */
-    void setPara(String text, byte paraLevel, byte[] embeddingLevels)
-    {
-        if (text == null) {
-            setPara(new char[0], paraLevel, embeddingLevels);
-        } else {
-            setPara(text.toCharArray(), paraLevel, embeddingLevels);
-        }
-    }
-
-    /**
-     * Perform the Unicode Bidi algorithm. It is defined in the
-     * Unicode Standard Annex #9,
-     * version 13,
-     * also described in The Unicode Standard, Version 4.0 .

-     *
-     * This method takes a piece of plain text containing one or more paragraphs,
-     * with or without externally specified embedding levels from styled
-     * text and computes the left-right-directionality of each character.

-     *
-     * If the entire text is all of the same directionality, then
-     * the method may not perform all the steps described by the algorithm,
-     * i.e., some levels may not be the same as if all steps were performed.
-     * This is not relevant for unidirectional text.

-     * For example, in pure LTR text with numbers the numbers would get
-     * a resolved level of 2 higher than the surrounding text according to
-     * the algorithm. This implementation may set all resolved levels to
-     * the same value in such a case.
-     *
-     * The text can be composed of multiple paragraphs. Occurrence of a block
-     * separator in the text terminates a paragraph, and whatever comes next starts
-     * a new paragraph. The exception to this rule is when a Carriage Return (CR)
-     * is followed by a Line Feed (LF). Both CR and LF are block separators, but
-     * in that case, the pair of characters is considered as terminating the
-     * preceding paragraph, and a new paragraph will be started by a character
-     * coming after the LF.
-     *
-     * The text is stored internally as an array of characters. Therefore the
-     * documentation will refer to indexes of the characters in the text.
-     *
-     * @param chars contains the text that the Bidi algorithm will be performed
-     *        on. This text can be retrieved with getText() or
-     *        getTextAsString.

-     *
-     * @param paraLevel specifies the default level for the text;
-     *        it is typically 0 (LTR) or 1 (RTL).
-     *        If the method shall determine the paragraph level from the text,
-     *        then paraLevel can be set to
-     *        either LEVEL_DEFAULT_LTR
-     *        or LEVEL_DEFAULT_RTL; if the text contains multiple
-     *        paragraphs, the paragraph level shall be determined separately for
-     *        each paragraph; if a paragraph does not include any strongly typed
-     *        character, then the desired default is used (0 for LTR or 1 for RTL).
-     *        Any other value between 0 and MAX_EXPLICIT_LEVEL
-     *        is also valid, with odd levels indicating RTL.
-     *
-     * @param embeddingLevels (in) may be used to preset the embedding and
-     *        override levels, ignoring characters like LRE and PDF in the text.
-     *        A level overrides the directional property of its corresponding
-     *        (same index) character if the level has the
-     *        LEVEL_OVERRIDE bit set.


-     *        Except for that bit, it must be
-     *        paraLevel<=embeddingLevels[]<=MAX_EXPLICIT_LEVEL,
-     *        with one exception: a level of zero may be specified for a
-     *        paragraph separator even if paraLevel>0 when multiple
-     *        paragraphs are submitted in the same call to setPara().


-     *        Caution: A reference to this array, not a copy
-     *        of the levels, will be stored in the Bidi object;
-     *        the embeddingLevels
-     *        should not be modified to avoid unexpected results on subsequent
-     *        Bidi operations. However, the setPara() and
-     *        setLine() methods may modify some or all of the
-     *        levels.


-     *        Note: the embeddingLevels array must
-     *        have one entry for each character in text.
-     *
-     * @throws IllegalArgumentException if the values in embeddingLevels are
-     *         not within the allowed range
-     *
-     * @see #LEVEL_DEFAULT_LTR
-     * @see #LEVEL_DEFAULT_RTL
-     * @see #LEVEL_OVERRIDE
-     * @see #MAX_EXPLICIT_LEVEL
-     * @stable ICU 3.8
-     */
-    void setPara(char[] chars, byte paraLevel, byte[] embeddingLevels)
-    {
-        /* check the argument values */
-        if (paraLevel < LEVEL_DEFAULT_LTR) {
-            verifyRange(paraLevel, 0, MAX_EXPLICIT_LEVEL + 1);
-        }
-        if (chars == null) {
-            chars = new char[0];
-        }
-
-        /* special treatment for RUNS_ONLY mode */
-        if (reorderingMode == REORDER_RUNS_ONLY) {
-            setParaRunsOnly(chars, paraLevel);
-            return;
-        }
-
-        /* initialize the Bidi object */
-        this.paraBidi = null;          /* mark unfinished setPara */
-        this.text = chars;
-        this.length = this.originalLength = this.resultLength = text.length;
-        this.paraLevel = paraLevel;
-        this.direction = (byte)(paraLevel & 1);
-        this.paraCount = 1;
-
-        /* Allocate zero-length arrays instead of setting to null here; then
-         * checks for null in various places can be eliminated.
-         */
-        dirProps = new byte[0];
-        levels = new byte[0];
-        runs = new BidiRun[0];
-        isGoodLogicalToVisualRunsMap = false;
-        insertPoints.size = 0;          /* clean up from last call */
-        insertPoints.confirmed = 0;     /* clean up from last call */
-
-        /*
-         * Save the original paraLevel if contextual; otherwise, set to 0.
-         */
-        defaultParaLevel = IsDefaultLevel(paraLevel) ? paraLevel : 0;
-
-        if (length == 0) {
-            /*
-             * For an empty paragraph, create a Bidi object with the paraLevel and
-             * the flags and the direction set but without allocating zero-length arrays.
-             * There is nothing more to do.
-             */
-            if (IsDefaultLevel(paraLevel)) {
-                this.paraLevel &= 1;
-                defaultParaLevel = 0;
-            }
-            flags = DirPropFlagLR(paraLevel);
-            runCount = 0;
-            paraCount = 0;
-            setParaSuccess();
-            return;
-        }
-
-        runCount = -1;
-
-        /*
-         * Get the directional properties,
-         * the flags bit-set, and
-         * determine the paragraph level if necessary.
-         */
-        getDirPropsMemory(length);
-        dirProps = dirPropsMemory;
-        getDirProps();
-        /* the processed length may have changed if OPTION_STREAMING is set */
-        trailingWSStart = length;  /* the levels[] will reflect the WS run */
-
-        /* are explicit levels specified? */
-        if (embeddingLevels == null) {
-            /* no: determine explicit levels according to the (Xn) rules */
-            getLevelsMemory(length);
-            levels = levelsMemory;
-            direction = resolveExplicitLevels();
-        } else {
-            /* set BN for all explicit codes, check that all levels are 0 or paraLevel..MAX_EXPLICIT_LEVEL */
-            levels = embeddingLevels;
-            direction = checkExplicitLevels();
-        }
-
-        /* allocate isolate memory */
-        if (isolateCount > 0) {
-            if (isolates == null || isolates.length < isolateCount)
-                isolates = new Isolate[isolateCount + 3];   /* keep some reserve */
-        }
-        isolateCount = -1;              /* current isolates stack entry == none */
-
-        /*
-         * The steps after (X9) in the Bidi algorithm are performed only if
-         * the paragraph text has mixed directionality!
-         */
-        switch (direction) {
-        case LTR:
-            /* all levels are implicitly at paraLevel (important for getLevels()) */
-            trailingWSStart = 0;
-            break;
-        case RTL:
-            /* all levels are implicitly at paraLevel (important for getLevels()) */
-            trailingWSStart = 0;
-            break;
-        default:
-            /*
-             *  Choose the right implicit state table
-             */
-            switch(reorderingMode) {
-            case REORDER_DEFAULT:
-                this.impTabPair = impTab_DEFAULT;
-                break;
-            case REORDER_NUMBERS_SPECIAL:
-                this.impTabPair = impTab_NUMBERS_SPECIAL;
-                break;
-            case REORDER_GROUP_NUMBERS_WITH_R:
-                this.impTabPair = impTab_GROUP_NUMBERS_WITH_R;
-                break;
-            case REORDER_RUNS_ONLY:
-                /* we should never get here */
-                throw new InternalError("Internal ICU error in setPara");
-                /* break; */
-            case REORDER_INVERSE_NUMBERS_AS_L:
-                this.impTabPair = impTab_INVERSE_NUMBERS_AS_L;
-                break;
-            case REORDER_INVERSE_LIKE_DIRECT:
-                if ((reorderingOptions & OPTION_INSERT_MARKS) != 0) {
-                    this.impTabPair = impTab_INVERSE_LIKE_DIRECT_WITH_MARKS;
-                } else {
-                    this.impTabPair = impTab_INVERSE_LIKE_DIRECT;
-                }
-                break;
-            case REORDER_INVERSE_FOR_NUMBERS_SPECIAL:
-                if ((reorderingOptions & OPTION_INSERT_MARKS) != 0) {
-                    this.impTabPair = impTab_INVERSE_FOR_NUMBERS_SPECIAL_WITH_MARKS;
-                } else {
-                    this.impTabPair = impTab_INVERSE_FOR_NUMBERS_SPECIAL;
-                }
-                break;
-            }
-            /*
-             * If there are no external levels specified and there
-             * are no significant explicit level codes in the text,
-             * then we can treat the entire paragraph as one run.
-             * Otherwise, we need to perform the following rules on runs of
-             * the text with the same embedding levels. (X10)
-             * "Significant" explicit level codes are ones that actually
-             * affect non-BN characters.
-             * Examples for "insignificant" ones are empty embeddings
-             * LRE-PDF, LRE-RLE-PDF-PDF, etc.
-             */
-            if (embeddingLevels == null && paraCount <= 1 &&
-                (flags & DirPropFlagMultiRuns) == 0) {
-                resolveImplicitLevels(0, length,
-                        GetLRFromLevel(GetParaLevelAt(0)),
-                        GetLRFromLevel(GetParaLevelAt(length - 1)));
-            } else {
-                /* sor, eor: start and end types of same-level-run */
-                int start, limit = 0;
-                byte level, nextLevel;
-                short sor, eor;
-
-                /* determine the first sor and set eor to it because of the loop body (sor=eor there) */
-                level = GetParaLevelAt(0);
-                nextLevel = levels[0];
-                if (level < nextLevel) {
-                    eor = GetLRFromLevel(nextLevel);
-                } else {
-                    eor = GetLRFromLevel(level);
-                }
-
-                do {
-                    /* determine start and limit of the run (end points just behind the run) */
-
-                    /* the values for this run's start are the same as for the previous run's end */
-                    start = limit;
-                    level = nextLevel;
-                    if ((start > 0) && (dirProps[start - 1] == B)) {
-                        /* except if this is a new paragraph, then set sor = para level */
-                        sor = GetLRFromLevel(GetParaLevelAt(start));
-                    } else {
-                        sor = eor;
-                    }
-
-                    /* search for the limit of this run */
-                    while ((++limit < length) &&
-                           ((levels[limit] == level) ||
-                            ((DirPropFlag(dirProps[limit]) & MASK_BN_EXPLICIT) != 0))) {}
-
-                    /* get the correct level of the next run */
-                    if (limit < length) {
-                        nextLevel = levels[limit];
-                    } else {
-                        nextLevel = GetParaLevelAt(length - 1);
-                    }
-
-                    /* determine eor from max(level, nextLevel); sor is last run's eor */
-                    if (NoOverride(level) < NoOverride(nextLevel)) {
-                        eor = GetLRFromLevel(nextLevel);
-                    } else {
-                        eor = GetLRFromLevel(level);
-                    }
-
-                    /* if the run consists of overridden directional types, then there
-                       are no implicit types to be resolved */
-                    if ((level & LEVEL_OVERRIDE) == 0) {
-                        resolveImplicitLevels(start, limit, sor, eor);
-                    } else {
-                        /* remove the LEVEL_OVERRIDE flags */
-                        do {
-                            levels[start++] &= ~LEVEL_OVERRIDE;
-                        } while (start < limit);
-                    }
-                } while (limit  < length);
-            }
-
-            /* reset the embedding levels for some non-graphic characters (L1), (X9) */
-            adjustWSLevels();
-
-            break;
-        }
-
-        /* add RLM for inverse Bidi with contextual orientation resolving
-         * to RTL which would not round-trip otherwise
-         */
-        if ((defaultParaLevel > 0) &&
-            ((reorderingOptions & OPTION_INSERT_MARKS) != 0) &&
-            ((reorderingMode == REORDER_INVERSE_LIKE_DIRECT) ||
-             (reorderingMode == REORDER_INVERSE_FOR_NUMBERS_SPECIAL))) {
-            int start, last;
-            byte level;
-            byte dirProp;
-            for (int i = 0; i < paraCount; i++) {
-                last = paras_limit[i] - 1;
-                level = paras_level[i];
-                if (level == 0)
-                    continue;           /* LTR paragraph */
-                start = i == 0 ? 0 : paras_limit[i - 1];
-                for (int j = last; j >= start; j--) {
-                    dirProp = dirProps[j];
-                    if (dirProp == L) {
-                        if (j < last) {
-                            while (dirProps[last] == B) {
-                                last--;
-                            }
-                        }
-                        addPoint(last, RLM_BEFORE);
-                        break;
-                    }
-                    if ((DirPropFlag(dirProp) & MASK_R_AL) != 0) {
-                        break;
-                    }
-                }
-            }
-        }
-
-        if ((reorderingOptions & OPTION_REMOVE_CONTROLS) != 0) {
-            resultLength -= controlCount;
-        } else {
-            resultLength += insertPoints.size;
-        }
-        setParaSuccess();
-    }
-
-    /**
-     * Perform the Unicode Bidi algorithm on a given paragraph, as defined in the
-     * Unicode Standard Annex #9,
-     * version 13,
-     * also described in The Unicode Standard, Version 4.0 .

-     *
-     * This method takes a paragraph of text and computes the
-     * left-right-directionality of each character. The text should not
-     * contain any Unicode block separators.

-     *
-     * The RUN_DIRECTION attribute in the text, if present, determines the base
-     * direction (left-to-right or right-to-left). If not present, the base
-     * direction is computed using the Unicode Bidirectional Algorithm,
-     * defaulting to left-to-right if there are no strong directional characters
-     * in the text. This attribute, if present, must be applied to all the text
-     * in the paragraph.

-     *
-     * The BIDI_EMBEDDING attribute in the text, if present, represents
-     * embedding level information. Negative values from -1 to -62 indicate
-     * overrides at the absolute value of the level. Positive values from 1 to
-     * 62 indicate embeddings. Where values are zero or not defined, the base
-     * embedding level as determined by the base direction is assumed.

-     *
-     * The NUMERIC_SHAPING attribute in the text, if present, converts European
-     * digits to other decimal digits before running the bidi algorithm. This
-     * attribute, if present, must be applied to all the text in the paragraph.
-     *
-     * If the entire text is all of the same directionality, then
-     * the method may not perform all the steps described by the algorithm,
-     * i.e., some levels may not be the same as if all steps were performed.
-     * This is not relevant for unidirectional text.

-     * For example, in pure LTR text with numbers the numbers would get
-     * a resolved level of 2 higher than the surrounding text according to
-     * the algorithm. This implementation may set all resolved levels to
-     * the same value in such a case.

-     *
-     * @param paragraph a paragraph of text with optional character and
-     *        paragraph attribute information
-     * @stable ICU 3.8
-     */
-    public void setPara(AttributedCharacterIterator paragraph)
-    {
-        byte paraLvl;
-        char ch = paragraph.first();
-        Boolean runDirection =
-          (Boolean) paragraph.getAttribute(TextAttributeConstants.RUN_DIRECTION);
-        Object shaper = paragraph.getAttribute(TextAttributeConstants.NUMERIC_SHAPING);
-
-        if (runDirection == null) {
-            paraLvl = LEVEL_DEFAULT_LTR;
-        } else {
-            paraLvl = (runDirection.equals(TextAttributeConstants.RUN_DIRECTION_LTR)) ?
-                        LTR : RTL;
-        }
-
-        byte[] lvls = null;
-        int len = paragraph.getEndIndex() - paragraph.getBeginIndex();
-        byte[] embeddingLevels = new byte[len];
-        char[] txt = new char[len];
-        int i = 0;
-        while (ch != AttributedCharacterIterator.DONE) {
-            txt[i] = ch;
-            Integer embedding =
-                (Integer) paragraph.getAttribute(TextAttributeConstants.BIDI_EMBEDDING);
-            if (embedding != null) {
-                byte level = embedding.byteValue();
-                if (level == 0) {
-                    /* no-op */
-                } else if (level < 0) {
-                    lvls = embeddingLevels;
-                    embeddingLevels[i] = (byte)((0 - level) | LEVEL_OVERRIDE);
-                } else {
-                    lvls = embeddingLevels;
-                    embeddingLevels[i] = level;
-                }
-            }
-            ch = paragraph.next();
-            ++i;
-        }
-
-        if (shaper != null) {
-            NumericShapings.shape(shaper, txt, 0, len);
-        }
-        setPara(txt, paraLvl, lvls);
-    }
-
-    /**
-     * Specify whether block separators must be allocated level zero,
-     * so that successive paragraphs will progress from left to right.
-     * This method must be called before setPara().
-     * Paragraph separators (B) may appear in the text.  Setting them to level zero
-     * means that all paragraph separators (including one possibly appearing
-     * in the last text position) are kept in the reordered text after the text
-     * that they follow in the source text.
-     * When this feature is not enabled, a paragraph separator at the last
-     * position of the text before reordering will go to the first position
-     * of the reordered text when the paragraph level is odd.
-     *
-     * @param ordarParaLTR specifies whether paragraph separators (B) must
-     * receive level 0, so that successive paragraphs progress from left to right.
-     *
-     * @see #setPara
-     * @stable ICU 3.8
-     */
-    public void orderParagraphsLTR(boolean ordarParaLTR) {
-        orderParagraphsLTR = ordarParaLTR;
-    }
-
-    /**
-     * Get the directionality of the text.
-     *
-     * @return a value of LTR, RTL or MIXED
-     *         that indicates if the entire text
-     *         represented by this object is unidirectional,
-     *         and which direction, or if it is mixed-directional.
-     *
-     * @throws IllegalStateException if this call is not preceded by a successful
-     *         call to setPara or setLine
-     *
-     * @see #LTR
-     * @see #RTL
-     * @see #MIXED
-     * @stable ICU 3.8
-     */
-    public byte getDirection()
-    {
-        verifyValidParaOrLine();
-        return direction;
-    }
-
-    /**
-     * Get the length of the text.
-     *
-     * @return The length of the text that the Bidi object was
-     *         created for.
-     *
-     * @throws IllegalStateException if this call is not preceded by a successful
-     *         call to setPara or setLine
-     * @stable ICU 3.8
-     */
-    public int getLength()
-    {
-        verifyValidParaOrLine();
-        return originalLength;
-    }
-
-    /* paragraphs API methods ------------------------------------------------- */
-
-    /**
-     * Get the paragraph level of the text.
-     *
-     * @return The paragraph level. If there are multiple paragraphs, their
-     *         level may vary if the required paraLevel is LEVEL_DEFAULT_LTR or
-     *         LEVEL_DEFAULT_RTL.  In that case, the level of the first paragraph
-     *         is returned.
-     *
-     * @throws IllegalStateException if this call is not preceded by a successful
-     *         call to setPara or setLine
-     *
-     * @see #LEVEL_DEFAULT_LTR
-     * @see #LEVEL_DEFAULT_RTL
-     * @see #getParagraph
-     * @see #getParagraphByIndex
-     * @stable ICU 3.8
-     */
-    public byte getParaLevel()
-    {
-        verifyValidParaOrLine();
-        return paraLevel;
-    }
-
-    /**
-     * Retrieves the Bidi class for a given code point.
-     * 
If a BidiClassifier is defined and returns a value
-     * other than CLASS_DEFAULT, that value is used; otherwise
-     * the default class determination mechanism is invoked.

-     *
-     * @param c The code point to get a Bidi class for.
-     *
-     * @return The Bidi class for the character c that is in effect
-     *         for this Bidi instance.
-     *
-     * @stable ICU 3.8
-     */
-    public int getCustomizedClass(int c) {
-        int dir;
-
-        dir = bdp.getClass(c);
-        if (dir >= CHAR_DIRECTION_COUNT)
-            dir = ON;
-        return dir;
-    }
-
-    /**
-     * setLine() returns a Bidi object to
-     * contain the reordering information, especially the resolved levels,
-     * for all the characters in a line of text. This line of text is
-     * specified by referring to a Bidi object representing
-     * this information for a piece of text containing one or more paragraphs,
-     * and by specifying a range of indexes in this text.

-     * In the new line object, the indexes will range from 0 to limit-start-1.

-     *
-     * This is used after calling setPara()
-     * for a piece of text, and after line-breaking on that text.
-     * It is not necessary if each paragraph is treated as a single line.

-     *
-     * After line-breaking, rules (L1) and (L2) for the treatment of
-     * trailing WS and for reordering are performed on
-     * a Bidi object that represents a line.

-     *
-     * Important: the line Bidi object may
-     * reference data within the global text Bidi object.
-     * You should not alter the content of the global text object until
-     * you are finished using the line object.
-     *
-     * @param start is the line's first index into the text.
-     *
-     * @param limit is just behind the line's last index into the text
-     *        (its last index +1).
-     *
-     * @return a Bidi object that will now represent a line of the text.
-     *
-     * @throws IllegalStateException if this call is not preceded by a successful
-     *         call to setPara
-     * @throws IllegalArgumentException if start and limit are not in the range
-     *         0<=start<limit<=getProcessedLength(),
-     *         or if the specified line crosses a paragraph boundary
-     *
-     * @see #setPara
-     * @see #getProcessedLength
-     * @stable ICU 3.8
-     */
-    public Bidi setLine(Bidi bidi, BidiBase bidiBase, Bidi newBidi, BidiBase newBidiBase, int start, int limit)
-    {
-        verifyValidPara();
-        verifyRange(start, 0, limit);
-        verifyRange(limit, 0, length+1);
-
-        return BidiLine.setLine(this, newBidi, newBidiBase, start, limit);
-    }
-
-    /**
-     * Get the level for one character.
-     *
-     * @param charIndex the index of a character.
-     *
-     * @return The level for the character at charIndex.
-     *
-     * @throws IllegalStateException if this call is not preceded by a successful
-     *         call to setPara or setLine
-     * @throws IllegalArgumentException if charIndex is not in the range
-     *         0<=charIndex<getProcessedLength()
-     *
-     * @see #getProcessedLength
-     * @stable ICU 3.8
-     */
-    public byte getLevelAt(int charIndex)
-    {
-        // for backward compatibility
-        if (charIndex < 0 || charIndex >= length) {
-            return (byte)getBaseLevel();
-        }
-
-        verifyValidParaOrLine();
-        verifyRange(charIndex, 0, length);
-        return BidiLine.getLevelAt(this, charIndex);
-    }
-
-    /**
-     * Get an array of levels for each character.

-     *
-     * Note that this method may allocate memory under some
-     * circumstances, unlike getLevelAt().
-     *
-     * @return The levels array for the text,
-     *         or null if an error occurs.
-     *
-     * @throws IllegalStateException if this call is not preceded by a successful
-     *         call to setPara or setLine
-     * @stable ICU 3.8
-     */
-    byte[] getLevels()
-    {
-        verifyValidParaOrLine();
-        if (length <= 0) {
-            return new byte[0];
-        }
-        return BidiLine.getLevels(this);
-    }
-
-    /**
-     * Get the number of runs.
-     * This method may invoke the actual reordering on the
-     * Bidi object, after setPara()
-     * may have resolved only the levels of the text. Therefore,
-     * countRuns() may have to allocate memory,
-     * and may throw an exception if it fails to do so.
-     *
-     * @return The number of runs.
-     *
-     * @throws IllegalStateException if this call is not preceded by a successful
-     *         call to setPara or setLine
-     * @stable ICU 3.8
-     */
-    public int countRuns()
-    {
-        verifyValidParaOrLine();
-        BidiLine.getRuns(this);
-        return runCount;
-    }
-
-    /**
-     *
-     * Get a BidiRun object according to its index. BidiRun methods
-     * may be used to retrieve the run's logical start, length and level,
-     * which can be even for an LTR run or odd for an RTL run.
-     * In an RTL run, the character at the logical start is
-     * visually on the right of the displayed run.
-     * The length is the number of characters in the run.

-     * countRuns() is normally called
-     * before the runs are retrieved.
-     *
-     * 

-     *  Example:
-     * 
-     *  Bidi bidi = new Bidi();
-     *  String text = "abc 123 DEFG xyz";
-     *  bidi.setPara(text, Bidi.RTL, null);
-     *  int i, count=bidi.countRuns(), logicalStart, visualIndex=0, length;
-     *  BidiRun run;
-     *  for (i = 0; i < count; ++i) {
-     *      run = bidi.getVisualRun(i);
-     *      logicalStart = run.getStart();
-     *      length = run.getLength();
-     *      if (Bidi.LTR == run.getEmbeddingLevel()) {
-     *          do { // LTR
-     *              show_char(text.charAt(logicalStart++), visualIndex++);
-     *          } while (--length > 0);
-     *      } else {
-     *          logicalStart += length;  // logicalLimit
-     *          do { // RTL
-     *              show_char(text.charAt(--logicalStart), visualIndex++);
-     *          } while (--length > 0);
-     *      }
-     *  }
-     * 

-     * 

-     * Note that in right-to-left runs, code like this places
-     * second surrogates before first ones (which is generally a bad idea)
-     * and combining characters before base characters.
-     * 

-     * Use of {@link #writeReordered}, optionally with the
-     * {@link #KEEP_BASE_COMBINING} option, can be considered in
-     * order to avoid these issues.
-     *
-     * @param runIndex is the number of the run in visual order, in the
-     *        range [0..countRuns()-1].
-     *
-     * @return a BidiRun object containing the details of the run. The
-     *         directionality of the run is
-     *         LTR==0 or RTL==1,
-     *         never MIXED.
-     *
-     * @throws IllegalStateException if this call is not preceded by a successful
-     *         call to setPara or setLine
-     * @throws IllegalArgumentException if runIndex is not in
-     *         the range 0<=runIndex<countRuns()
-     *
-     * @see #countRuns()
-     * @see com.ibm.icu.text.BidiRun
-     * @see com.ibm.icu.text.BidiRun#getStart()
-     * @see com.ibm.icu.text.BidiRun#getLength()
-     * @see com.ibm.icu.text.BidiRun#getEmbeddingLevel()
-     * @stable ICU 3.8
-     */
-    BidiRun getVisualRun(int runIndex)
-    {
-        verifyValidParaOrLine();
-        BidiLine.getRuns(this);
-        verifyRange(runIndex, 0, runCount);
-        return BidiLine.getVisualRun(this, runIndex);
-    }
-
-    /**
-     * Get a visual-to-logical index map (array) for the characters in the
-     * Bidi (paragraph or line) object.
-     * 

-     * Some values in the map may be MAP_NOWHERE if the
-     * corresponding text characters are Bidi marks inserted in the visual
-     * output by the option OPTION_INSERT_MARKS.
-     * 

-     * When the visual output is altered by using options of
-     * writeReordered() such as INSERT_LRM_FOR_NUMERIC,
-     * KEEP_BASE_COMBINING, OUTPUT_REVERSE,
-     * REMOVE_BIDI_CONTROLS, the logical positions returned may not
-     * be correct. It is advised to use, when possible, reordering options
-     * such as {@link #OPTION_INSERT_MARKS} and {@link #OPTION_REMOVE_CONTROLS}.
-     *
-     * @return an array of getResultLength()
-     *        indexes which will reflect the reordering of the characters.


-     *        The index map will result in
-     *        indexMap[visualIndex]==logicalIndex, where
-     *        indexMap represents the returned array.
-     *
-     * @throws IllegalStateException if this call is not preceded by a successful
-     *         call to setPara or setLine
-     *
-     * @see #getLogicalMap
-     * @see #getLogicalIndex
-     * @see #getResultLength
-     * @see #MAP_NOWHERE
-     * @see #OPTION_INSERT_MARKS
-     * @see #writeReordered
-     * @stable ICU 3.8
-     */
-    private int[] getVisualMap()
-    {
-        /* countRuns() checks successful call to setPara/setLine */
-        countRuns();
-        if (resultLength <= 0) {
-            return new int[0];
-        }
-        return BidiLine.getVisualMap(this);
-    }
-
-    /**
-     * This is a convenience method that does not use a Bidi object.
-     * It is intended to be used for when an application has determined the levels
-     * of objects (character sequences) and just needs to have them reordered (L2).
-     * This is equivalent to using getVisualMap() on a
-     * Bidi object.
-     *
-     * @param levels is an array of levels that have been determined by
-     *        the application.
-     *
-     * @return an array of levels.length
-     *        indexes which will reflect the reordering of the characters.

-     *        The index map will result in
-     *        indexMap[visualIndex]==logicalIndex, where
-     *        indexMap represents the returned array.
-     *
-     * @stable ICU 3.8
-     */
-    private static int[] reorderVisual(byte[] levels)
-    {
-        return BidiLine.reorderVisual(levels);
-    }
-
-    /**
-     * Constant indicating that the base direction depends on the first strong
-     * directional character in the text according to the Unicode Bidirectional
-     * Algorithm. If no strong directional character is present, the base
-     * direction is right-to-left.
-     * @stable ICU 3.8
-     */
-    public static final int DIRECTION_DEFAULT_RIGHT_TO_LEFT = LEVEL_DEFAULT_RTL;
-
-    /**
-     * Create Bidi from the given text, embedding, and direction information.
-     * The embeddings array may be null. If present, the values represent
-     * embedding level information. Negative values from -1 to -61 indicate
-     * overrides at the absolute value of the level. Positive values from 1 to
-     * 61 indicate embeddings. Where values are zero, the base embedding level
-     * as determined by the base direction is assumed.

-     *
-     * Note: this constructor calls setPara() internally.
-     *
-     * @param text an array containing the paragraph of text to process.
-     * @param textStart the index into the text array of the start of the
-     *        paragraph.
-     * @param embeddings an array containing embedding values for each character
-     *        in the paragraph. This can be null, in which case it is assumed
-     *        that there is no external embedding information.
-     * @param embStart the index into the embedding array of the start of the
-     *        paragraph.
-     * @param paragraphLength the length of the paragraph in the text and
-     *        embeddings arrays.
-     * @param flags a collection of flags that control the algorithm. The
-     *        algorithm understands the flags DIRECTION_LEFT_TO_RIGHT,
-     *        DIRECTION_RIGHT_TO_LEFT, DIRECTION_DEFAULT_LEFT_TO_RIGHT, and
-     *        DIRECTION_DEFAULT_RIGHT_TO_LEFT. Other values are reserved.
-     *
-     * @throws IllegalArgumentException if the values in embeddings are
-     *         not within the allowed range
-     *
-     * @see #DIRECTION_LEFT_TO_RIGHT
-     * @see #DIRECTION_RIGHT_TO_LEFT
-     * @see #DIRECTION_DEFAULT_LEFT_TO_RIGHT
-     * @see #DIRECTION_DEFAULT_RIGHT_TO_LEFT
-     * @stable ICU 3.8
-     */
-    public BidiBase(char[] text,
-            int textStart,
-            byte[] embeddings,
-            int embStart,
-            int paragraphLength,
-            int flags)
-    {
-        this(0, 0);
-        byte paraLvl;
-        switch (flags) {
-        case Bidi.DIRECTION_LEFT_TO_RIGHT:
-        default:
-            paraLvl = LTR;
-            break;
-        case Bidi.DIRECTION_RIGHT_TO_LEFT:
-            paraLvl = RTL;
-            break;
-        case Bidi.DIRECTION_DEFAULT_LEFT_TO_RIGHT:
-            paraLvl = LEVEL_DEFAULT_LTR;
-            break;
-        case Bidi.DIRECTION_DEFAULT_RIGHT_TO_LEFT:
-            paraLvl = LEVEL_DEFAULT_RTL;
-            break;
-        }
-        byte[] paraEmbeddings;
-        if (embeddings == null) {
-            paraEmbeddings = null;
-        } else {
-            paraEmbeddings = new byte[paragraphLength];
-            byte lev;
-            for (int i = 0; i < paragraphLength; i++) {
-                lev = embeddings[i + embStart];
-                if (lev < 0) {
-                    lev = (byte)((- lev) | LEVEL_OVERRIDE);
-                } else if (lev == 0) {
-                    lev = paraLvl;
-                    if (paraLvl > MAX_EXPLICIT_LEVEL) {
-                        lev &= 1;
-                    }
-                }
-                paraEmbeddings[i] = lev;
-            }
-        }
-
-        char[] paraText = new char[paragraphLength];
-        System.arraycopy(text, textStart, paraText, 0, paragraphLength);
-        setPara(paraText, paraLvl, paraEmbeddings);
-    }
-
-    /**
-     * Return true if the line is not left-to-right or right-to-left. This means
-     * it either has mixed runs of left-to-right and right-to-left text, or the
-     * base direction differs from the direction of the only run of text.
-     *
-     * @return true if the line is not left-to-right or right-to-left.
-     *
-     * @throws IllegalStateException if this call is not preceded by a successful
-     *         call to setPara
-     * @stable ICU 3.8
-     */
-    public boolean isMixed()
-    {
-        return (!isLeftToRight() && !isRightToLeft());
-    }
-
-    /**
-     * Return true if the line is all left-to-right text and the base direction
-     * is left-to-right.
-     *
-     * @return true if the line is all left-to-right text and the base direction
-     *         is left-to-right.
-     *
-     * @throws IllegalStateException if this call is not preceded by a successful
-     *         call to setPara
-     * @stable ICU 3.8
-     */
-    public boolean isLeftToRight()
-    {
-        return (getDirection() == LTR && (paraLevel & 1) == 0);
-    }
-
-    /**
-     * Return true if the line is all right-to-left text, and the base direction
-     * is right-to-left
-     *
-     * @return true if the line is all right-to-left text, and the base
-     *         direction is right-to-left
-     *
-     * @throws IllegalStateException if this call is not preceded by a successful
-     *         call to setPara
-     * @stable ICU 3.8
-     */
-    public boolean isRightToLeft()
-    {
-        return (getDirection() == RTL && (paraLevel & 1) == 1);
-    }
-
-    /**
-     * Return true if the base direction is left-to-right
-     *
-     * @return true if the base direction is left-to-right
-     *
-     * @throws IllegalStateException if this call is not preceded by a successful
-     *         call to setPara or setLine
-     *
-     * @stable ICU 3.8
-     */
-    public boolean baseIsLeftToRight()
-    {
-        return (getParaLevel() == LTR);
-    }
-
-    /**
-     * Return the base level (0 if left-to-right, 1 if right-to-left).
-     *
-     * @return the base level
-     *
-     * @throws IllegalStateException if this call is not preceded by a successful
-     *         call to setPara or setLine
-     *
-     * @stable ICU 3.8
-     */
-    public int getBaseLevel()
-    {
-        return getParaLevel();
-    }
-
-    /**
-     * Compute the logical to visual run mapping
-     */
-     void getLogicalToVisualRunsMap()
-     {
-        if (isGoodLogicalToVisualRunsMap) {
-            return;
-        }
-        int count = countRuns();
-        if ((logicalToVisualRunsMap == null) ||
-            (logicalToVisualRunsMap.length < count)) {
-            logicalToVisualRunsMap = new int[count];
-        }
-        int i;
-        long[] keys = new long[count];
-        for (i = 0; i < count; i++) {
-            keys[i] = ((long)(runs[i].start)<<32) + i;
-        }
-        Arrays.sort(keys);
-        for (i = 0; i < count; i++) {
-            logicalToVisualRunsMap[i] = (int)(keys[i] & 0x00000000FFFFFFFF);
-        }
-        isGoodLogicalToVisualRunsMap = true;
-     }
-
-    /**
-     * Return the level of the nth logical run in this line.
-     *
-     * @param run the index of the run, between 0 and countRuns()-1
-     *
-     * @return the level of the run
-     *
-     * @throws IllegalStateException if this call is not preceded by a successful
-     *         call to setPara or setLine
-     * @throws IllegalArgumentException if run is not in
-     *         the range 0<=run<countRuns()
-     * @stable ICU 3.8
-     */
-    public int getRunLevel(int run)
-    {
-        verifyValidParaOrLine();
-        BidiLine.getRuns(this);
-
-        // for backward compatibility
-        if (run < 0 || run >= runCount) {
-            return getParaLevel();
-        }
-
-        getLogicalToVisualRunsMap();
-        return runs[logicalToVisualRunsMap[run]].level;
-    }
-
-    /**
-     * Return the index of the character at the start of the nth logical run in
-     * this line, as an offset from the start of the line.
-     *
-     * @param run the index of the run, between 0 and countRuns()
-     *
-     * @return the start of the run
-     *
-     * @throws IllegalStateException if this call is not preceded by a successful
-     *         call to setPara or setLine
-     * @throws IllegalArgumentException if run is not in
-     *         the range 0<=run<countRuns()
-     * @stable ICU 3.8
-     */
-    public int getRunStart(int run)
-    {
-        verifyValidParaOrLine();
-        BidiLine.getRuns(this);
-
-        // for backward compatibility
-        if (runCount == 1) {
-            return 0;
-        } else if (run == runCount) {
-            return length;
-        }
-
-        getLogicalToVisualRunsMap();
-        return runs[logicalToVisualRunsMap[run]].start;
-    }
-
-    /**
-     * Return the index of the character past the end of the nth logical run in
-     * this line, as an offset from the start of the line. For example, this
-     * will return the length of the line for the last run on the line.
-     *
-     * @param run the index of the run, between 0 and countRuns()
-     *
-     * @return the limit of the run
-     *
-     * @throws IllegalStateException if this call is not preceded by a successful
-     *         call to setPara or setLine
-     * @throws IllegalArgumentException if run is not in
-     *         the range 0<=run<countRuns()
-     * @stable ICU 3.8
-     */
-    public int getRunLimit(int run)
-    {
-        verifyValidParaOrLine();
-        BidiLine.getRuns(this);
-
-        // for backward compatibility
-        if (runCount == 1) {
-            return length;
-        }
-
-        getLogicalToVisualRunsMap();
-        int idx = logicalToVisualRunsMap[run];
-        int len = idx == 0 ? runs[idx].limit :
-                                runs[idx].limit - runs[idx-1].limit;
-        return runs[idx].start + len;
-    }
-
-    /**
-     * Return true if the specified text requires bidi analysis. If this returns
-     * false, the text will display left-to-right. Clients can then avoid
-     * constructing a Bidi object. Text in the Arabic Presentation Forms area of
-     * Unicode is presumed to already be shaped and ordered for display, and so
-     * will not cause this method to return true.
-     *
-     * @param text the text containing the characters to test
-     * @param start the start of the range of characters to test
-     * @param limit the limit of the range of characters to test
-     *
-     * @return true if the range of characters requires bidi analysis
-     *
-     * @stable ICU 3.8
-     */
-    public static boolean requiresBidi(char[] text,
-            int start,
-            int limit)
-    {
-        final int RTLMask = (1 << R |
-                1 << AL |
-                1 << RLE |
-                1 << RLO |
-                1 << AN);
-
-        if (0 > start || start > limit || limit > text.length) {
-            throw new IllegalArgumentException("Value start " + start +
-                      " is out of range 0 to " + limit);
-        }
-
-        for (int i = start; i < limit; ++i) {
-            if (Character.isHighSurrogate(text[i]) && i < (limit-1) &&
-                Character.isLowSurrogate(text[i+1])) {
-                if (((1 << UCharacter.getDirection(Character.codePointAt(text, i))) & RTLMask) != 0) {
-                    return true;
-                }
-            } else if (((1 << UCharacter.getDirection(text[i])) & RTLMask) != 0) {
-                return true;
-            }
-        }
-
-        return false;
-    }
-
-    /**
-     * Reorder the objects in the array into visual order based on their levels.
-     * This is a utility method to use when you have a collection of objects
-     * representing runs of text in logical order, each run containing text at a
-     * single level. The elements at index from
-     * objectStart up to objectStart + count in the
-     * objects array will be reordered into visual order assuming
-     * each run of text has the level indicated by the corresponding element in
-     * the levels array (at index - objectStart + levelStart).
-     *
-     * @param levels an array representing the bidi level of each object
-     * @param levelStart the start position in the levels array
-     * @param objects the array of objects to be reordered into visual order
-     * @param objectStart the start position in the objects array
-     * @param count the number of objects to reorder
-     * @stable ICU 3.8
-     */
-    public static void reorderVisually(byte[] levels,
-            int levelStart,
-            Object[] objects,
-            int objectStart,
-            int count)
-    {
-        // for backward compatibility
-        if (0 > levelStart || levels.length <= levelStart) {
-          throw new IllegalArgumentException("Value levelStart " +
-                      levelStart + " is out of range 0 to " +
-                      (levels.length-1));
-        }
-        if (0 > objectStart || objects.length <= objectStart) {
-            throw new IllegalArgumentException("Value objectStart " +
-                      levelStart + " is out of range 0 to " +
-                      (objects.length-1));
-        }
-        if (0 > count || objects.length < (objectStart+count)) {
-            throw new IllegalArgumentException("Value count " +
-                      levelStart + " is out of range 0 to " +
-                      (objects.length - objectStart));
-        }
-
-        byte[] reorderLevels = new byte[count];
-        System.arraycopy(levels, levelStart, reorderLevels, 0, count);
-        int[] indexMap = reorderVisual(reorderLevels);
-        Object[] temp = new Object[count];
-        System.arraycopy(objects, objectStart, temp, 0, count);
-        for (int i = 0; i < count; ++i) {
-            objects[objectStart + i] = temp[indexMap[i]];
-        }
-    }
-
-    /**
-     * Take a Bidi object containing the reordering
-     * information for a piece of text (one or more paragraphs) set by
-     * setPara() or for a line of text set by setLine()
-     * and return a string containing the reordered text.
-     *
-     * 
The text may have been aliased (only a reference was stored
-     * without copying the contents), thus it must not have been modified
-     * since the setPara() call.

-     *
-     * This method preserves the integrity of characters with multiple
-     * code units and (optionally) combining characters.
-     * Characters in RTL runs can be replaced by mirror-image characters
-     * in the returned string. Note that "real" mirroring has to be done in a
-     * rendering engine by glyph selection and that for many "mirrored"
-     * characters there are no Unicode characters as mirror-image equivalents.
-     * There are also options to insert or remove Bidi control
-     * characters; see the descriptions of the return value and the
-     * options parameter, and of the option bit flags.
-     *
-     * @param options A bit set of options for the reordering that control
-     *                how the reordered text is written.
-     *                The options include mirroring the characters on a code
-     *                point basis and inserting LRM characters, which is used
-     *                especially for transforming visually stored text
-     *                to logically stored text (although this is still an
-     *                imperfect implementation of an "inverse Bidi" algorithm
-     *                because it uses the "forward Bidi" algorithm at its core).
-     *                The available options are:
-     *                DO_MIRRORING,
-     *                INSERT_LRM_FOR_NUMERIC,
-     *                KEEP_BASE_COMBINING,
-     *                OUTPUT_REVERSE,
-     *                REMOVE_BIDI_CONTROLS,
-     *                STREAMING
-     *
-     * @return The reordered text.
-     *         If the INSERT_LRM_FOR_NUMERIC option is set, then
-     *         the length of the returned string could be as large as
-     *         getLength()+2*countRuns().

-     *         If the REMOVE_BIDI_CONTROLS option is set, then the
-     *         length of the returned string may be less than
-     *         getLength().

-     *         If none of these options is set, then the length of the returned
-     *         string will be exactly getProcessedLength().
-     *
-     * @throws IllegalStateException if this call is not preceded by a successful
-     *         call to setPara or setLine
-     *
-     * @see #DO_MIRRORING
-     * @see #INSERT_LRM_FOR_NUMERIC
-     * @see #KEEP_BASE_COMBINING
-     * @see #OUTPUT_REVERSE
-     * @see #REMOVE_BIDI_CONTROLS
-     * @see #OPTION_STREAMING
-     * @see #getProcessedLength
-     * @stable ICU 3.8
-     */
-    public String writeReordered(int options)
-    {
-        verifyValidParaOrLine();
-        if (length == 0) {
-            /* nothing to do */
-            return "";
-        }
-        return BidiWriter.writeReordered(this, options);
-    }
-
-    /**
-     * Display the bidi internal state, used in debugging.
-     */
-    public String toString() {
-        StringBuilder buf = new StringBuilder(getClass().getName());
-
-        buf.append("[dir: ");
-        buf.append(direction);
-        buf.append(" baselevel: ");
-        buf.append(paraLevel);
-        buf.append(" length: ");
-        buf.append(length);
-        buf.append(" runs: ");
-        if (levels == null) {
-            buf.append("none");
-        } else {
-            buf.append('[');
-            buf.append(levels[0]);
-            for (int i = 1; i < levels.length; i++) {
-                buf.append(' ');
-                buf.append(levels[i]);
-            }
-            buf.append(']');
-        }
-        buf.append(" text: [0x");
-        buf.append(Integer.toHexString(text[0]));
-        for (int i = 1; i < text.length; i++) {
-            buf.append(" 0x");
-            buf.append(Integer.toHexString(text[i]));
-        }
-        buf.append("]]");
-
-        return buf.toString();
-    }
-
-    /**
-     * A class that provides access to constants defined by
-     * java.awt.font.TextAttribute without creating a static dependency.
-     */
-    private static class TextAttributeConstants {
-        // Make sure to load the AWT's TextAttribute class before using the constants, if any.
-        static {
-            try {
-                Class.forName("java.awt.font.TextAttribute", true, null);
-            } catch (ClassNotFoundException e) {}
-        }
-        static final JavaAWTFontAccess jafa = SharedSecrets.getJavaAWTFontAccess();
-
-        /**
-         * TextAttribute instances (or a fake Attribute type if
-         * java.awt.font.TextAttribute is not present)
-         */
-        static final AttributedCharacterIterator.Attribute RUN_DIRECTION =
-            getTextAttribute("RUN_DIRECTION");
-        static final AttributedCharacterIterator.Attribute NUMERIC_SHAPING =
-            getTextAttribute("NUMERIC_SHAPING");
-        static final AttributedCharacterIterator.Attribute BIDI_EMBEDDING =
-            getTextAttribute("BIDI_EMBEDDING");
-
-        /**
-         * TextAttribute.RUN_DIRECTION_LTR
-         */
-        static final Boolean RUN_DIRECTION_LTR = (jafa == null) ?
-            Boolean.FALSE : (Boolean)jafa.getTextAttributeConstant("RUN_DIRECTION_LTR");
-
-        @SuppressWarnings("serial")
-        private static AttributedCharacterIterator.Attribute
-            getTextAttribute(String name)
-        {
-            if (jafa == null) {
-                // fake attribute
-                return new AttributedCharacterIterator.Attribute(name) { };
-            } else {
-                return (AttributedCharacterIterator.Attribute)jafa.getTextAttributeConstant(name);
-            }
-        }
-    }
-
-    /**
-     * A class that provides access to java.awt.font.NumericShaper without
-     * creating a static dependency.
-     */
-    private static class NumericShapings {
-        // Make sure to load the AWT's NumericShaper class before calling shape, if any.
-        static {
-            try {
-                Class.forName("java.awt.font.NumericShaper", true, null);
-            } catch (ClassNotFoundException e) {}
-        }
-        static final JavaAWTFontAccess jafa = SharedSecrets.getJavaAWTFontAccess();
-
-        /**
-         * Invokes NumericShaping shape(text,start,count) method.
-         */
-        static void shape(Object shaper, char[] text, int start, int count) {
-            if (jafa != null) {
-                jafa.shape(shaper, text, start, count);
-            }
-        }
-    }
-
-}
--- /dev/null	2020-01-10 15:57:52.000000000 -0800
+++ new/src/java.base/share/classes/jdk/internal/icu/text/BidiBase.java	2020-01-10 15:57:52.000000000 -0800
@@ -0,0 +1,4780 @@
+/*
+ * Copyright (c) 2009, 2020, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.  Oracle designates this
+ * particular file as subject to the "Classpath" exception as provided
+ * by Oracle in the LICENSE file that accompanied this code.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+/*
+*******************************************************************************
+*   Copyright (C) 2001-2014, International Business Machines
+*   Corporation and others.  All Rights Reserved.
+*******************************************************************************
+*/
+
+/* FOOD FOR THOUGHT: currently the reordering modes are a mixture of
+ * algorithm for direct BiDi, algorithm for inverse Bidi and the bizarre
+ * concept of RUNS_ONLY which is a double operation.
+ * It could be advantageous to divide this into 3 concepts:
+ * a) Operation: direct / inverse / RUNS_ONLY
+ * b) Direct algorithm: default / NUMBERS_SPECIAL / GROUP_NUMBERS_WITH_L
+ * c) Inverse algorithm: default / INVERSE_LIKE_DIRECT / NUMBERS_SPECIAL
+ * This would allow combinations not possible today like RUNS_ONLY with
+ * NUMBERS_SPECIAL.
+ * Also allow to set INSERT_MARKS for the direct step of RUNS_ONLY and
+ * REMOVE_CONTROLS for the inverse step.
+ * Not all combinations would be supported, and probably not all do make sense.
+ * This would need to document which ones are supported and what are the
+ * fallbacks for unsupported combinations.
+ */
+
+package jdk.internal.icu.text;
+
+import java.lang.reflect.Array;
+import java.text.AttributedCharacterIterator;
+import java.text.Bidi;
+import java.util.Arrays;
+import jdk.internal.access.JavaAWTFontAccess;
+import jdk.internal.access.SharedSecrets;
+import jdk.internal.icu.lang.UCharacter;
+import jdk.internal.icu.impl.UBiDiProps;
+
+/**
+ *
+ * 
Bidi algorithm for ICU

+ *
+ * This is an implementation of the Unicode Bidirectional Algorithm. The
+ * algorithm is defined in the Unicode Standard Annex #9.
+ * 

+ *
+ * Note: Libraries that perform a bidirectional algorithm and reorder strings
+ * accordingly are sometimes called "Storage Layout Engines". ICU's Bidi and
+ * shaping (ArabicShaping) classes can be used at the core of such "Storage
+ * Layout Engines".
+ *
+ * 
General remarks about the API:

+ *
+ * The "limit" of a sequence of characters is the position just after
+ * their last character, i.e., one more than that position.
+ * 

+ *
+ * Some of the API methods provide access to "runs". Such a
+ * "run" is defined as a sequence of characters that are at the same
+ * embedding level after performing the Bidi algorithm.
+ *
+ * 
Basic concept: paragraph

+ * A piece of text can be divided into several paragraphs by characters
+ * with the Bidi class Block Separator. For handling of
+ * paragraphs, see:
+ * 
    
+ * 
  • {@link #countParagraphs}
+ * 
  • {@link #getParaLevel}
+ * 
  • {@link #getParagraph}
+ * 
  • {@link #getParagraphByIndex}
+ * 

+ *
+ * 
Basic concept: text direction

+ * The direction of a piece of text may be:
+ * 
    
+ * 
  • {@link #LTR}
+ * 
  • {@link #RTL}
+ * 
  • {@link #MIXED}
+ * 
  • {@link #NEUTRAL}
+ * 

+ *
+ * 
Basic concept: levels

+ *
+ * Levels in this API represent embedding levels according to the Unicode
+ * Bidirectional Algorithm.
+ * Their low-order bit (even/odd value) indicates the visual direction.

+ *
+ * Levels can be abstract values when used for the
+ * paraLevel and embeddingLevels
+ * arguments of setPara(); there:
+ * 
    
+ * 
  • the high-order bit of an embeddingLevels[]
+ * value indicates whether the using application is
+ * specifying the level of a character to override whatever the
+ * Bidi implementation would resolve it to.
    • 
+ * 
  • paraLevel can be set to the
+ * pseudo-level values LEVEL_DEFAULT_LTR
+ * and LEVEL_DEFAULT_RTL.
    • 
+ * 

+ *
+ * 
The related constants are not real, valid level values.
+ * DEFAULT_XXX can be used to specify
+ * a default for the paragraph level for
+ * when the setPara() method
+ * shall determine it but there is no
+ * strongly typed character in the input.

+ *
+ * Note that the value for LEVEL_DEFAULT_LTR is even
+ * and the one for LEVEL_DEFAULT_RTL is odd,
+ * just like with normal LTR and RTL level values -
+ * these special values are designed that way. Also, the implementation
+ * assumes that MAX_EXPLICIT_LEVEL is odd.
+ *
+ * 
See Also:
+ * 
    
+ * 
  • {@link #LEVEL_DEFAULT_LTR}
+ * 
  • {@link #LEVEL_DEFAULT_RTL}
+ * 
  • {@link #LEVEL_OVERRIDE}
+ * 
  • {@link #MAX_EXPLICIT_LEVEL}
+ * 
  • {@link #setPara}
+ * 

+ *
+ * 
Basic concept: Reordering Mode

+ * Reordering mode values indicate which variant of the Bidi algorithm to
+ * use.
+ *
+ * 
See Also:
+ * 
    
+ * 
  • {@link #setReorderingMode}
+ * 
  • {@link #REORDER_DEFAULT}
+ * 
  • {@link #REORDER_NUMBERS_SPECIAL}
+ * 
  • {@link #REORDER_GROUP_NUMBERS_WITH_R}
+ * 
  • {@link #REORDER_RUNS_ONLY}
+ * 
  • {@link #REORDER_INVERSE_NUMBERS_AS_L}
+ * 
  • {@link #REORDER_INVERSE_LIKE_DIRECT}
+ * 
  • {@link #REORDER_INVERSE_FOR_NUMBERS_SPECIAL}
+ * 

+ *
+ * 
Basic concept: Reordering Options

+ * Reordering options can be applied during Bidi text transformations.
+ *
+ * 
See Also:
+ * 
    
+ * 
  • {@link #setReorderingOptions}
+ * 
  • {@link #OPTION_DEFAULT}
+ * 
  • {@link #OPTION_INSERT_MARKS}
+ * 
  • {@link #OPTION_REMOVE_CONTROLS}
+ * 
  • {@link #OPTION_STREAMING}
+ * 

+ *
+ *
+ * @author Simon Montagu, Matitiahu Allouche (ported from C code written by Markus W. Scherer)
+ * @stable ICU 3.8
+ *
+ *
+ * 
 Sample code for the ICU Bidi API 

+ *
+ * 
Rendering a paragraph with the ICU Bidi API

+ *
+ * This is (hypothetical) sample code that illustrates how the ICU Bidi API
+ * could be used to render a paragraph of text. Rendering code depends highly on
+ * the graphics system, therefore this sample code must make a lot of
+ * assumptions, which may or may not match any existing graphics system's
+ * properties.
+ *
+ * 

+ * The basic assumptions are:
+ * 

+ * 
    
+ * 
  • Rendering is done from left to right on a horizontal line.
    • 
+ * 
  • A run of single-style, unidirectional text can be rendered at once.
+ * 
    • 
+ * 
  • Such a run of text is passed to the graphics system with characters
+ * (code units) in logical order.
    • 
+ * 
  • The line-breaking algorithm is very complicated and Locale-dependent -
+ * and therefore its implementation omitted from this sample code.
    • 
+ * 

+ *
+ * 
{@code
+ *
+ *  package com.ibm.icu.dev.test.bidi;
+ *
+ *  import com.ibm.icu.text.Bidi;
+ *  import com.ibm.icu.text.BidiRun;
+ *
+ *  public class Sample {
+ *
+ *      static final int styleNormal = 0;
+ *      static final int styleSelected = 1;
+ *      static final int styleBold = 2;
+ *      static final int styleItalics = 4;
+ *      static final int styleSuper=8;
+ *      static final int styleSub = 16;
+ *
+ *      static class StyleRun {
+ *          int limit;
+ *          int style;
+ *
+ *          public StyleRun(int limit, int style) {
+ *              this.limit = limit;
+ *              this.style = style;
+ *          }
+ *      }
+ *
+ *      static class Bounds {
+ *          int start;
+ *          int limit;
+ *
+ *          public Bounds(int start, int limit) {
+ *              this.start = start;
+ *              this.limit = limit;
+ *          }
+ *      }
+ *
+ *      static int getTextWidth(String text, int start, int limit,
+ *                              StyleRun[] styleRuns, int styleRunCount) {
+ *          // simplistic way to compute the width
+ *          return limit - start;
+ *      }
+ *
+ *      // set limit and StyleRun limit for a line
+ *      // from text[start] and from styleRuns[styleRunStart]
+ *      // using Bidi.getLogicalRun(...)
+ *      // returns line width
+ *      static int getLineBreak(String text, Bounds line, Bidi para,
+ *                              StyleRun styleRuns[], Bounds styleRun) {
+ *          // dummy return
+ *          return 0;
+ *      }
+ *
+ *      // render runs on a line sequentially, always from left to right
+ *
+ *      // prepare rendering a new line
+ *      static void startLine(byte textDirection, int lineWidth) {
+ *          System.out.println();
+ *      }
+ *
+ *      // render a run of text and advance to the right by the run width
+ *      // the text[start..limit-1] is always in logical order
+ *      static void renderRun(String text, int start, int limit,
+ *                            byte textDirection, int style) {
+ *      }
+ *
+ *      // We could compute a cross-product
+ *      // from the style runs with the directional runs
+ *      // and then reorder it.
+ *      // Instead, here we iterate over each run type
+ *      // and render the intersections -
+ *      // with shortcuts in simple (and common) cases.
+ *      // renderParagraph() is the main function.
+ *
+ *      // render a directional run with
+ *      // (possibly) multiple style runs intersecting with it
+ *      static void renderDirectionalRun(String text, int start, int limit,
+ *                                       byte direction, StyleRun styleRuns[],
+ *                                       int styleRunCount) {
+ *          int i;
+ *
+ *          // iterate over style runs
+ *          if (direction == Bidi.LTR) {
+ *              int styleLimit;
+ *              for (i = 0; i < styleRunCount; ++i) {
+ *                  styleLimit = styleRuns[i].limit;
+ *                  if (start < styleLimit) {
+ *                      if (styleLimit > limit) {
+ *                          styleLimit = limit;
+ *                      }
+ *                      renderRun(text, start, styleLimit,
+ *                                direction, styleRuns[i].style);
+ *                      if (styleLimit == limit) {
+ *                          break;
+ *                      }
+ *                      start = styleLimit;
+ *                  }
+ *              }
+ *          } else {
+ *              int styleStart;
+ *
+ *              for (i = styleRunCount-1; i >= 0; --i) {
+ *                  if (i > 0) {
+ *                      styleStart = styleRuns[i-1].limit;
+ *                  } else {
+ *                      styleStart = 0;
+ *                  }
+ *                  if (limit >= styleStart) {
+ *                      if (styleStart < start) {
+ *                          styleStart = start;
+ *                      }
+ *                      renderRun(text, styleStart, limit, direction,
+ *                                styleRuns[i].style);
+ *                      if (styleStart == start) {
+ *                          break;
+ *                      }
+ *                      limit = styleStart;
+ *                  }
+ *              }
+ *          }
+ *      }
+ *
+ *      // the line object represents text[start..limit-1]
+ *      static void renderLine(Bidi line, String text, int start, int limit,
+ *                             StyleRun styleRuns[], int styleRunCount) {
+ *          byte direction = line.getDirection();
+ *          if (direction != Bidi.MIXED) {
+ *              // unidirectional
+ *              if (styleRunCount <= 1) {
+ *                  renderRun(text, start, limit, direction, styleRuns[0].style);
+ *              } else {
+ *                  renderDirectionalRun(text, start, limit, direction,
+ *                                       styleRuns, styleRunCount);
+ *              }
+ *          } else {
+ *              // mixed-directional
+ *              int count, i;
+ *              BidiRun run;
+ *
+ *              try {
+ *                  count = line.countRuns();
+ *              } catch (IllegalStateException e) {
+ *                  e.printStackTrace();
+ *                  return;
+ *              }
+ *              if (styleRunCount <= 1) {
+ *                  int style = styleRuns[0].style;
+ *
+ *                  // iterate over directional runs
+ *                  for (i = 0; i < count; ++i) {
+ *                      run = line.getVisualRun(i);
+ *                      renderRun(text, run.getStart(), run.getLimit(),
+ *                                run.getDirection(), style);
+ *                  }
+ *              } else {
+ *                  // iterate over both directional and style runs
+ *                  for (i = 0; i < count; ++i) {
+ *                      run = line.getVisualRun(i);
+ *                      renderDirectionalRun(text, run.getStart(),
+ *                                           run.getLimit(), run.getDirection(),
+ *                                           styleRuns, styleRunCount);
+ *                  }
+ *              }
+ *          }
+ *      }
+ *
+ *      static void renderParagraph(String text, byte textDirection,
+ *                                  StyleRun styleRuns[], int styleRunCount,
+ *                                  int lineWidth) {
+ *          int length = text.length();
+ *          Bidi para = new Bidi();
+ *          try {
+ *              para.setPara(text,
+ *                           textDirection != 0 ? Bidi.LEVEL_DEFAULT_RTL
+ *                                              : Bidi.LEVEL_DEFAULT_LTR,
+ *                           null);
+ *          } catch (Exception e) {
+ *              e.printStackTrace();
+ *              return;
+ *          }
+ *          byte paraLevel = (byte)(1 & para.getParaLevel());
+ *          StyleRun styleRun = new StyleRun(length, styleNormal);
+ *
+ *          if (styleRuns == null || styleRunCount <= 0) {
+ *              styleRuns = new StyleRun[1];
+ *              styleRunCount = 1;
+ *              styleRuns[0] = styleRun;
+ *          }
+ *          // assume styleRuns[styleRunCount-1].limit>=length
+ *
+ *          int width = getTextWidth(text, 0, length, styleRuns, styleRunCount);
+ *          if (width <= lineWidth) {
+ *              // everything fits onto one line
+ *
+ *              // prepare rendering a new line from either left or right
+ *              startLine(paraLevel, width);
+ *
+ *              renderLine(para, text, 0, length, styleRuns, styleRunCount);
+ *          } else {
+ *              // we need to render several lines
+ *              Bidi line = new Bidi(length, 0);
+ *              int start = 0, limit;
+ *              int styleRunStart = 0, styleRunLimit;
+ *
+ *              for (;;) {
+ *                  limit = length;
+ *                  styleRunLimit = styleRunCount;
+ *                  width = getLineBreak(text, new Bounds(start, limit),
+ *                                       para, styleRuns,
+ *                                       new Bounds(styleRunStart, styleRunLimit));
+ *                  try {
+ *                      line = para.setLine(start, limit);
+ *                  } catch (Exception e) {
+ *                      e.printStackTrace();
+ *                      return;
+ *                  }
+ *                  // prepare rendering a new line
+ *                  // from either left or right
+ *                  startLine(paraLevel, width);
+ *
+ *                  if (styleRunStart > 0) {
+ *                      int newRunCount = styleRuns.length - styleRunStart;
+ *                      StyleRun[] newRuns = new StyleRun[newRunCount];
+ *                      System.arraycopy(styleRuns, styleRunStart, newRuns, 0,
+ *                                       newRunCount);
+ *                      renderLine(line, text, start, limit, newRuns,
+ *                                 styleRunLimit - styleRunStart);
+ *                  } else {
+ *                      renderLine(line, text, start, limit, styleRuns,
+ *                                 styleRunLimit - styleRunStart);
+ *                  }
+ *                  if (limit == length) {
+ *                      break;
+ *                  }
+ *                  start = limit;
+ *                  styleRunStart = styleRunLimit - 1;
+ *                  if (start >= styleRuns[styleRunStart].limit) {
+ *                      ++styleRunStart;
+ *                  }
+ *              }
+ *          }
+ *      }
+ *
+ *      public static void main(String[] args)
+ *      {
+ *          renderParagraph("Some Latin text...", Bidi.LTR, null, 0, 80);
+ *          renderParagraph("Some Hebrew text...", Bidi.RTL, null, 0, 60);
+ *      }
+ *  }
+ *
+ * }

+ */
+
+/*
+ * General implementation notes:
+ *
+ * Throughout the implementation, there are comments like (W2) that refer to
+ * rules of the BiDi algorithm, in this example to the second rule of the
+ * resolution of weak types.
+ *
+ * For handling surrogate pairs, where two UChar's form one "abstract" (or UTF-32)
+ * character according to UTF-16, the second UChar gets the directional property of
+ * the entire character assigned, while the first one gets a BN, a boundary
+ * neutral, type, which is ignored by most of the algorithm according to
+ * rule (X9) and the implementation suggestions of the BiDi algorithm.
+ *
+ * Later, adjustWSLevels() will set the level for each BN to that of the
+ * following character (UChar), which results in surrogate pairs getting the
+ * same level on each of their surrogates.
+ *
+ * In a UTF-8 implementation, the same thing could be done: the last byte of
+ * a multi-byte sequence would get the "real" property, while all previous
+ * bytes of that sequence would get BN.
+ *
+ * It is not possible to assign all those parts of a character the same real
+ * property because this would fail in the resolution of weak types with rules
+ * that look at immediately surrounding types.
+ *
+ * As a related topic, this implementation does not remove Boundary Neutral
+ * types from the input, but ignores them wherever this is relevant.
+ * For example, the loop for the resolution of the weak types reads
+ * types until it finds a non-BN.
+ * Also, explicit embedding codes are neither changed into BN nor removed.
+ * They are only treated the same way real BNs are.
+ * As stated before, adjustWSLevels() takes care of them at the end.
+ * For the purpose of conformance, the levels of all these codes
+ * do not matter.
+ *
+ * Note that this implementation modifies the dirProps
+ * after the initial setup, when applying X5c (replace FSI by LRI or RLI),
+ * X6, N0 (replace paired brackets by L or R).
+ *
+ * In this implementation, the resolution of weak types (W1 to W6),
+ * neutrals (N1 and N2), and the assignment of the resolved level (In)
+ * are all done in one single loop, in resolveImplicitLevels().
+ * Changes of dirProp values are done on the fly, without writing
+ * them back to the dirProps array.
+ *
+ *
+ * This implementation contains code that allows to bypass steps of the
+ * algorithm that are not needed on the specific paragraph
+ * in order to speed up the most common cases considerably,
+ * like text that is entirely LTR, or RTL text without numbers.
+ *
+ * Most of this is done by setting a bit for each directional property
+ * in a flags variable and later checking for whether there are
+ * any LTR characters or any RTL characters, or both, whether
+ * there are any explicit embedding codes, etc.
+ *
+ * If the (Xn) steps are performed, then the flags are re-evaluated,
+ * because they will then not contain the embedding codes any more
+ * and will be adjusted for override codes, so that subsequently
+ * more bypassing may be possible than what the initial flags suggested.
+ *
+ * If the text is not mixed-directional, then the
+ * algorithm steps for the weak type resolution are not performed,
+ * and all levels are set to the paragraph level.
+ *
+ * If there are no explicit embedding codes, then the (Xn) steps
+ * are not performed.
+ *
+ * If embedding levels are supplied as a parameter, then all
+ * explicit embedding codes are ignored, and the (Xn) steps
+ * are not performed.
+ *
+ * White Space types could get the level of the run they belong to,
+ * and are checked with a test of (flags&MASK_EMBEDDING) to
+ * consider if the paragraph direction should be considered in
+ * the flags variable.
+ *
+ * If there are no White Space types in the paragraph, then
+ * (L1) is not necessary in adjustWSLevels().
+ */
+
+// Original filename in ICU4J: Bidi.java
+public class BidiBase {
+
+    static class Point {
+        int pos;    /* position in text */
+        int flag;   /* flag for LRM/RLM, before/after */
+    }
+
+    static class InsertPoints {
+        int size;
+        int confirmed;
+        Point[] points = new Point[0];
+    }
+
+    static class Opening {
+        int   position;                 /* position of opening bracket */
+        int   match;                    /* matching char or -position of closing bracket */
+        int   contextPos;               /* position of last strong char found before opening */
+        short flags;                    /* bits for L or R/AL found within the pair */
+        byte  contextDir;               /* L or R according to last strong char before opening */
+    }
+
+    static class IsoRun {
+        int   contextPos;               /* position of char determining context */
+        short start;                    /* index of first opening entry for this run */
+        short limit;                    /* index after last opening entry for this run */
+        byte  level;                    /* level of this run */
+        byte  lastStrong;               /* bidi class of last strong char found in this run */
+        byte  lastBase;                 /* bidi class of last base char found in this run */
+        byte  contextDir;               /* L or R to use as context for following openings */
+    }
+
+    static class BracketData {
+        Opening[] openings = new Opening[SIMPLE_PARAS_COUNT];
+        int   isoRunLast;               /* index of last used entry */
+        /* array of nested isolated sequence entries; can never excess UBIDI_MAX_EXPLICIT_LEVEL
+           + 1 for index 0, + 1 for before the first isolated sequence */
+        IsoRun[]  isoRuns = new IsoRun[MAX_EXPLICIT_LEVEL+2];
+        boolean   isNumbersSpecial;     /*reordering mode for NUMBERS_SPECIAL */
+    }
+
+    static class Isolate {
+        int   startON;
+        int   start1;
+        short stateImp;
+        short state;
+    }
+
+    /** Paragraph level setting

+     *
+     * Constant indicating that the base direction depends on the first strong
+     * directional character in the text according to the Unicode Bidirectional
+     * Algorithm. If no strong directional character is present,
+     * then set the paragraph level to 0 (left-to-right).

+     *
+     * If this value is used in conjunction with reordering modes
+     * REORDER_INVERSE_LIKE_DIRECT or
+     * REORDER_INVERSE_FOR_NUMBERS_SPECIAL, the text to reorder
+     * is assumed to be visual LTR, and the text after reordering is required
+     * to be the corresponding logical string with appropriate contextual
+     * direction. The direction of the result string will be RTL if either
+     * the rightmost or leftmost strong character of the source text is RTL
+     * or Arabic Letter, the direction will be LTR otherwise.

+     *
+     * If reordering option OPTION_INSERT_MARKS is set, an RLM may
+     * be added at the beginning of the result string to ensure round trip
+     * (that the result string, when reordered back to visual, will produce
+     * the original source text).
+     * @see #REORDER_INVERSE_LIKE_DIRECT
+     * @see #REORDER_INVERSE_FOR_NUMBERS_SPECIAL
+     * @stable ICU 3.8
+     */
+    public static final byte LEVEL_DEFAULT_LTR = (byte)0x7e;
+
+    /** Paragraph level setting

+     *
+     * Constant indicating that the base direction depends on the first strong
+     * directional character in the text according to the Unicode Bidirectional
+     * Algorithm. If no strong directional character is present,
+     * then set the paragraph level to 1 (right-to-left).

+     *
+     * If this value is used in conjunction with reordering modes
+     * REORDER_INVERSE_LIKE_DIRECT or
+     * REORDER_INVERSE_FOR_NUMBERS_SPECIAL, the text to reorder
+     * is assumed to be visual LTR, and the text after reordering is required
+     * to be the corresponding logical string with appropriate contextual
+     * direction. The direction of the result string will be RTL if either
+     * the rightmost or leftmost strong character of the source text is RTL
+     * or Arabic Letter, or if the text contains no strong character;
+     * the direction will be LTR otherwise.

+     *
+     * If reordering option OPTION_INSERT_MARKS is set, an RLM may
+     * be added at the beginning of the result string to ensure round trip
+     * (that the result string, when reordered back to visual, will produce
+     * the original source text).
+     * @see #REORDER_INVERSE_LIKE_DIRECT
+     * @see #REORDER_INVERSE_FOR_NUMBERS_SPECIAL
+     * @stable ICU 3.8
+     */
+    public static final byte LEVEL_DEFAULT_RTL = (byte)0x7f;
+
+    /**
+     * Maximum explicit embedding level.
+     * (The maximum resolved level can be up to MAX_EXPLICIT_LEVEL+1).
+     * @stable ICU 3.8
+     */
+    public static final byte MAX_EXPLICIT_LEVEL = 125;
+
+    /**
+     * Bit flag for level input.
+     * Overrides directional properties.
+     * @stable ICU 3.8
+     */
+    public static final byte LEVEL_OVERRIDE = (byte)0x80;
+
+    /**
+     * Special value which can be returned by the mapping methods when a
+     * logical index has no corresponding visual index or vice-versa. This may
+     * happen for the logical-to-visual mapping of a Bidi control when option
+     * OPTION_REMOVE_CONTROLS is
+     * specified. This can also happen for the visual-to-logical mapping of a
+     * Bidi mark (LRM or RLM) inserted by option
+     * OPTION_INSERT_MARKS.
+     * @see #getVisualIndex
+     * @see #getVisualMap
+     * @see #getLogicalIndex
+     * @see #getLogicalMap
+     * @see #OPTION_INSERT_MARKS
+     * @see #OPTION_REMOVE_CONTROLS
+     * @stable ICU 3.8
+     */
+    public static final int MAP_NOWHERE = -1;
+
+    /**
+     * Left-to-right text.
+     * 
    
+     * 
  • As return value for getDirection(), it means
+     *     that the source string contains no right-to-left characters, or
+     *     that the source string is empty and the paragraph level is even.
+     * 
  • As return value for getBaseDirection(), it
+     *     means that the first strong character of the source string has
+     *     a left-to-right direction.
+     * 

+     * @stable ICU 3.8
+     */
+    public static final byte LTR = 0;
+
+    /**
+     * Right-to-left text.
+     * 
    
+     * 
  • As return value for getDirection(), it means
+     *     that the source string contains no left-to-right characters, or
+     *     that the source string is empty and the paragraph level is odd.
+     * 
  • As return value for getBaseDirection(), it
+     *     means that the first strong character of the source string has
+     *     a right-to-left direction.
+     * 

+     * @stable ICU 3.8
+     */
+    public static final byte RTL = 1;
+
+    /**
+     * Mixed-directional text.
+     * 
As return value for getDirection(), it means
+     *    that the source string contains both left-to-right and
+     *    right-to-left characters.
+     * @stable ICU 3.8
+     */
+    public static final byte MIXED = 2;
+
+    /**
+     * option bit for writeReordered():
+     * keep combining characters after their base characters in RTL runs
+     *
+     * @see #writeReordered
+     * @stable ICU 3.8
+     */
+    public static final short KEEP_BASE_COMBINING = 1;
+
+    /**
+     * option bit for writeReordered():
+     * replace characters with the "mirrored" property in RTL runs
+     * by their mirror-image mappings
+     *
+     * @see #writeReordered
+     * @stable ICU 3.8
+     */
+    public static final short DO_MIRRORING = 2;
+
+    /**
+     * option bit for writeReordered():
+     * surround the run with LRMs if necessary;
+     * this is part of the approximate "inverse Bidi" algorithm
+     *
+     * 
This option does not imply corresponding adjustment of the index
+     * mappings.

+     *
+     * @see #setInverse
+     * @see #writeReordered
+     * @stable ICU 3.8
+     */
+    public static final short INSERT_LRM_FOR_NUMERIC = 4;
+
+    /**
+     * option bit for writeReordered():
+     * remove Bidi control characters
+     * (this does not affect INSERT_LRM_FOR_NUMERIC)
+     *
+     * 
This option does not imply corresponding adjustment of the index
+     * mappings.

+     *
+     * @see #writeReordered
+     * @see #INSERT_LRM_FOR_NUMERIC
+     * @stable ICU 3.8
+     */
+    public static final short REMOVE_BIDI_CONTROLS = 8;
+
+    /**
+     * option bit for writeReordered():
+     * write the output in reverse order
+     *
+     * 
This has the same effect as calling writeReordered()
+     * first without this option, and then calling
+     * writeReverse() without mirroring.
+     * Doing this in the same step is faster and avoids a temporary buffer.
+     * An example for using this option is output to a character terminal that
+     * is designed for RTL scripts and stores text in reverse order.

+     *
+     * @see #writeReordered
+     * @stable ICU 3.8
+     */
+    public static final short OUTPUT_REVERSE = 16;
+
+    /** Reordering mode: Regular Logical to Visual Bidi algorithm according to Unicode.
+     * @see #setReorderingMode
+     * @stable ICU 3.8
+     */
+    private static final short REORDER_DEFAULT = 0;
+
+    /** Reordering mode: Logical to Visual algorithm which handles numbers in
+     * a way which mimicks the behavior of Windows XP.
+     * @see #setReorderingMode
+     * @stable ICU 3.8
+     */
+    private static final short REORDER_NUMBERS_SPECIAL = 1;
+
+    /** Reordering mode: Logical to Visual algorithm grouping numbers with
+     * adjacent R characters (reversible algorithm).
+     * @see #setReorderingMode
+     * @stable ICU 3.8
+     */
+    private static final short REORDER_GROUP_NUMBERS_WITH_R = 2;
+
+    /** Reordering mode: Reorder runs only to transform a Logical LTR string
+     * to the logical RTL string with the same display, or vice-versa.

+     * If this mode is set together with option
+     * OPTION_INSERT_MARKS, some Bidi controls in the source
+     * text may be removed and other controls may be added to produce the
+     * minimum combination which has the required display.
+     * @see #OPTION_INSERT_MARKS
+     * @see #setReorderingMode
+     * @stable ICU 3.8
+     */
+    static final short REORDER_RUNS_ONLY = 3;
+
+    /** Reordering mode: Visual to Logical algorithm which handles numbers
+     * like L (same algorithm as selected by setInverse(true).
+     * @see #setInverse
+     * @see #setReorderingMode
+     * @stable ICU 3.8
+     */
+    static final short REORDER_INVERSE_NUMBERS_AS_L = 4;
+
+    /** Reordering mode: Visual to Logical algorithm equivalent to the regular
+     * Logical to Visual algorithm.
+     * @see #setReorderingMode
+     * @stable ICU 3.8
+     */
+    static final short REORDER_INVERSE_LIKE_DIRECT = 5;
+
+    /** Reordering mode: Inverse Bidi (Visual to Logical) algorithm for the
+     * REORDER_NUMBERS_SPECIAL Bidi algorithm.
+     * @see #setReorderingMode
+     * @stable ICU 3.8
+     */
+    static final short REORDER_INVERSE_FOR_NUMBERS_SPECIAL = 6;
+
+    /* Reordering mode values must be ordered so that all the regular logical to
+     * visual modes come first, and all inverse Bidi modes come last.
+     */
+    private static final short REORDER_LAST_LOGICAL_TO_VISUAL =
+            REORDER_NUMBERS_SPECIAL;
+
+    /**
+     * Option bit for setReorderingOptions:
+     * insert Bidi marks (LRM or RLM) when needed to ensure correct result of
+     * a reordering to a Logical order
+     *
+     * 
This option must be set or reset before calling
+     * setPara.

+     *
+     * 
This option is significant only with reordering modes which generate
+     * a result with Logical order, specifically.

+     * 
    
+     *   
  • REORDER_RUNS_ONLY
    • 
+     *   
  • REORDER_INVERSE_NUMBERS_AS_L
    • 
+     *   
  • REORDER_INVERSE_LIKE_DIRECT
    • 
+     *   
  • REORDER_INVERSE_FOR_NUMBERS_SPECIAL
    • 
+     * 

+     *
+     * 
If this option is set in conjunction with reordering mode
+     * REORDER_INVERSE_NUMBERS_AS_L or with calling
+     * setInverse(true), it implies option
+     * INSERT_LRM_FOR_NUMERIC in calls to method
+     * writeReordered().

+     *
+     * 
For other reordering modes, a minimum number of LRM or RLM characters
+     * will be added to the source text after reordering it so as to ensure
+     * round trip, i.e. when applying the inverse reordering mode on the
+     * resulting logical text with removal of Bidi marks
+     * (option OPTION_REMOVE_CONTROLS set before calling
+     * setPara() or option
+     * REMOVE_BIDI_CONTROLS in
+     * writeReordered), the result will be identical to the
+     * source text in the first transformation.
+     *
+     * 
This option will be ignored if specified together with option
+     * OPTION_REMOVE_CONTROLS. It inhibits option
+     * REMOVE_BIDI_CONTROLS in calls to method
+     * writeReordered() and it implies option
+     * INSERT_LRM_FOR_NUMERIC in calls to method
+     * writeReordered() if the reordering mode is
+     * REORDER_INVERSE_NUMBERS_AS_L.

+     *
+     * @see #setReorderingMode
+     * @see #setReorderingOptions
+     * @see #INSERT_LRM_FOR_NUMERIC
+     * @see #REMOVE_BIDI_CONTROLS
+     * @see #OPTION_REMOVE_CONTROLS
+     * @see #REORDER_RUNS_ONLY
+     * @see #REORDER_INVERSE_NUMBERS_AS_L
+     * @see #REORDER_INVERSE_LIKE_DIRECT
+     * @see #REORDER_INVERSE_FOR_NUMBERS_SPECIAL
+     * @stable ICU 3.8
+     */
+    static final int OPTION_INSERT_MARKS = 1;
+
+    /**
+     * Option bit for setReorderingOptions:
+     * remove Bidi control characters
+     *
+     * 
This option must be set or reset before calling
+     * setPara.

+     *
+     * 
This option nullifies option
+     * OPTION_INSERT_MARKS. It inhibits option
+     * INSERT_LRM_FOR_NUMERIC in calls to method
+     * writeReordered() and it implies option
+     * REMOVE_BIDI_CONTROLS in calls to that method.

+     *
+     * @see #setReorderingMode
+     * @see #setReorderingOptions
+     * @see #OPTION_INSERT_MARKS
+     * @see #INSERT_LRM_FOR_NUMERIC
+     * @see #REMOVE_BIDI_CONTROLS
+     * @stable ICU 3.8
+     */
+    static final int OPTION_REMOVE_CONTROLS = 2;
+
+    /**
+     * Option bit for setReorderingOptions:
+     * process the output as part of a stream to be continued
+     *
+     * 
This option must be set or reset before calling
+     * setPara.

+     *
+     * 
This option specifies that the caller is interested in processing
+     * large text object in parts. The results of the successive calls are
+     * expected to be concatenated by the caller. Only the call for the last
+     * part will have this option bit off.

+     *
+     * 
When this option bit is on, setPara() may process
+     * less than the full source text in order to truncate the text at a
+     * meaningful boundary. The caller should call
+     * getProcessedLength() immediately after calling
+     * setPara() in order to determine how much of the source
+     * text has been processed. Source text beyond that length should be
+     * resubmitted in following calls to setPara. The
+     * processed length may be less than the length of the source text if a
+     * character preceding the last character of the source text constitutes a
+     * reasonable boundary (like a block separator) for text to be continued.

+     * If the last character of the source text constitutes a reasonable
+     * boundary, the whole text will be processed at once.

+     * If nowhere in the source text there exists
+     * such a reasonable boundary, the processed length will be zero.

+     * The caller should check for such an occurrence and do one of the following:
+     * 
  • submit a larger amount of text with a better chance to include
+     *         a reasonable boundary.
    • 
+     *     
  • resubmit the same text after turning off option
+     *         OPTION_STREAMING.

+     * In all cases, this option should be turned off before processing the last
+     * part of the text.

+     *
+     * 
When the OPTION_STREAMING option is used, it is
+     * recommended to call orderParagraphsLTR(true) before calling
+     * setPara() so that later paragraphs may be concatenated to
+     * previous paragraphs on the right.
+     * 

+     *
+     * @see #setReorderingMode
+     * @see #setReorderingOptions
+     * @see #getProcessedLength
+     * @stable ICU 3.8
+     */
+    private static final int OPTION_STREAMING = 4;
+
+    /*
+     *   Comparing the description of the Bidi algorithm with this implementation
+     *   is easier with the same names for the Bidi types in the code as there.
+     *   See UCharacterDirection
+     */
+    /* private */ static final byte L   = 0;
+    private static final byte R   = 1;
+    private static final byte EN  = 2;
+    private static final byte ES  = 3;
+    private static final byte ET  = 4;
+    private static final byte AN  = 5;
+    private static final byte CS  = 6;
+    static final byte B   = 7;
+    private static final byte S   = 8;
+    private static final byte WS  = 9;
+    private static final byte ON  = 10;
+    private static final byte LRE = 11;
+    private static final byte LRO = 12;
+    private static final byte AL  = 13;
+    private static final byte RLE = 14;
+    private static final byte RLO = 15;
+    private static final byte PDF = 16;
+    private static final byte NSM = 17;
+    private static final byte BN  = 18;
+    private static final byte FSI = 19;
+    private static final byte LRI = 20;
+    private static final byte RLI = 21;
+    private static final byte PDI = 22;
+    private static final byte ENL = PDI + 1;    /* EN after W7 */
+    private static final byte ENR = ENL + 1;    /* EN not subject to W7 */
+
+    // Number of directional types
+    private static final int CHAR_DIRECTION_COUNT       = 23;
+
+    /**
+     * Enumerated property Bidi_Paired_Bracket_Type (new in Unicode 6.3).
+     * Used in UAX #9: Unicode Bidirectional Algorithm
+     * (http://www.unicode.org/reports/tr9/)
+     * Returns UCharacter.BidiPairedBracketType values.
+     * @stable ICU 52
+     */
+    public static final int BIDI_PAIRED_BRACKET_TYPE = 0x1015;
+
+    /**
+     * Bidi Paired Bracket Type constants.
+     *
+     * @see UProperty#BIDI_PAIRED_BRACKET_TYPE
+     * @stable ICU 52
+     */
+    public static interface BidiPairedBracketType {
+        /**
+         * Not a paired bracket.
+         * @stable ICU 52
+         */
+        public static final int NONE = 0;
+        /**
+         * Open paired bracket.
+         * @stable ICU 52
+         */
+        public static final int OPEN = 1;
+        /**
+         * Close paired bracket.
+         * @stable ICU 52
+         */
+        public static final int CLOSE = 2;
+        /**
+         * @stable ICU 52
+         */
+        public static final int COUNT = 3;
+    }
+
+    /* number of paras entries allocated initially */
+    static final int SIMPLE_PARAS_COUNT = 10;
+
+    private static final char CR = '\r';
+    private static final char LF = '\n';
+
+    static final int LRM_BEFORE = 1;
+    static final int LRM_AFTER = 2;
+    static final int RLM_BEFORE = 4;
+    static final int RLM_AFTER = 8;
+
+    /* flags for Opening.flags */
+    static final byte FOUND_L = (byte)DirPropFlag(L);
+    static final byte FOUND_R = (byte)DirPropFlag(R);
+
+    /*
+     * The following bit is used for the directional isolate status.
+     * Stack entries corresponding to isolate sequences are greater than ISOLATE.
+     */
+    static final int ISOLATE = 0x0100;
+
+    /*
+     * reference to parent paragraph object (reference to self if this object is
+     * a paragraph object); set to null in a newly opened object; set to a
+     * real value after a successful execution of setPara or setLine
+     */
+    BidiBase            paraBidi;
+
+    final UBiDiProps    bdp;
+
+    /* character array representing the current text */
+    char[]              text;
+
+    /* length of the current text */
+    int                 originalLength;
+
+    /* if the option OPTION_STREAMING is set, this is the length of
+     * text actually processed by setPara, which may be shorter
+     * than the original length. Otherwise, it is identical to the original
+     * length.
+     */
+    public int                 length;
+
+    /* if option OPTION_REMOVE_CONTROLS is set, and/or Bidi
+     * marks are allowed to be inserted in one of the reordering modes, the
+     * length of the result string may be different from the processed length.
+     */
+    int                 resultLength;
+
+    /* indicators for whether memory may be allocated after construction */
+    boolean             mayAllocateText;
+    boolean             mayAllocateRuns;
+
+    /* arrays with one value per text-character */
+    byte[]              dirPropsMemory = new byte[1];
+    byte[]              levelsMemory = new byte[1];
+    byte[]              dirProps;
+    byte[]              levels;
+
+    /* are we performing an approximation of the "inverse Bidi" algorithm? */
+    boolean             isInverse;
+
+    /* are we using the basic algorithm or its variation? */
+    int                 reorderingMode;
+
+    /* bitmask for reordering options */
+    int                 reorderingOptions;
+
+    /* must block separators receive level 0? */
+    boolean             orderParagraphsLTR;
+
+    /* the paragraph level */
+    byte                paraLevel;
+
+    /* original paraLevel when contextual */
+    /* must be one of DEFAULT_xxx or 0 if not contextual */
+    byte                defaultParaLevel;
+
+    /* the following is set in setPara, used in processPropertySeq */
+
+    ImpTabPair          impTabPair;  /* reference to levels state table pair */
+
+    /* the overall paragraph or line directionality*/
+    byte                direction;
+
+    /* flags is a bit set for which directional properties are in the text */
+    int                 flags;
+
+    /* lastArabicPos is index to the last AL in the text, -1 if none */
+    int                 lastArabicPos;
+
+    /* characters after trailingWSStart are WS and are */
+    /* implicitly at the paraLevel (rule (L1)) - levels may not reflect that */
+    int                 trailingWSStart;
+
+    /* fields for paragraph handling, set in getDirProps() */
+    int                 paraCount;
+    int[]               paras_limit = new int[SIMPLE_PARAS_COUNT];
+    byte[]              paras_level = new byte[SIMPLE_PARAS_COUNT];
+
+    /* fields for line reordering */
+    int                 runCount;     /* ==-1: runs not set up yet */
+    BidiRun[]           runsMemory = new BidiRun[0];
+    BidiRun[]           runs;
+
+    /* for non-mixed text, we only need a tiny array of runs (no allocation) */
+    BidiRun[]           simpleRuns = {new BidiRun()};
+
+    /* fields for managing isolate sequences */
+    Isolate[]           isolates;
+
+    /* maximum or current nesting depth of isolate sequences */
+    /* Within resolveExplicitLevels() and checkExplicitLevels(), this is the maximal
+       nesting encountered.
+       Within resolveImplicitLevels(), this is the index of the current isolates
+       stack entry. */
+    int                 isolateCount;
+
+    /* mapping of runs in logical order to visual order */
+    int[]               logicalToVisualRunsMap;
+    /* flag to indicate that the map has been updated */
+    boolean             isGoodLogicalToVisualRunsMap;
+
+    /* for inverse Bidi with insertion of directional marks */
+    InsertPoints        insertPoints = new InsertPoints();
+
+    /* for option OPTION_REMOVE_CONTROLS */
+    int                 controlCount;
+
+    /*
+     * Sometimes, bit values are more appropriate
+     * to deal with directionality properties.
+     * Abbreviations in these method names refer to names
+     * used in the Bidi algorithm.
+     */
+    static int DirPropFlag(byte dir) {
+        return (1 << dir);
+    }
+
+    boolean testDirPropFlagAt(int flag, int index) {
+        return ((DirPropFlag(dirProps[index]) & flag) != 0);
+    }
+
+    static final int DirPropFlagMultiRuns = DirPropFlag((byte)31);
+
+    /* to avoid some conditional statements, use tiny constant arrays */
+    static final int DirPropFlagLR[] = { DirPropFlag(L), DirPropFlag(R) };
+    static final int DirPropFlagE[] = { DirPropFlag(LRE), DirPropFlag(RLE) };
+    static final int DirPropFlagO[] = { DirPropFlag(LRO), DirPropFlag(RLO) };
+
+    static final int DirPropFlagLR(byte level) { return DirPropFlagLR[level & 1]; }
+    static final int DirPropFlagE(byte level)  { return DirPropFlagE[level & 1]; }
+    static final int DirPropFlagO(byte level)  { return DirPropFlagO[level & 1]; }
+    static final byte DirFromStrong(byte strong) { return strong == L ? L : R; }
+    static final byte NoOverride(byte level) { return (byte)(level & ~LEVEL_OVERRIDE); }
+
+    /*  are there any characters that are LTR or RTL? */
+    static final int MASK_LTR =
+        DirPropFlag(L)|DirPropFlag(EN)|DirPropFlag(ENL)|DirPropFlag(ENR)|DirPropFlag(AN)|DirPropFlag(LRE)|DirPropFlag(LRO)|DirPropFlag(LRI);
+    static final int MASK_RTL = DirPropFlag(R)|DirPropFlag(AL)|DirPropFlag(RLE)|DirPropFlag(RLO)|DirPropFlag(RLI);
+
+    static final int MASK_R_AL = DirPropFlag(R)|DirPropFlag(AL);
+
+    /* explicit embedding codes */
+    private static final int MASK_EXPLICIT = DirPropFlag(LRE)|DirPropFlag(LRO)|DirPropFlag(RLE)|DirPropFlag(RLO)|DirPropFlag(PDF);
+    private static final int MASK_BN_EXPLICIT = DirPropFlag(BN)|MASK_EXPLICIT;
+
+    /* explicit isolate codes */
+    private static final int MASK_ISO = DirPropFlag(LRI)|DirPropFlag(RLI)|DirPropFlag(FSI)|DirPropFlag(PDI);
+
+    /* paragraph and segment separators */
+    private static final int MASK_B_S = DirPropFlag(B)|DirPropFlag(S);
+
+    /* all types that are counted as White Space or Neutral in some steps */
+    static final int MASK_WS = MASK_B_S|DirPropFlag(WS)|MASK_BN_EXPLICIT|MASK_ISO;
+
+    /* types that are neutrals or could becomes neutrals in (Wn) */
+    private static final int MASK_POSSIBLE_N = DirPropFlag(ON)|DirPropFlag(CS)|DirPropFlag(ES)|DirPropFlag(ET)|MASK_WS;
+
+    /*
+     * These types may be changed to "e",
+     * the embedding type (L or R) of the run,
+     * in the Bidi algorithm (N2)
+     */
+    private static final int MASK_EMBEDDING = DirPropFlag(NSM)|MASK_POSSIBLE_N;
+
+    /*
+     *  the dirProp's L and R are defined to 0 and 1 values in UCharacterDirection.java
+     */
+    private static byte GetLRFromLevel(byte level)
+    {
+        return (byte)(level & 1);
+    }
+
+    private static boolean IsDefaultLevel(byte level)
+    {
+        return ((level & LEVEL_DEFAULT_LTR) == LEVEL_DEFAULT_LTR);
+    }
+
+    static boolean IsBidiControlChar(int c)
+    {
+        /* check for range 0x200c to 0x200f (ZWNJ, ZWJ, LRM, RLM) or
+                           0x202a to 0x202e (LRE, RLE, PDF, LRO, RLO) */
+        return (((c & 0xfffffffc) == 0x200c) || ((c >= 0x202a) && (c <= 0x202e))
+                                             || ((c >= 0x2066) && (c <= 0x2069)));
+    }
+
+    void verifyValidPara()
+    {
+        if (!(this == this.paraBidi)) {
+            throw new IllegalStateException();
+        }
+    }
+
+    void verifyValidParaOrLine()
+    {
+        BidiBase para = this.paraBidi;
+        /* verify Para */
+        if (this == para) {
+            return;
+        }
+        /* verify Line */
+        if ((para == null) || (para != para.paraBidi)) {
+            throw new IllegalStateException();
+        }
+    }
+
+    void verifyRange(int index, int start, int limit)
+    {
+        if (index < start || index >= limit) {
+            throw new IllegalArgumentException("Value " + index +
+                      " is out of range " + start + " to " + limit);
+        }
+    }
+
+    /**
+     * Allocate a Bidi object with preallocated memory
+     * for internal structures.
+     * This method provides a Bidi object like the default constructor
+     * but it also preallocates memory for internal structures
+     * according to the sizings supplied by the caller.

+     * The preallocation can be limited to some of the internal memory
+     * by setting some values to 0 here. That means that if, e.g.,
+     * maxRunCount cannot be reasonably predetermined and should not
+     * be set to maxLength (the only failproof value) to avoid
+     * wasting  memory, then maxRunCount could be set to 0 here
+     * and the internal structures that are associated with it will be allocated
+     * on demand, just like with the default constructor.
+     *
+     * @param maxLength is the maximum text or line length that internal memory
+     *        will be preallocated for. An attempt to associate this object with a
+     *        longer text will fail, unless this value is 0, which leaves the allocation
+     *        up to the implementation.
+     *
+     * @param maxRunCount is the maximum anticipated number of same-level runs
+     *        that internal memory will be preallocated for. An attempt to access
+     *        visual runs on an object that was not preallocated for as many runs
+     *        as the text was actually resolved to will fail,
+     *        unless this value is 0, which leaves the allocation up to the implementation.


+     *        The number of runs depends on the actual text and maybe anywhere between
+     *        1 and maxLength. It is typically small.
+     *
+     * @throws IllegalArgumentException if maxLength or maxRunCount is less than 0
+     * @stable ICU 3.8
+     */
+    public BidiBase(int maxLength, int maxRunCount)
+    {
+        /* check the argument values */
+        if (maxLength < 0 || maxRunCount < 0) {
+            throw new IllegalArgumentException();
+        }
+
+        /* reset the object, all reference variables null, all flags false,
+           all sizes 0.
+           In fact, we don't need to do anything, since class members are
+           initialized as zero when an instance is created.
+         */
+        /*
+        mayAllocateText = false;
+        mayAllocateRuns = false;
+        orderParagraphsLTR = false;
+        paraCount = 0;
+        runCount = 0;
+        trailingWSStart = 0;
+        flags = 0;
+        paraLevel = 0;
+        defaultParaLevel = 0;
+        direction = 0;
+        */
+        /* get Bidi properties */
+        bdp = UBiDiProps.INSTANCE;
+
+        /* allocate memory for arrays as requested */
+        if (maxLength > 0) {
+            getInitialDirPropsMemory(maxLength);
+            getInitialLevelsMemory(maxLength);
+        } else {
+            mayAllocateText = true;
+        }
+
+        if (maxRunCount > 0) {
+            // if maxRunCount == 1, use simpleRuns[]
+            if (maxRunCount > 1) {
+                getInitialRunsMemory(maxRunCount);
+            }
+        } else {
+            mayAllocateRuns = true;
+        }
+    }
+
+    /*
+     * We are allowed to allocate memory if object==null or
+     * mayAllocate==true for each array that we need.
+     *
+     * Assume sizeNeeded>0.
+     * If object != null, then assume size > 0.
+     */
+    private Object getMemory(String label, Object array, Class arrayClass,
+            boolean mayAllocate, int sizeNeeded)
+    {
+        int len = Array.getLength(array);
+
+        /* we have at least enough memory and must not allocate */
+        if (sizeNeeded == len) {
+            return array;
+        }
+        if (!mayAllocate) {
+            /* we must not allocate */
+            if (sizeNeeded <= len) {
+                return array;
+            }
+            throw new OutOfMemoryError("Failed to allocate memory for "
+                                       + label);
+        }
+        /* we may try to grow or shrink */
+        /* FOOD FOR THOUGHT: when shrinking it should be possible to avoid
+           the allocation altogether and rely on this.length */
+        try {
+            return Array.newInstance(arrayClass, sizeNeeded);
+        } catch (Exception e) {
+            throw new OutOfMemoryError("Failed to allocate memory for "
+                                       + label);
+        }
+    }
+
+    /* helper methods for each allocated array */
+    private void getDirPropsMemory(boolean mayAllocate, int len)
+    {
+        Object array = getMemory("DirProps", dirPropsMemory, Byte.TYPE, mayAllocate, len);
+        dirPropsMemory = (byte[]) array;
+    }
+
+    void getDirPropsMemory(int len)
+    {
+        getDirPropsMemory(mayAllocateText, len);
+    }
+
+    private void getLevelsMemory(boolean mayAllocate, int len)
+    {
+        Object array = getMemory("Levels", levelsMemory, Byte.TYPE, mayAllocate, len);
+        levelsMemory = (byte[]) array;
+    }
+
+    void getLevelsMemory(int len)
+    {
+        getLevelsMemory(mayAllocateText, len);
+    }
+
+    private void getRunsMemory(boolean mayAllocate, int len)
+    {
+        Object array = getMemory("Runs", runsMemory, BidiRun.class, mayAllocate, len);
+        runsMemory = (BidiRun[]) array;
+    }
+
+    void getRunsMemory(int len)
+    {
+        getRunsMemory(mayAllocateRuns, len);
+    }
+
+    /* additional methods used by constructor - always allow allocation */
+    private void getInitialDirPropsMemory(int len)
+    {
+        getDirPropsMemory(true, len);
+    }
+
+    private void getInitialLevelsMemory(int len)
+    {
+        getLevelsMemory(true, len);
+    }
+
+    private void getInitialRunsMemory(int len)
+    {
+        getRunsMemory(true, len);
+    }
+
+    /**
+     * Is this Bidi object set to perform the inverse Bidi
+     * algorithm?
+     * 
Note: calling this method after setting the reordering mode with
+     * setReorderingMode will return true if the
+     * reordering mode was set to
+     * REORDER_INVERSE_NUMBERS_AS_L, false
+     * for all other values.

+     *
+     * @return true if the Bidi object is set to
+     * perform the inverse Bidi algorithm by handling numbers as L.
+     *
+     * @see #setInverse
+     * @see #setReorderingMode
+     * @see #REORDER_INVERSE_NUMBERS_AS_L
+     * @stable ICU 3.8
+     */
+    public boolean isInverse() {
+        return isInverse;
+    }
+
+    /* perform (P2)..(P3) ------------------------------------------------------- */
+
+    /*
+     * Check that there are enough entries in the arrays paras_limit and paras_level
+     */
+    private void checkParaCount() {
+        int[] saveLimits;
+        byte[] saveLevels;
+        int count = paraCount;
+        if (count <= paras_level.length)
+            return;
+        int oldLength = paras_level.length;
+        saveLimits = paras_limit;
+        saveLevels = paras_level;
+        try {
+            paras_limit = new int[count * 2];
+            paras_level = new byte[count * 2];
+        } catch (Exception e) {
+            throw new OutOfMemoryError("Failed to allocate memory for paras");
+        }
+        System.arraycopy(saveLimits, 0, paras_limit, 0, oldLength);
+        System.arraycopy(saveLevels, 0, paras_level, 0, oldLength);
+    }
+
+    /*
+     * Get the directional properties for the text, calculate the flags bit-set, and
+     * determine the paragraph level if necessary (in paras_level[i]).
+     * FSI initiators are also resolved and their dirProp replaced with LRI or RLI.
+     * When encountering an FSI, it is initially replaced with an LRI, which is the
+     * default. Only if a strong R or AL is found within its scope will the LRI be
+     * replaced by an RLI.
+     */
+    static final int NOT_SEEKING_STRONG = 0;        /* 0: not contextual paraLevel, not after FSI */
+    static final int SEEKING_STRONG_FOR_PARA = 1;   /* 1: looking for first strong char in para */
+    static final int SEEKING_STRONG_FOR_FSI = 2;    /* 2: looking for first strong after FSI */
+    static final int LOOKING_FOR_PDI = 3;           /* 3: found strong after FSI, looking for PDI */
+
+    private void getDirProps()
+    {
+        int i = 0, i0, i1;
+        flags = 0;          /* collect all directionalities in the text */
+        int uchar;
+        byte dirProp;
+        byte defaultParaLevel = 0;   /* initialize to avoid compiler warnings */
+        boolean isDefaultLevel = IsDefaultLevel(paraLevel);
+        /* for inverse Bidi, the default para level is set to RTL if there is a
+           strong R or AL character at either end of the text                */
+        boolean isDefaultLevelInverse=isDefaultLevel &&
+                (reorderingMode == REORDER_INVERSE_LIKE_DIRECT ||
+                 reorderingMode == REORDER_INVERSE_FOR_NUMBERS_SPECIAL);
+        lastArabicPos = -1;
+        int controlCount = 0;
+        boolean removeBidiControls = (reorderingOptions & OPTION_REMOVE_CONTROLS) != 0;
+
+        byte state;
+        byte lastStrong = ON;           /* for default level & inverse Bidi */
+    /* The following stacks are used to manage isolate sequences. Those
+       sequences may be nested, but obviously never more deeply than the
+       maximum explicit embedding level.
+       lastStack is the index of the last used entry in the stack. A value of -1
+       means that there is no open isolate sequence.
+       lastStack is reset to -1 on paragraph boundaries. */
+    /* The following stack contains the position of the initiator of
+       each open isolate sequence */
+        int[] isolateStartStack= new int[MAX_EXPLICIT_LEVEL+1];
+    /* The following stack contains the last known state before
+       encountering the initiator of an isolate sequence */
+        byte[] previousStateStack = new byte[MAX_EXPLICIT_LEVEL+1];
+        int  stackLast=-1;
+
+        if ((reorderingOptions & OPTION_STREAMING) != 0)
+            length = 0;
+        defaultParaLevel = (byte)(paraLevel & 1);
+
+        if (isDefaultLevel) {
+            paras_level[0] = defaultParaLevel;
+            lastStrong = defaultParaLevel;
+            state = SEEKING_STRONG_FOR_PARA;
+        } else {
+            paras_level[0] = paraLevel;
+            state = NOT_SEEKING_STRONG;
+        }
+        /* count paragraphs and determine the paragraph level (P2..P3) */
+        /*
+         * see comment on constant fields:
+         * the LEVEL_DEFAULT_XXX values are designed so that
+         * their low-order bit alone yields the intended default
+         */
+
+        for (i = 0; i < originalLength; /* i is incremented in the loop */) {
+            i0 = i;                     /* index of first code unit */
+            uchar = UTF16.charAt(text, 0, originalLength, i);
+            i += UTF16.getCharCount(uchar);
+            i1 = i - 1; /* index of last code unit, gets the directional property */
+
+            dirProp = (byte)getCustomizedClass(uchar);
+            flags |= DirPropFlag(dirProp);
+            dirProps[i1] = dirProp;
+            if (i1 > i0) {     /* set previous code units' properties to BN */
+                flags |= DirPropFlag(BN);
+                do {
+                    dirProps[--i1] = BN;
+                } while (i1 > i0);
+            }
+            if (removeBidiControls && IsBidiControlChar(uchar)) {
+                controlCount++;
+            }
+            if (dirProp == L) {
+                if (state == SEEKING_STRONG_FOR_PARA) {
+                    paras_level[paraCount - 1] = 0;
+                    state = NOT_SEEKING_STRONG;
+                }
+                else if (state == SEEKING_STRONG_FOR_FSI) {
+                    if (stackLast <= MAX_EXPLICIT_LEVEL) {
+                        /* no need for next statement, already set by default */
+                        /* dirProps[isolateStartStack[stackLast]] = LRI; */
+                        flags |= DirPropFlag(LRI);
+                    }
+                    state = LOOKING_FOR_PDI;
+                }
+                lastStrong = L;
+                continue;
+            }
+            if (dirProp == R || dirProp == AL) {
+                if (state == SEEKING_STRONG_FOR_PARA) {
+                    paras_level[paraCount - 1] = 1;
+                    state = NOT_SEEKING_STRONG;
+                }
+                else if (state == SEEKING_STRONG_FOR_FSI) {
+                    if (stackLast <= MAX_EXPLICIT_LEVEL) {
+                        dirProps[isolateStartStack[stackLast]] = RLI;
+                        flags |= DirPropFlag(RLI);
+                    }
+                    state = LOOKING_FOR_PDI;
+                }
+                lastStrong = R;
+                if (dirProp == AL)
+                    lastArabicPos = i - 1;
+                continue;
+            }
+            if (dirProp >= FSI && dirProp <= RLI) { /* FSI, LRI or RLI */
+                stackLast++;
+                if (stackLast <= MAX_EXPLICIT_LEVEL) {
+                    isolateStartStack[stackLast] = i - 1;
+                    previousStateStack[stackLast] = state;
+                }
+                if (dirProp == FSI) {
+                    dirProps[i-1] = LRI;    /* default if no strong char */
+                    state = SEEKING_STRONG_FOR_FSI;
+                }
+                else
+                    state = LOOKING_FOR_PDI;
+                continue;
+            }
+            if (dirProp == PDI) {
+                if (state == SEEKING_STRONG_FOR_FSI) {
+                    if (stackLast <= MAX_EXPLICIT_LEVEL) {
+                        /* no need for next statement, already set by default */
+                        /* dirProps[isolateStartStack[stackLast]] = LRI; */
+                        flags |= DirPropFlag(LRI);
+                    }
+                }
+                if (stackLast >= 0) {
+                    if (stackLast <= MAX_EXPLICIT_LEVEL)
+                        state = previousStateStack[stackLast];
+                    stackLast--;
+                }
+                continue;
+            }
+            if (dirProp == B) {
+                if (i < originalLength && uchar == CR && text[i] == LF) /* do nothing on the CR */
+                    continue;
+                paras_limit[paraCount - 1] = i;
+                if (isDefaultLevelInverse && lastStrong == R)
+                    paras_level[paraCount - 1] = 1;
+                if ((reorderingOptions & OPTION_STREAMING) != 0) {
+                /* When streaming, we only process whole paragraphs
+                   thus some updates are only done on paragraph boundaries */
+                   length = i;          /* i is index to next character */
+                   this.controlCount = controlCount;
+                }
+                if (i < originalLength) {       /* B not last char in text */
+                    paraCount++;
+                    checkParaCount();   /* check that there is enough memory for a new para entry */
+                    if (isDefaultLevel) {
+                        paras_level[paraCount - 1] = defaultParaLevel;
+                        state = SEEKING_STRONG_FOR_PARA;
+                        lastStrong = defaultParaLevel;
+                    } else {
+                        paras_level[paraCount - 1] = paraLevel;
+                        state = NOT_SEEKING_STRONG;
+                    }
+                    stackLast = -1;
+                }
+                continue;
+            }
+        }
+        /* +Ignore still open isolate sequences with overflow */
+      if (stackLast > MAX_EXPLICIT_LEVEL) {
+            stackLast = MAX_EXPLICIT_LEVEL;
+            state=SEEKING_STRONG_FOR_FSI;   /* to be on the safe side */
+        }
+        /* Resolve direction of still unresolved open FSI sequences */
+        while (stackLast >= 0) {
+            if (state == SEEKING_STRONG_FOR_FSI) {
+                /* no need for next statement, already set by default */
+                /* dirProps[isolateStartStack[stackLast]] = LRI; */
+                flags |= DirPropFlag(LRI);
+                break;
+            }
+            state = previousStateStack[stackLast];
+            stackLast--;
+        }
+        /* When streaming, ignore text after the last paragraph separator */
+        if ((reorderingOptions & OPTION_STREAMING) != 0) {
+            if (length < originalLength)
+                paraCount--;
+        } else {
+            paras_limit[paraCount - 1] = originalLength;
+            this.controlCount = controlCount;
+        }
+        /* For inverse bidi, default para direction is RTL if there is
+           a strong R or AL at either end of the paragraph */
+        if (isDefaultLevelInverse && lastStrong == R) {
+            paras_level[paraCount - 1] = 1;
+        }
+        if (isDefaultLevel) {
+            paraLevel = paras_level[0];
+        }
+        /* The following is needed to resolve the text direction for default level
+           paragraphs containing no strong character */
+        for (i = 0; i < paraCount; i++)
+            flags |= DirPropFlagLR(paras_level[i]);
+
+        if (orderParagraphsLTR && (flags & DirPropFlag(B)) != 0) {
+            flags |= DirPropFlag(L);
+        }
+    }
+
+    /* determine the paragraph level at position index */
+    byte GetParaLevelAt(int pindex)
+    {
+        if (defaultParaLevel == 0 || pindex < paras_limit[0])
+            return paraLevel;
+        int i;
+        for (i = 1; i < paraCount; i++)
+            if (pindex < paras_limit[i])
+                break;
+        if (i >= paraCount)
+            i = paraCount - 1;
+        return paras_level[i];
+    }
+
+    /* Functions for handling paired brackets ----------------------------------- */
+
+    /* In the isoRuns array, the first entry is used for text outside of any
+       isolate sequence.  Higher entries are used for each more deeply nested
+       isolate sequence. isoRunLast is the index of the last used entry.  The
+       openings array is used to note the data of opening brackets not yet
+       matched by a closing bracket, or matched but still susceptible to change
+       level.
+       Each isoRun entry contains the index of the first and
+       one-after-last openings entries for pending opening brackets it
+       contains.  The next openings entry to use is the one-after-last of the
+       most deeply nested isoRun entry.
+       isoRun entries also contain their current embedding level and the last
+       encountered strong character, since these will be needed to resolve
+       the level of paired brackets.  */
+
+    private void bracketInit(BracketData bd) {
+        bd.isoRunLast = 0;
+        bd.isoRuns[0] = new IsoRun();
+        bd.isoRuns[0].start = 0;
+        bd.isoRuns[0].limit = 0;
+        bd.isoRuns[0].level = GetParaLevelAt(0);
+        bd.isoRuns[0].lastStrong = bd.isoRuns[0].lastBase = bd.isoRuns[0].contextDir = (byte)(GetParaLevelAt(0) & 1);
+        bd.isoRuns[0].contextPos = 0;
+        bd.openings = new Opening[SIMPLE_PARAS_COUNT];
+        bd.isNumbersSpecial = reorderingMode == REORDER_NUMBERS_SPECIAL ||
+                              reorderingMode == REORDER_INVERSE_FOR_NUMBERS_SPECIAL;
+    }
+
+    /* paragraph boundary */
+    private void bracketProcessB(BracketData bd, byte level) {
+        bd.isoRunLast = 0;
+        bd.isoRuns[0].limit = 0;
+        bd.isoRuns[0].level = level;
+        bd.isoRuns[0].lastStrong = bd.isoRuns[0].lastBase = bd.isoRuns[0].contextDir = (byte)(level & 1);
+        bd.isoRuns[0].contextPos = 0;
+    }
+
+    /* LRE, LRO, RLE, RLO, PDF */
+    private void bracketProcessBoundary(BracketData bd, int lastCcPos,
+                                        byte contextLevel, byte embeddingLevel) {
+        IsoRun pLastIsoRun = bd.isoRuns[bd.isoRunLast];
+        if ((DirPropFlag(dirProps[lastCcPos]) & MASK_ISO) != 0) /* after an isolate */
+            return;
+        if (NoOverride(embeddingLevel) > NoOverride(contextLevel))  /* not a PDF */
+            contextLevel = embeddingLevel;
+        pLastIsoRun.limit = pLastIsoRun.start;
+        pLastIsoRun.level = embeddingLevel;
+        pLastIsoRun.lastStrong = pLastIsoRun.lastBase = pLastIsoRun.contextDir = (byte)(contextLevel & 1);
+        pLastIsoRun.contextPos = lastCcPos;
+    }
+
+    /* LRI or RLI */
+    private void bracketProcessLRI_RLI(BracketData bd, byte level) {
+        IsoRun pLastIsoRun = bd.isoRuns[bd.isoRunLast];
+        short lastLimit;
+        pLastIsoRun.lastBase = ON;
+        lastLimit = pLastIsoRun.limit;
+        bd.isoRunLast++;
+        pLastIsoRun = bd.isoRuns[bd.isoRunLast];
+        if (pLastIsoRun == null)
+            pLastIsoRun = bd.isoRuns[bd.isoRunLast] = new IsoRun();
+        pLastIsoRun.start = pLastIsoRun.limit = lastLimit;
+        pLastIsoRun.level = level;
+        pLastIsoRun.lastStrong = pLastIsoRun.lastBase = pLastIsoRun.contextDir = (byte)(level & 1);
+        pLastIsoRun.contextPos = 0;
+    }
+
+    /* PDI */
+    private void bracketProcessPDI(BracketData bd) {
+        IsoRun pLastIsoRun;
+        bd.isoRunLast--;
+        pLastIsoRun = bd.isoRuns[bd.isoRunLast];
+        pLastIsoRun.lastBase = ON;
+    }
+
+    /* newly found opening bracket: create an openings entry */
+    private void bracketAddOpening(BracketData bd, char match, int position) {
+        IsoRun pLastIsoRun = bd.isoRuns[bd.isoRunLast];
+        Opening pOpening;
+        if (pLastIsoRun.limit >= bd.openings.length) {  /* no available new entry */
+            Opening[] saveOpenings = bd.openings;
+            int count;
+            try {
+                count = bd.openings.length;
+                bd.openings = new Opening[count * 2];
+            } catch (Exception e) {
+                throw new OutOfMemoryError("Failed to allocate memory for openings");
+            }
+            System.arraycopy(saveOpenings, 0, bd.openings, 0, count);
+        }
+        pOpening = bd.openings[pLastIsoRun.limit];
+        if (pOpening == null)
+            pOpening = bd.openings[pLastIsoRun.limit]= new Opening();
+        pOpening.position = position;
+        pOpening.match = match;
+        pOpening.contextDir = pLastIsoRun.contextDir;
+        pOpening.contextPos = pLastIsoRun.contextPos;
+        pOpening.flags = 0;
+        pLastIsoRun.limit++;
+    }
+
+    /* change N0c1 to N0c2 when a preceding bracket is assigned the embedding level */
+    private void fixN0c(BracketData bd, int openingIndex, int newPropPosition, byte newProp) {
+        /* This function calls itself recursively */
+        IsoRun pLastIsoRun = bd.isoRuns[bd.isoRunLast];
+        Opening qOpening;
+        int k, openingPosition, closingPosition;
+        for (k = openingIndex+1; k < pLastIsoRun.limit; k++) {
+            qOpening = bd.openings[k];
+            if (qOpening.match >= 0)    /* not an N0c match */
+                continue;
+            if (newPropPosition < qOpening.contextPos)
+                break;
+            if (newPropPosition >= qOpening.position)
+                continue;
+            if (newProp == qOpening.contextDir)
+                break;
+            openingPosition = qOpening.position;
+            dirProps[openingPosition] = newProp;
+            closingPosition = -(qOpening.match);
+            dirProps[closingPosition] = newProp;
+            qOpening.match = 0;                                 /* prevent further changes */
+            fixN0c(bd, k, openingPosition, newProp);
+            fixN0c(bd, k, closingPosition, newProp);
+        }
+    }
+
+    /* process closing bracket; return L or R if N0b or N0c, ON if N0d */
+    private byte bracketProcessClosing(BracketData bd, int openIdx, int position) {
+        IsoRun pLastIsoRun = bd.isoRuns[bd.isoRunLast];
+        Opening pOpening, qOpening;
+        byte direction;
+        boolean stable;
+        byte newProp;
+        pOpening = bd.openings[openIdx];
+        direction = (byte)(pLastIsoRun.level & 1);
+        stable = true;          /* assume stable until proved otherwise */
+
+        /* The stable flag is set when brackets are paired and their
+           level is resolved and cannot be changed by what will be
+           found later in the source string.
+           An unstable match can occur only when applying N0c, where
+           the resolved level depends on the preceding context, and
+           this context may be affected by text occurring later.
+           Example: RTL paragraph containing:  abc[(latin) HEBREW]
+           When the closing parenthesis is encountered, it appears
+           that N0c1 must be applied since 'abc' sets an opposite
+           direction context and both parentheses receive level 2.
+           However, when the closing square bracket is processed,
+           N0b applies because of 'HEBREW' being included within the
+           brackets, thus the square brackets are treated like R and
+           receive level 1. However, this changes the preceding
+           context of the opening parenthesis, and it now appears
+           that N0c2 must be applied to the parentheses rather than
+           N0c1. */
+
+            if ((direction == 0 && (pOpening.flags & FOUND_L) > 0) ||
+                (direction == 1 && (pOpening.flags & FOUND_R) > 0)) {   /* N0b */
+                newProp = direction;
+            }
+            else if ((pOpening.flags & (FOUND_L | FOUND_R)) != 0) {     /* N0c */
+                    /* it is stable if there is no preceding text or in
+                       conditions too complicated and not worth checking */
+                    stable = (openIdx == pLastIsoRun.start);
+                if (direction != pOpening.contextDir)
+                    newProp = pOpening.contextDir;                      /* N0c1 */
+                else
+                    newProp = direction;                                /* N0c2 */
+            } else {
+            /* forget this and any brackets nested within this pair */
+            pLastIsoRun.limit = (short)openIdx;
+            return ON;                                                  /* N0d */
+        }
+        dirProps[pOpening.position] = newProp;
+        dirProps[position] = newProp;
+        /* Update nested N0c pairs that may be affected */
+        fixN0c(bd, openIdx, pOpening.position, newProp);
+        if (stable) {
+            pLastIsoRun.limit = (short)openIdx; /* forget any brackets nested within this pair */
+            /* remove lower located synonyms if any */
+            while (pLastIsoRun.limit > pLastIsoRun.start &&
+                   bd.openings[pLastIsoRun.limit - 1].position == pOpening.position)
+                pLastIsoRun.limit--;
+        } else {
+            int k;
+            pOpening.match = -position;
+            /* neutralize lower located synonyms if any */
+            k = openIdx - 1;
+            while (k >= pLastIsoRun.start &&
+                   bd.openings[k].position == pOpening.position)
+                bd.openings[k--].match = 0;
+            /* neutralize any unmatched opening between the current pair;
+               this will also neutralize higher located synonyms if any */
+            for (k = openIdx + 1; k < pLastIsoRun.limit; k++) {
+                qOpening =bd.openings[k];
+                if (qOpening.position >= position)
+                    break;
+                if (qOpening.match > 0)
+                    qOpening.match = 0;
+            }
+        }
+        return newProp;
+    }
+
+    /* handle strong characters, digits and candidates for closing brackets */
+    private void bracketProcessChar(BracketData bd, int position) {
+        IsoRun pLastIsoRun = bd.isoRuns[bd.isoRunLast];
+        byte dirProp, newProp;
+        byte level;
+        dirProp = dirProps[position];
+        if (dirProp == ON) {
+            char c, match;
+            int idx;
+            /* First see if it is a matching closing bracket. Hopefully, this is
+               more efficient than checking if it is a closing bracket at all */
+            c = text[position];
+            for (idx = pLastIsoRun.limit - 1; idx >= pLastIsoRun.start; idx--) {
+                if (bd.openings[idx].match != c)
+                    continue;
+                /* We have a match */
+                newProp = bracketProcessClosing(bd, idx, position);
+                if(newProp == ON) {         /* N0d */
+                    c = 0;          /* prevent handling as an opening */
+                    break;
+                }
+                pLastIsoRun.lastBase = ON;
+                pLastIsoRun.contextDir = newProp;
+                pLastIsoRun.contextPos = position;
+                level = levels[position];
+                if ((level & LEVEL_OVERRIDE) != 0) {    /* X4, X5 */
+                    short flag;
+                    int i;
+                    newProp = (byte)(level & 1);
+                    pLastIsoRun.lastStrong = newProp;
+                    flag = (short)DirPropFlag(newProp);
+                    for (i = pLastIsoRun.start; i < idx; i++)
+                        bd.openings[i].flags |= flag;
+                    /* matching brackets are not overridden by LRO/RLO */
+                    levels[position] &= ~LEVEL_OVERRIDE;
+                }
+                /* matching brackets are not overridden by LRO/RLO */
+                levels[bd.openings[idx].position] &= ~LEVEL_OVERRIDE;
+                return;
+            }
+            /* We get here only if the ON character is not a matching closing
+               bracket or it is a case of N0d */
+            /* Now see if it is an opening bracket */
+            if (c != 0) {
+                match = (char)UCharacter.getBidiPairedBracket(c); /* get the matching char */
+            } else {
+                match = 0;
+            }
+            if (match != c &&               /* has a matching char */
+                UCharacter.getIntPropertyValue(c, BIDI_PAIRED_BRACKET_TYPE) ==
+                    /* opening bracket */         BidiPairedBracketType.OPEN) {
+                /* special case: process synonyms
+                   create an opening entry for each synonym */
+                if (match == 0x232A) {      /* RIGHT-POINTING ANGLE BRACKET */
+                    bracketAddOpening(bd, (char)0x3009, position);
+                }
+                else if (match == 0x3009) { /* RIGHT ANGLE BRACKET */
+                    bracketAddOpening(bd, (char)0x232A, position);
+                }
+                bracketAddOpening(bd, match, position);
+            }
+        }
+        level = levels[position];
+        if ((level & LEVEL_OVERRIDE) != 0) {    /* X4, X5 */
+            newProp = (byte)(level & 1);
+            if (dirProp != S && dirProp != WS && dirProp != ON)
+                dirProps[position] = newProp;
+            pLastIsoRun.lastBase = newProp;
+            pLastIsoRun.lastStrong = newProp;
+            pLastIsoRun.contextDir = newProp;
+            pLastIsoRun.contextPos = position;
+        }
+        else if (dirProp <= R || dirProp == AL) {
+            newProp = DirFromStrong(dirProp);
+            pLastIsoRun.lastBase = dirProp;
+            pLastIsoRun.lastStrong = dirProp;
+            pLastIsoRun.contextDir = newProp;
+            pLastIsoRun.contextPos = position;
+        }
+        else if(dirProp == EN) {
+            pLastIsoRun.lastBase = EN;
+            if (pLastIsoRun.lastStrong == L) {
+                newProp = L;                    /* W7 */
+                if (!bd.isNumbersSpecial)
+                    dirProps[position] = ENL;
+                pLastIsoRun.contextDir = L;
+                pLastIsoRun.contextPos = position;
+            }
+            else {
+                newProp = R;                    /* N0 */
+                if (pLastIsoRun.lastStrong == AL)
+                    dirProps[position] = AN;    /* W2 */
+                else
+                    dirProps[position] = ENR;
+                pLastIsoRun.contextDir = R;
+                pLastIsoRun.contextPos = position;
+            }
+        }
+        else if (dirProp == AN) {
+            newProp = R;                        /* N0 */
+            pLastIsoRun.lastBase = AN;
+            pLastIsoRun.contextDir = R;
+            pLastIsoRun.contextPos = position;
+        }
+        else if (dirProp == NSM) {
+            /* if the last real char was ON, change NSM to ON so that it
+               will stay ON even if the last real char is a bracket which
+               may be changed to L or R */
+            newProp = pLastIsoRun.lastBase;
+            if (newProp == ON)
+                dirProps[position] = newProp;
+        }
+        else {
+            newProp = dirProp;
+            pLastIsoRun.lastBase = dirProp;
+        }
+        if (newProp <= R || newProp == AL) {
+            int i;
+            short flag = (short)DirPropFlag(DirFromStrong(newProp));
+            for (i = pLastIsoRun.start; i < pLastIsoRun.limit; i++)
+                if (position > bd.openings[i].position)
+                    bd.openings[i].flags |= flag;
+        }
+    }
+
+    /* perform (X1)..(X9) ------------------------------------------------------- */
+
+    /* determine if the text is mixed-directional or single-directional */
+    private byte directionFromFlags() {
+
+        /* if the text contains AN and neutrals, then some neutrals may become RTL */
+        if (!((flags & MASK_RTL) != 0 ||
+              ((flags & DirPropFlag(AN)) != 0 &&
+               (flags & MASK_POSSIBLE_N) != 0))) {
+            return LTR;
+        } else if ((flags & MASK_LTR) == 0) {
+            return RTL;
+        } else {
+            return MIXED;
+        }
+    }
+
+    /*
+     * Resolve the explicit levels as specified by explicit embedding codes.
+     * Recalculate the flags to have them reflect the real properties
+     * after taking the explicit embeddings into account.
+     *
+     * The BiDi algorithm is designed to result in the same behavior whether embedding
+     * levels are externally specified (from "styled text", supposedly the preferred
+     * method) or set by explicit embedding codes (LRx, RLx, PDF, FSI, PDI) in the plain text.
+     * That is why (X9) instructs to remove all not-isolate explicit codes (and BN).
+     * However, in a real implementation, the removal of these codes and their index
+     * positions in the plain text is undesirable since it would result in
+     * reallocated, reindexed text.
+     * Instead, this implementation leaves the codes in there and just ignores them
+     * in the subsequent processing.
+     * In order to get the same reordering behavior, positions with a BN or a not-isolate
+     * explicit embedding code just get the same level assigned as the last "real"
+     * character.
+     *
+     * Some implementations, not this one, then overwrite some of these
+     * directionality properties at "real" same-level-run boundaries by
+     * L or R codes so that the resolution of weak types can be performed on the
+     * entire paragraph at once instead of having to parse it once more and
+     * perform that resolution on same-level-runs.
+     * This limits the scope of the implicit rules in effectively
+     * the same way as the run limits.
+     *
+     * Instead, this implementation does not modify these codes, except for
+     * paired brackets whose properties (ON) may be replaced by L or R.
+     * On one hand, the paragraph has to be scanned for same-level-runs, but
+     * on the other hand, this saves another loop to reset these codes,
+     * or saves making and modifying a copy of dirProps[].
+     *
+     *
+     * Note that (Pn) and (Xn) changed significantly from version 4 of the BiDi algorithm.
+     *
+     *
+     * Handling the stack of explicit levels (Xn):
+     *
+     * With the BiDi stack of explicit levels, as pushed with each
+     * LRE, RLE, LRO, RLO, LRI, RLI and FSI and popped with each PDF and PDI,
+     * the explicit level must never exceed MAX_EXPLICIT_LEVEL.
+     *
+     * In order to have a correct push-pop semantics even in the case of overflows,
+     * overflow counters and a valid isolate counter are used as described in UAX#9
+     * section 3.3.2 "Explicit Levels and Directions".
+     *
+     * This implementation assumes that MAX_EXPLICIT_LEVEL is odd.
+     *
+     * Returns the direction
+     *
+     */
+    private byte resolveExplicitLevels() {
+        int i = 0;
+        byte dirProp;
+        byte level = GetParaLevelAt(0);
+        byte dirct;
+        isolateCount = 0;
+
+        /* determine if the text is mixed-directional or single-directional */
+        dirct = directionFromFlags();
+
+        /* we may not need to resolve any explicit levels */
+        if (dirct != MIXED) {
+            /* not mixed directionality: levels don't matter - trailingWSStart will be 0 */
+            return dirct;
+        }
+
+        if (reorderingMode > REORDER_LAST_LOGICAL_TO_VISUAL) {
+            /* inverse BiDi: mixed, but all characters are at the same embedding level */
+            /* set all levels to the paragraph level */
+            int paraIndex, start, limit;
+            for (paraIndex = 0; paraIndex < paraCount; paraIndex++) {
+                if (paraIndex == 0)
+                    start = 0;
+                else
+                    start = paras_limit[paraIndex - 1];
+                limit = paras_limit[paraIndex];
+                level = paras_level[paraIndex];
+                for (i = start; i < limit; i++)
+                    levels[i] =level;
+            }
+            return dirct;               /* no bracket matching for inverse BiDi */
+        }
+        if ((flags & (MASK_EXPLICIT | MASK_ISO)) == 0) {
+            /* no embeddings, set all levels to the paragraph level */
+            /* we still have to perform bracket matching */
+            int paraIndex, start, limit;
+            BracketData bracketData = new BracketData();
+            bracketInit(bracketData);
+            for (paraIndex = 0; paraIndex < paraCount; paraIndex++) {
+                if (paraIndex == 0)
+                    start = 0;
+                else
+                    start = paras_limit[paraIndex-1];
+                limit = paras_limit[paraIndex];
+                level = paras_level[paraIndex];
+                for (i = start; i < limit; i++) {
+                    levels[i] = level;
+                    dirProp = dirProps[i];
+                    if (dirProp == BN)
+                        continue;
+                    if (dirProp == B) {
+                        if ((i + 1) < length) {
+                            if (text[i] == CR && text[i + 1] == LF)
+                                continue;   /* skip CR when followed by LF */
+                            bracketProcessB(bracketData, level);
+                        }
+                        continue;
+                    }
+                    bracketProcessChar(bracketData, i);
+                }
+            }
+            return dirct;
+        }
+        /* continue to perform (Xn) */
+
+        /* (X1) level is set for all codes, embeddingLevel keeps track of the push/pop operations */
+        /* both variables may carry the LEVEL_OVERRIDE flag to indicate the override status */
+        byte embeddingLevel = level, newLevel;
+        byte previousLevel = level; /* previous level for regular (not CC) characters */
+        int lastCcPos = 0;          /* index of last effective LRx,RLx, PDx */
+
+        /* The following stack remembers the embedding level and the ISOLATE flag of level runs.
+           stackLast points to its current entry. */
+        short[] stack = new short[MAX_EXPLICIT_LEVEL + 2];  /* we never push anything >= MAX_EXPLICIT_LEVEL
+                                                               but we need one more entry as base */
+        int stackLast = 0;
+        int overflowIsolateCount = 0;
+        int overflowEmbeddingCount = 0;
+        int validIsolateCount = 0;
+        BracketData bracketData = new BracketData();
+        bracketInit(bracketData);
+        stack[0] = level;       /* initialize base entry to para level, no override, no isolate */
+
+        /* recalculate the flags */
+        flags = 0;
+
+        for (i = 0; i < length; i++) {
+            dirProp = dirProps[i];
+            switch (dirProp) {
+            case LRE:
+            case RLE:
+            case LRO:
+            case RLO:
+                /* (X2, X3, X4, X5) */
+                flags |= DirPropFlag(BN);
+                levels[i] = previousLevel;
+                if (dirProp == LRE || dirProp == LRO) {
+                    /* least greater even level */
+                    newLevel = (byte)((embeddingLevel+2) & ~(LEVEL_OVERRIDE | 1));
+                } else {
+                    /* least greater odd level */
+                    newLevel = (byte)((NoOverride(embeddingLevel) + 1) | 1);
+                }
+                if (newLevel <= MAX_EXPLICIT_LEVEL && overflowIsolateCount == 0 &&
+                                                      overflowEmbeddingCount == 0) {
+                    lastCcPos = i;
+                    embeddingLevel = newLevel;
+                    if (dirProp == LRO || dirProp == RLO)
+                        embeddingLevel |= LEVEL_OVERRIDE;
+                    stackLast++;
+                    stack[stackLast] = embeddingLevel;
+                    /* we don't need to set LEVEL_OVERRIDE off for LRE and RLE
+                       since this has already been done for newLevel which is
+                       the source for embeddingLevel.
+                     */
+                } else {
+                    if (overflowIsolateCount == 0)
+                        overflowEmbeddingCount++;
+                }
+                break;
+            case PDF:
+                /* (X7) */
+                flags |= DirPropFlag(BN);
+                levels[i] = previousLevel;
+                /* handle all the overflow cases first */
+                if (overflowIsolateCount > 0) {
+                    break;
+                }
+                if (overflowEmbeddingCount > 0) {
+                    overflowEmbeddingCount--;
+                    break;
+                }
+                if (stackLast > 0 && stack[stackLast] < ISOLATE) {   /* not an isolate entry */
+                    lastCcPos = i;
+                    stackLast--;
+                    embeddingLevel = (byte)stack[stackLast];
+                }
+                break;
+            case LRI:
+            case RLI:
+                flags |= DirPropFlag(ON) | DirPropFlagLR(embeddingLevel);
+                levels[i] = NoOverride(embeddingLevel);
+                if (NoOverride(embeddingLevel) != NoOverride(previousLevel)) {
+                    bracketProcessBoundary(bracketData, lastCcPos,
+                                           previousLevel, embeddingLevel);
+                    flags |= DirPropFlagMultiRuns;
+                }
+                previousLevel = embeddingLevel;
+                /* (X5a, X5b) */
+                if (dirProp == LRI)
+                    /* least greater even level */
+                    newLevel=(byte)((embeddingLevel+2)&~(LEVEL_OVERRIDE|1));
+                else
+                    /* least greater odd level */
+                    newLevel=(byte)((NoOverride(embeddingLevel)+1)|1);
+                if (newLevel <= MAX_EXPLICIT_LEVEL && overflowIsolateCount == 0
+                                                   && overflowEmbeddingCount == 0) {
+                    flags |= DirPropFlag(dirProp);
+                    lastCcPos = i;
+                    validIsolateCount++;
+                    if (validIsolateCount > isolateCount)
+                        isolateCount = validIsolateCount;
+                    embeddingLevel = newLevel;
+                    /* we can increment stackLast without checking because newLevel
+                       will exceed UBIDI_MAX_EXPLICIT_LEVEL before stackLast overflows */
+                    stackLast++;
+                    stack[stackLast] = (short)(embeddingLevel + ISOLATE);
+                    bracketProcessLRI_RLI(bracketData, embeddingLevel);
+                } else {
+                    /* make it WS so that it is handled by adjustWSLevels() */
+                    dirProps[i] = WS;
+                    overflowIsolateCount++;
+                }
+                break;
+            case PDI:
+                if (NoOverride(embeddingLevel) != NoOverride(previousLevel)) {
+                    bracketProcessBoundary(bracketData, lastCcPos,
+                                           previousLevel, embeddingLevel);
+                    flags |= DirPropFlagMultiRuns;
+                }
+                /* (X6a) */
+                if (overflowIsolateCount > 0) {
+                    overflowIsolateCount--;
+                    /* make it WS so that it is handled by adjustWSLevels() */
+                    dirProps[i] = WS;
+                }
+                else if (validIsolateCount > 0) {
+                    flags |= DirPropFlag(PDI);
+                    lastCcPos = i;
+                    overflowEmbeddingCount = 0;
+                    while (stack[stackLast] < ISOLATE)  /* pop embedding entries */
+                        stackLast--;                    /* until the last isolate entry */
+                    stackLast--;                        /* pop also the last isolate entry */
+                    validIsolateCount--;
+                    bracketProcessPDI(bracketData);
+                } else
+                    /* make it WS so that it is handled by adjustWSLevels() */
+                    dirProps[i] = WS;
+                embeddingLevel = (byte)(stack[stackLast] & ~ISOLATE);
+                flags |= DirPropFlag(ON) | DirPropFlagLR(embeddingLevel);
+                previousLevel = embeddingLevel;
+                levels[i] = NoOverride(embeddingLevel);
+                break;
+            case B:
+                flags |= DirPropFlag(B);
+                levels[i] = GetParaLevelAt(i);
+                if ((i + 1) < length) {
+                    if (text[i] == CR && text[i + 1] == LF)
+                        break;          /* skip CR when followed by LF */
+                    overflowEmbeddingCount = overflowIsolateCount = 0;
+                    validIsolateCount = 0;
+                    stackLast = 0;
+                    previousLevel = embeddingLevel = GetParaLevelAt(i + 1);
+                    stack[0] = embeddingLevel;   /* initialize base entry to para level, no override, no isolate */
+                    bracketProcessB(bracketData, embeddingLevel);
+                }
+                break;
+            case BN:
+                /* BN, LRE, RLE, and PDF are supposed to be removed (X9) */
+                /* they will get their levels set correctly in adjustWSLevels() */
+                levels[i] = previousLevel;
+                flags |= DirPropFlag(BN);
+                break;
+            default:
+                /* all other types are normal characters and get the "real" level */
+                if (NoOverride(embeddingLevel) != NoOverride(previousLevel)) {
+                    bracketProcessBoundary(bracketData, lastCcPos,
+                                           previousLevel, embeddingLevel);
+                    flags |= DirPropFlagMultiRuns;
+                    if ((embeddingLevel & LEVEL_OVERRIDE) != 0)
+                        flags |= DirPropFlagO(embeddingLevel);
+                    else
+                        flags |= DirPropFlagE(embeddingLevel);
+                }
+                previousLevel = embeddingLevel;
+                levels[i] = embeddingLevel;
+                bracketProcessChar(bracketData, i);
+                /* the dirProp may have been changed in bracketProcessChar() */
+                flags |= DirPropFlag(dirProps[i]);
+                break;
+            }
+        }
+        if ((flags & MASK_EMBEDDING) != 0) {
+            flags |= DirPropFlagLR(paraLevel);
+        }
+        if (orderParagraphsLTR && (flags & DirPropFlag(B)) != 0) {
+            flags |= DirPropFlag(L);
+        }
+        /* again, determine if the text is mixed-directional or single-directional */
+        dirct = directionFromFlags();
+
+        return dirct;
+    }
+
+    /*
+     * Use a pre-specified embedding levels array:
+     *
+     * Adjust the directional properties for overrides (->LEVEL_OVERRIDE),
+     * ignore all explicit codes (X9),
+     * and check all the preset levels.
+     *
+     * Recalculate the flags to have them reflect the real properties
+     * after taking the explicit embeddings into account.
+     */
+    private byte checkExplicitLevels() {
+        byte dirProp;
+        int i;
+        int isolateCount = 0;
+
+        this.flags = 0;     /* collect all directionalities in the text */
+        byte level;
+        this.isolateCount = 0;
+
+        for (i = 0; i < length; ++i) {
+            if (levels[i] == 0) {
+               levels[i] = paraLevel;
+            }
+
+            // for backward compatibility
+            if (MAX_EXPLICIT_LEVEL < (levels[i]&0x7f)) {
+                if ((levels[i] & LEVEL_OVERRIDE) != 0) {
+                    levels[i] =  (byte)(paraLevel|LEVEL_OVERRIDE);
+                } else {
+                    levels[i] = paraLevel;
+                }
+            }
+
+            level = levels[i];
+            dirProp = dirProps[i];
+            if (dirProp == LRI || dirProp == RLI) {
+                isolateCount++;
+                if (isolateCount > this.isolateCount)
+                    this.isolateCount = isolateCount;
+            }
+            else if (dirProp == PDI) {
+                isolateCount--;
+            } else if (dirProp == B) {
+                isolateCount = 0;
+            }
+            if ((level & LEVEL_OVERRIDE) != 0) {
+                /* keep the override flag in levels[i] but adjust the flags */
+                level &= ~LEVEL_OVERRIDE;     /* make the range check below simpler */
+                flags |= DirPropFlagO(level);
+            } else {
+                /* set the flags */
+                flags |= DirPropFlagE(level) | DirPropFlag(dirProp);
+            }
+            if ((level < GetParaLevelAt(i) &&
+                    !((0 == level) && (dirProp == B))) ||
+                    (MAX_EXPLICIT_LEVEL < level)) {
+                /* level out of bounds */
+                throw new IllegalArgumentException("level " + level +
+                                                   " out of bounds at " + i);
+            }
+        }
+        if ((flags & MASK_EMBEDDING) != 0) {
+            flags |= DirPropFlagLR(paraLevel);
+        }
+        /* determine if the text is mixed-directional or single-directional */
+        return directionFromFlags();
+    }
+
+    /*********************************************************************/
+    /* The Properties state machine table                                */
+    /*********************************************************************/
+    /*                                                                   */
+    /* All table cells are 8 bits:                                       */
+    /*      bits 0..4:  next state                                       */
+    /*      bits 5..7:  action to perform (if > 0)                       */
+    /*                                                                   */
+    /* Cells may be of format "n" where n represents the next state      */
+    /* (except for the rightmost column).                                */
+    /* Cells may also be of format "_(x,y)" where x represents an action */
+    /* to perform and y represents the next state.                       */
+    /*                                                                   */
+    /*********************************************************************/
+    /* Definitions and type for properties state tables                  */
+    /*********************************************************************/
+    private static final int IMPTABPROPS_COLUMNS = 16;
+    private static final int IMPTABPROPS_RES = IMPTABPROPS_COLUMNS - 1;
+    private static short GetStateProps(short cell) {
+        return (short)(cell & 0x1f);
+    }
+    private static short GetActionProps(short cell) {
+        return (short)(cell >> 5);
+    }
+
+    private static final short groupProp[] =          /* dirProp regrouped */
+    {
+        /*  L   R   EN  ES  ET  AN  CS  B   S   WS  ON  LRE LRO AL  RLE RLO PDF NSM BN  FSI LRI RLI PDI ENL ENR */
+            0,  1,  2,  7,  8,  3,  9,  6,  5,  4,  4,  10, 10, 12, 10, 10, 10, 11, 10, 4,  4,  4,  4,  13, 14
+    };
+    private static final short _L  = 0;
+    private static final short _R  = 1;
+    private static final short _EN = 2;
+    private static final short _AN = 3;
+    private static final short _ON = 4;
+    private static final short _S  = 5;
+    private static final short _B  = 6; /* reduced dirProp */
+
+    /*********************************************************************/
+    /*                                                                   */
+    /*      PROPERTIES  STATE  TABLE                                     */
+    /*                                                                   */
+    /* In table impTabProps,                                             */
+    /*      - the ON column regroups ON and WS, FSI, RLI, LRI and PDI    */
+    /*      - the BN column regroups BN, LRE, RLE, LRO, RLO, PDF         */
+    /*      - the Res column is the reduced property assigned to a run   */
+    /*                                                                   */
+    /* Action 1: process current run1, init new run1                     */
+    /*        2: init new run2                                           */
+    /*        3: process run1, process run2, init new run1               */
+    /*        4: process run1, set run1=run2, init new run2              */
+    /*                                                                   */
+    /* Notes:                                                            */
+    /*  1) This table is used in resolveImplicitLevels().                */
+    /*  2) This table triggers actions when there is a change in the Bidi*/
+    /*     property of incoming characters (action 1).                   */
+    /*  3) Most such property sequences are processed immediately (in    */
+    /*     fact, passed to processPropertySeq().                         */
+    /*  4) However, numbers are assembled as one sequence. This means    */
+    /*     that undefined situations (like CS following digits, until    */
+    /*     it is known if the next char will be a digit) are held until  */
+    /*     following chars define them.                                  */
+    /*     Example: digits followed by CS, then comes another CS or ON;  */
+    /*              the digits will be processed, then the CS assigned   */
+    /*              as the start of an ON sequence (action 3).           */
+    /*  5) There are cases where more than one sequence must be          */
+    /*     processed, for instance digits followed by CS followed by L:  */
+    /*     the digits must be processed as one sequence, and the CS      */
+    /*     must be processed as an ON sequence, all this before starting */
+    /*     assembling chars for the opening L sequence.                  */
+    /*                                                                   */
+    /*                                                                   */
+    private static final short impTabProps[][] =
+    {
+/*                        L,     R,    EN,    AN,    ON,     S,     B,    ES,    ET,    CS,    BN,   NSM,    AL,   ENL,   ENR,   Res */
+/* 0 Init        */ {     1,     2,     4,     5,     7,    15,    17,     7,     9,     7,     0,     7,     3,    18,    21,   _ON },
+/* 1 L           */ {     1,  32+2,  32+4,  32+5,  32+7, 32+15, 32+17,  32+7,  32+9,  32+7,     1,     1,  32+3, 32+18, 32+21,    _L },
+/* 2 R           */ {  32+1,     2,  32+4,  32+5,  32+7, 32+15, 32+17,  32+7,  32+9,  32+7,     2,     2,  32+3, 32+18, 32+21,    _R },
+/* 3 AL          */ {  32+1,  32+2,  32+6,  32+6,  32+8, 32+16, 32+17,  32+8,  32+8,  32+8,     3,     3,     3, 32+18, 32+21,    _R },
+/* 4 EN          */ {  32+1,  32+2,     4,  32+5,  32+7, 32+15, 32+17, 64+10,    11, 64+10,     4,     4,  32+3,    18,    21,   _EN },
+/* 5 AN          */ {  32+1,  32+2,  32+4,     5,  32+7, 32+15, 32+17,  32+7,  32+9, 64+12,     5,     5,  32+3, 32+18, 32+21,   _AN },
+/* 6 AL:EN/AN    */ {  32+1,  32+2,     6,     6,  32+8, 32+16, 32+17,  32+8,  32+8, 64+13,     6,     6,  32+3,    18,    21,   _AN },
+/* 7 ON          */ {  32+1,  32+2,  32+4,  32+5,     7, 32+15, 32+17,     7, 64+14,     7,     7,     7,  32+3, 32+18, 32+21,   _ON },
+/* 8 AL:ON       */ {  32+1,  32+2,  32+6,  32+6,     8, 32+16, 32+17,     8,     8,     8,     8,     8,  32+3, 32+18, 32+21,   _ON },
+/* 9 ET          */ {  32+1,  32+2,     4,  32+5,     7, 32+15, 32+17,     7,     9,     7,     9,     9,  32+3,    18,    21,   _ON },
+/*10 EN+ES/CS    */ {  96+1,  96+2,     4,  96+5, 128+7, 96+15, 96+17, 128+7,128+14, 128+7,    10, 128+7,  96+3,    18,    21,   _EN },
+/*11 EN+ET       */ {  32+1,  32+2,     4,  32+5,  32+7, 32+15, 32+17,  32+7,    11,  32+7,    11,    11,  32+3,    18,    21,   _EN },
+/*12 AN+CS       */ {  96+1,  96+2,  96+4,     5, 128+7, 96+15, 96+17, 128+7,128+14, 128+7,    12, 128+7,  96+3, 96+18, 96+21,   _AN },
+/*13 AL:EN/AN+CS */ {  96+1,  96+2,     6,     6, 128+8, 96+16, 96+17, 128+8, 128+8, 128+8,    13, 128+8,  96+3,    18,    21,   _AN },
+/*14 ON+ET       */ {  32+1,  32+2, 128+4,  32+5,     7, 32+15, 32+17,     7,    14,     7,    14,    14,  32+3,128+18,128+21,   _ON },
+/*15 S           */ {  32+1,  32+2,  32+4,  32+5,  32+7,    15, 32+17,  32+7,  32+9,  32+7,    15,  32+7,  32+3, 32+18, 32+21,    _S },
+/*16 AL:S        */ {  32+1,  32+2,  32+6,  32+6,  32+8,    16, 32+17,  32+8,  32+8,  32+8,    16,  32+8,  32+3, 32+18, 32+21,    _S },
+/*17 B           */ {  32+1,  32+2,  32+4,  32+5,  32+7, 32+15,    17,  32+7,  32+9,  32+7,    17,  32+7,  32+3, 32+18, 32+21,    _B },
+/*18 ENL         */ {  32+1,  32+2,    18,  32+5,  32+7, 32+15, 32+17, 64+19,    20, 64+19,    18,    18,  32+3,    18,    21,    _L },
+/*19 ENL+ES/CS   */ {  96+1,  96+2,    18,  96+5, 128+7, 96+15, 96+17, 128+7,128+14, 128+7,    19, 128+7,  96+3,    18,    21,    _L },
+/*20 ENL+ET      */ {  32+1,  32+2,    18,  32+5,  32+7, 32+15, 32+17,  32+7,    20,  32+7,    20,    20,  32+3,    18,    21,    _L },
+/*21 ENR         */ {  32+1,  32+2,    21,  32+5,  32+7, 32+15, 32+17, 64+22,    23, 64+22,    21,    21,  32+3,    18,    21,   _AN },
+/*22 ENR+ES/CS   */ {  96+1,  96+2,    21,  96+5, 128+7, 96+15, 96+17, 128+7,128+14, 128+7,    22, 128+7,  96+3,    18,    21,   _AN },
+/*23 ENR+ET      */ {  32+1,  32+2,    21,  32+5,  32+7, 32+15, 32+17,  32+7,    23,  32+7,    23,    23,  32+3,    18,    21,   _AN }
+    };
+
+    /*********************************************************************/
+    /* The levels state machine tables                                   */
+    /*********************************************************************/
+    /*                                                                   */
+    /* All table cells are 8 bits:                                       */
+    /*      bits 0..3:  next state                                       */
+    /*      bits 4..7:  action to perform (if > 0)                       */
+    /*                                                                   */
+    /* Cells may be of format "n" where n represents the next state      */
+    /* (except for the rightmost column).                                */
+    /* Cells may also be of format "_(x,y)" where x represents an action */
+    /* to perform and y represents the next state.                       */
+    /*                                                                   */
+    /* This format limits each table to 16 states each and to 15 actions.*/
+    /*                                                                   */
+    /*********************************************************************/
+    /* Definitions and type for levels state tables                      */
+    /*********************************************************************/
+    private static final int IMPTABLEVELS_COLUMNS = _B + 2;
+    private static final int IMPTABLEVELS_RES = IMPTABLEVELS_COLUMNS - 1;
+    private static short GetState(byte cell) { return (short)(cell & 0x0f); }
+    private static short GetAction(byte cell) { return (short)(cell >> 4); }
+
+    private static class ImpTabPair {
+        byte[][][] imptab;
+        short[][] impact;
+
+        ImpTabPair(byte[][] table1, byte[][] table2,
+                   short[] act1, short[] act2) {
+            imptab = new byte[][][] {table1, table2};
+            impact = new short[][] {act1, act2};
+        }
+    }
+
+    /*********************************************************************/
+    /*                                                                   */
+    /*      LEVELS  STATE  TABLES                                        */
+    /*                                                                   */
+    /* In all levels state tables,                                       */
+    /*      - state 0 is the initial state                               */
+    /*      - the Res column is the increment to add to the text level   */
+    /*        for this property sequence.                                */
+    /*                                                                   */
+    /* The impact arrays for each table of a pair map the local action   */
+    /* numbers of the table to the total list of actions. For instance,  */
+    /* action 2 in a given table corresponds to the action number which  */
+    /* appears in entry [2] of the impact array for that table.          */
+    /* The first entry of all impact arrays must be 0.                   */
+    /*                                                                   */
+    /* Action 1: init conditional sequence                               */
+    /*        2: prepend conditional sequence to current sequence        */
+    /*        3: set ON sequence to new level - 1                        */
+    /*        4: init EN/AN/ON sequence                                  */
+    /*        5: fix EN/AN/ON sequence followed by R                     */
+    /*        6: set previous level sequence to level 2                  */
+    /*                                                                   */
+    /* Notes:                                                            */
+    /*  1) These tables are used in processPropertySeq(). The input      */
+    /*     is property sequences as determined by resolveImplicitLevels. */
+    /*  2) Most such property sequences are processed immediately        */
+    /*     (levels are assigned).                                        */
+    /*  3) However, some sequences cannot be assigned a final level till */
+    /*     one or more following sequences are received. For instance,   */
+    /*     ON following an R sequence within an even-level paragraph.    */
+    /*     If the following sequence is R, the ON sequence will be       */
+    /*     assigned basic run level+1, and so will the R sequence.       */
+    /*  4) S is generally handled like ON, since its level will be fixed */
+    /*     to paragraph level in adjustWSLevels().                       */
+    /*                                                                   */
+
+    private static final byte impTabL_DEFAULT[][] = /* Even paragraph level */
+        /*  In this table, conditional sequences receive the lower possible level
+            until proven otherwise.
+        */
+    {
+        /*                         L,     R,    EN,    AN,    ON,     S,     B, Res */
+        /* 0 : init       */ {     0,     1,     0,     2,     0,     0,     0,  0 },
+        /* 1 : R          */ {     0,     1,     3,     3,  0x14,  0x14,     0,  1 },
+        /* 2 : AN         */ {     0,     1,     0,     2,  0x15,  0x15,     0,  2 },
+        /* 3 : R+EN/AN    */ {     0,     1,     3,     3,  0x14,  0x14,     0,  2 },
+        /* 4 : R+ON       */ {     0,  0x21,  0x33,  0x33,     4,     4,     0,  0 },
+        /* 5 : AN+ON      */ {     0,  0x21,     0,  0x32,     5,     5,     0,  0 }
+    };
+
+    private static final byte impTabR_DEFAULT[][] = /* Odd  paragraph level */
+        /*  In this table, conditional sequences receive the lower possible level
+            until proven otherwise.
+        */
+    {
+        /*                         L,     R,    EN,    AN,    ON,     S,     B, Res */
+        /* 0 : init       */ {     1,     0,     2,     2,     0,     0,     0,  0 },
+        /* 1 : L          */ {     1,     0,     1,     3,  0x14,  0x14,     0,  1 },
+        /* 2 : EN/AN      */ {     1,     0,     2,     2,     0,     0,     0,  1 },
+        /* 3 : L+AN       */ {     1,     0,     1,     3,     5,     5,     0,  1 },
+        /* 4 : L+ON       */ {  0x21,     0,  0x21,     3,     4,     4,     0,  0 },
+        /* 5 : L+AN+ON    */ {     1,     0,     1,     3,     5,     5,     0,  0 }
+    };
+
+    private static final short[] impAct0 = {0,1,2,3,4};
+
+    private static final ImpTabPair impTab_DEFAULT = new ImpTabPair(
+            impTabL_DEFAULT, impTabR_DEFAULT, impAct0, impAct0);
+
+    private static final byte impTabL_NUMBERS_SPECIAL[][] = { /* Even paragraph level */
+        /* In this table, conditional sequences receive the lower possible
+           level until proven otherwise.
+        */
+        /*                         L,     R,    EN,    AN,    ON,     S,     B, Res */
+        /* 0 : init       */ {     0,     2,  0x11,  0x11,     0,     0,     0,  0 },
+        /* 1 : L+EN/AN    */ {     0,  0x42,     1,     1,     0,     0,     0,  0 },
+        /* 2 : R          */ {     0,     2,     4,     4,  0x13,  0x13,     0,  1 },
+        /* 3 : R+ON       */ {     0,  0x22,  0x34,  0x34,     3,     3,     0,  0 },
+        /* 4 : R+EN/AN    */ {     0,     2,     4,     4,  0x13,  0x13,     0,  2 }
+    };
+    private static final ImpTabPair impTab_NUMBERS_SPECIAL = new ImpTabPair(
+            impTabL_NUMBERS_SPECIAL, impTabR_DEFAULT, impAct0, impAct0);
+
+    private static final byte impTabL_GROUP_NUMBERS_WITH_R[][] = {
+        /* In this table, EN/AN+ON sequences receive levels as if associated with R
+           until proven that there is L or sor/eor on both sides. AN is handled like EN.
+        */
+        /*                         L,     R,    EN,    AN,    ON,     S,     B, Res */
+        /* 0 init         */ {     0,     3,  0x11,  0x11,     0,     0,     0,  0 },
+        /* 1 EN/AN        */ {  0x20,     3,     1,     1,     2,  0x20,  0x20,  2 },
+        /* 2 EN/AN+ON     */ {  0x20,     3,     1,     1,     2,  0x20,  0x20,  1 },
+        /* 3 R            */ {     0,     3,     5,     5,  0x14,     0,     0,  1 },
+        /* 4 R+ON         */ {  0x20,     3,     5,     5,     4,  0x20,  0x20,  1 },
+        /* 5 R+EN/AN      */ {     0,     3,     5,     5,  0x14,     0,     0,  2 }
+    };
+    private static final byte impTabR_GROUP_NUMBERS_WITH_R[][] = {
+        /*  In this table, EN/AN+ON sequences receive levels as if associated with R
+            until proven that there is L on both sides. AN is handled like EN.
+        */
+        /*                         L,     R,    EN,    AN,    ON,     S,     B, Res */
+        /* 0 init         */ {     2,     0,     1,     1,     0,     0,     0,  0 },
+        /* 1 EN/AN        */ {     2,     0,     1,     1,     0,     0,     0,  1 },
+        /* 2 L            */ {     2,     0,  0x14,  0x14,  0x13,     0,     0,  1 },
+        /* 3 L+ON         */ {  0x22,     0,     4,     4,     3,     0,     0,  0 },
+        /* 4 L+EN/AN      */ {  0x22,     0,     4,     4,     3,     0,     0,  1 }
+    };
+    private static final ImpTabPair impTab_GROUP_NUMBERS_WITH_R = new
+            ImpTabPair(impTabL_GROUP_NUMBERS_WITH_R,
+                       impTabR_GROUP_NUMBERS_WITH_R, impAct0, impAct0);
+
+    private static final byte impTabL_INVERSE_NUMBERS_AS_L[][] = {
+        /* This table is identical to the Default LTR table except that EN and AN
+           are handled like L.
+        */
+        /*                         L,     R,    EN,    AN,    ON,     S,     B, Res */
+        /* 0 : init       */ {     0,     1,     0,     0,     0,     0,     0,  0 },
+        /* 1 : R          */ {     0,     1,     0,     0,  0x14,  0x14,     0,  1 },
+        /* 2 : AN         */ {     0,     1,     0,     0,  0x15,  0x15,     0,  2 },
+        /* 3 : R+EN/AN    */ {     0,     1,     0,     0,  0x14,  0x14,     0,  2 },
+        /* 4 : R+ON       */ {  0x20,     1,  0x20,  0x20,     4,     4,  0x20,  1 },
+        /* 5 : AN+ON      */ {  0x20,     1,  0x20,  0x20,     5,     5,  0x20,  1 }
+    };
+    private static final byte impTabR_INVERSE_NUMBERS_AS_L[][] = {
+        /* This table is identical to the Default RTL table except that EN and AN
+           are handled like L.
+        */
+        /*                         L,     R,    EN,    AN,    ON,     S,     B, Res */
+        /* 0 : init       */ {     1,     0,     1,     1,     0,     0,     0,  0 },
+        /* 1 : L          */ {     1,     0,     1,     1,  0x14,  0x14,     0,  1 },
+        /* 2 : EN/AN      */ {     1,     0,     1,     1,     0,     0,     0,  1 },
+        /* 3 : L+AN       */ {     1,     0,     1,     1,     5,     5,     0,  1 },
+        /* 4 : L+ON       */ {  0x21,     0,  0x21,  0x21,     4,     4,     0,  0 },
+        /* 5 : L+AN+ON    */ {     1,     0,     1,     1,     5,     5,     0,  0 }
+    };
+    private static final ImpTabPair impTab_INVERSE_NUMBERS_AS_L = new ImpTabPair
+            (impTabL_INVERSE_NUMBERS_AS_L, impTabR_INVERSE_NUMBERS_AS_L,
+             impAct0, impAct0);
+
+    private static final byte impTabR_INVERSE_LIKE_DIRECT[][] = {  /* Odd  paragraph level */
+        /*  In this table, conditional sequences receive the lower possible level
+            until proven otherwise.
+        */
+        /*                         L,     R,    EN,    AN,    ON,     S,     B, Res */
+        /* 0 : init       */ {     1,     0,     2,     2,     0,     0,     0,  0 },
+        /* 1 : L          */ {     1,     0,     1,     2,  0x13,  0x13,     0,  1 },
+        /* 2 : EN/AN      */ {     1,     0,     2,     2,     0,     0,     0,  1 },
+        /* 3 : L+ON       */ {  0x21,  0x30,     6,     4,     3,     3,  0x30,  0 },
+        /* 4 : L+ON+AN    */ {  0x21,  0x30,     6,     4,     5,     5,  0x30,  3 },
+        /* 5 : L+AN+ON    */ {  0x21,  0x30,     6,     4,     5,     5,  0x30,  2 },
+        /* 6 : L+ON+EN    */ {  0x21,  0x30,     6,     4,     3,     3,  0x30,  1 }
+    };
+    private static final short[] impAct1 = {0,1,13,14};
+    private static final ImpTabPair impTab_INVERSE_LIKE_DIRECT = new ImpTabPair(
+            impTabL_DEFAULT, impTabR_INVERSE_LIKE_DIRECT, impAct0, impAct1);
+
+    private static final byte impTabL_INVERSE_LIKE_DIRECT_WITH_MARKS[][] = {
+        /* The case handled in this table is (visually):  R EN L
+         */
+        /*                         L,     R,    EN,    AN,    ON,     S,     B, Res */
+        /* 0 : init       */ {     0,  0x63,     0,     1,     0,     0,     0,  0 },
+        /* 1 : L+AN       */ {     0,  0x63,     0,     1,  0x12,  0x30,     0,  4 },
+        /* 2 : L+AN+ON    */ {  0x20,  0x63,  0x20,     1,     2,  0x30,  0x20,  3 },
+        /* 3 : R          */ {     0,  0x63,  0x55,  0x56,  0x14,  0x30,     0,  3 },
+        /* 4 : R+ON       */ {  0x30,  0x43,  0x55,  0x56,     4,  0x30,  0x30,  3 },
+        /* 5 : R+EN       */ {  0x30,  0x43,     5,  0x56,  0x14,  0x30,  0x30,  4 },
+        /* 6 : R+AN       */ {  0x30,  0x43,  0x55,     6,  0x14,  0x30,  0x30,  4 }
+    };
+    private static final byte impTabR_INVERSE_LIKE_DIRECT_WITH_MARKS[][] = {
+        /* The cases handled in this table are (visually):  R EN L
+                                                            R L AN L
+        */
+        /*                         L,     R,    EN,    AN,    ON,     S,     B, Res */
+        /* 0 : init       */ {  0x13,     0,     1,     1,     0,     0,     0,  0 },
+        /* 1 : R+EN/AN    */ {  0x23,     0,     1,     1,     2,  0x40,     0,  1 },
+        /* 2 : R+EN/AN+ON */ {  0x23,     0,     1,     1,     2,  0x40,     0,  0 },
+        /* 3 : L          */ {     3,     0,     3,  0x36,  0x14,  0x40,     0,  1 },
+        /* 4 : L+ON       */ {  0x53,  0x40,     5,  0x36,     4,  0x40,  0x40,  0 },
+        /* 5 : L+ON+EN    */ {  0x53,  0x40,     5,  0x36,     4,  0x40,  0x40,  1 },
+        /* 6 : L+AN       */ {  0x53,  0x40,     6,     6,     4,  0x40,  0x40,  3 }
+    };
+    private static final short[] impAct2 = {0,1,2,5,6,7,8};
+    private static final short[] impAct3 = {0,1,9,10,11,12};
+    private static final ImpTabPair impTab_INVERSE_LIKE_DIRECT_WITH_MARKS =
+            new ImpTabPair(impTabL_INVERSE_LIKE_DIRECT_WITH_MARKS,
+                           impTabR_INVERSE_LIKE_DIRECT_WITH_MARKS, impAct2, impAct3);
+
+    private static final ImpTabPair impTab_INVERSE_FOR_NUMBERS_SPECIAL = new ImpTabPair(
+            impTabL_NUMBERS_SPECIAL, impTabR_INVERSE_LIKE_DIRECT, impAct0, impAct1);
+
+    private static final byte impTabL_INVERSE_FOR_NUMBERS_SPECIAL_WITH_MARKS[][] = {
+        /*  The case handled in this table is (visually):  R EN L
+        */
+        /*                         L,     R,    EN,    AN,    ON,     S,     B, Res */
+        /* 0 : init       */ {     0,  0x62,     1,     1,     0,     0,     0,  0 },
+        /* 1 : L+EN/AN    */ {     0,  0x62,     1,     1,     0,  0x30,     0,  4 },
+        /* 2 : R          */ {     0,  0x62,  0x54,  0x54,  0x13,  0x30,     0,  3 },
+        /* 3 : R+ON       */ {  0x30,  0x42,  0x54,  0x54,     3,  0x30,  0x30,  3 },
+        /* 4 : R+EN/AN    */ {  0x30,  0x42,     4,     4,  0x13,  0x30,  0x30,  4 }
+    };
+    private static final ImpTabPair impTab_INVERSE_FOR_NUMBERS_SPECIAL_WITH_MARKS = new
+            ImpTabPair(impTabL_INVERSE_FOR_NUMBERS_SPECIAL_WITH_MARKS,
+                       impTabR_INVERSE_LIKE_DIRECT_WITH_MARKS, impAct2, impAct3);
+
+    private static class LevState {
+        byte[][] impTab;                /* level table pointer          */
+        short[] impAct;                 /* action map array             */
+        int startON;                    /* start of ON sequence         */
+        int startL2EN;                  /* start of level 2 sequence    */
+        int lastStrongRTL;              /* index of last found R or AL  */
+        int runStart;                   /* start position of the run    */
+        short state;                    /* current state                */
+        byte runLevel;                  /* run level before implicit solving */
+    }
+
+    /*------------------------------------------------------------------------*/
+
+    static final int FIRSTALLOC = 10;
+    /*
+     *  param pos:     position where to insert
+     *  param flag:    one of LRM_BEFORE, LRM_AFTER, RLM_BEFORE, RLM_AFTER
+     */
+    private void addPoint(int pos, int flag)
+    {
+        Point point = new Point();
+
+        int len = insertPoints.points.length;
+        if (len == 0) {
+            insertPoints.points = new Point[FIRSTALLOC];
+            len = FIRSTALLOC;
+        }
+        if (insertPoints.size >= len) { /* no room for new point */
+            Point[] savePoints = insertPoints.points;
+            insertPoints.points = new Point[len * 2];
+            System.arraycopy(savePoints, 0, insertPoints.points, 0, len);
+        }
+        point.pos = pos;
+        point.flag = flag;
+        insertPoints.points[insertPoints.size] = point;
+        insertPoints.size++;
+    }
+
+    private void setLevelsOutsideIsolates(int start, int limit, byte level)
+    {
+        byte dirProp;
+        int  isolateCount = 0, k;
+        for (k = start; k < limit; k++) {
+            dirProp = dirProps[k];
+            if (dirProp == PDI)
+                isolateCount--;
+            if (isolateCount == 0) {
+                levels[k] = level;
+            }
+            if (dirProp == LRI || dirProp == RLI)
+                isolateCount++;
+        }
+    }
+
+    /* perform rules (Wn), (Nn), and (In) on a run of the text ------------------ */
+
+    /*
+     * This implementation of the (Wn) rules applies all rules in one pass.
+     * In order to do so, it needs a look-ahead of typically 1 character
+     * (except for W5: sequences of ET) and keeps track of changes
+     * in a rule Wp that affect a later Wq (p= 0) {
+                    addPoint(levState.startL2EN, LRM_BEFORE);
+                }
+                levState.startL2EN = -1;  /* not within previous if since could also be -2 */
+                /* check if we had any relevant EN/AN after R/AL */
+                if ((insertPoints.points.length == 0) ||
+                        (insertPoints.size <= insertPoints.confirmed)) {
+                    /* nothing, just clean up */
+                    levState.lastStrongRTL = -1;
+                    /* check if we have a pending conditional segment */
+                    level = impTab[oldStateSeq][IMPTABLEVELS_RES];
+                    if ((level & 1) != 0 && levState.startON > 0) { /* after ON */
+                        start = levState.startON;   /* reset to basic run level */
+                    }
+                    if (_prop == _S) {              /* add LRM before S */
+                        addPoint(start0, LRM_BEFORE);
+                        insertPoints.confirmed = insertPoints.size;
+                    }
+                    break;
+                }
+                /* reset previous RTL cont to level for LTR text */
+                for (k = levState.lastStrongRTL + 1; k < start0; k++) {
+                    /* reset odd level, leave runLevel+2 as is */
+                    levels[k] = (byte)((levels[k] - 2) & ~1);
+                }
+                /* mark insert points as confirmed */
+                insertPoints.confirmed = insertPoints.size;
+                levState.lastStrongRTL = -1;
+                if (_prop == _S) {           /* add LRM before S */
+                    addPoint(start0, LRM_BEFORE);
+                    insertPoints.confirmed = insertPoints.size;
+                }
+                break;
+
+            case 6:                     /* R/AL after possible relevant EN/AN */
+                /* just clean up */
+                if (insertPoints.points.length > 0)
+                    /* remove all non confirmed insert points */
+                    insertPoints.size = insertPoints.confirmed;
+                levState.startON = -1;
+                levState.startL2EN = -1;
+                levState.lastStrongRTL = limit - 1;
+                break;
+
+            case 7:                     /* EN/AN after R/AL + possible cont */
+                /* check for real AN */
+
+                if ((_prop == _AN) && (dirProps[start0] == AN) &&
+                (reorderingMode != REORDER_INVERSE_FOR_NUMBERS_SPECIAL))
+                {
+                    /* real AN */
+                    if (levState.startL2EN == -1) { /* if no relevant EN already found */
+                        /* just note the rightmost digit as a strong RTL */
+                        levState.lastStrongRTL = limit - 1;
+                        break;
+                    }
+                    if (levState.startL2EN >= 0)  { /* after EN, no AN */
+                        addPoint(levState.startL2EN, LRM_BEFORE);
+                        levState.startL2EN = -2;
+                    }
+                    /* note AN */
+                    addPoint(start0, LRM_BEFORE);
+                    break;
+                }
+                /* if first EN/AN after R/AL */
+                if (levState.startL2EN == -1) {
+                    levState.startL2EN = start0;
+                }
+                break;
+
+            case 8:                     /* note location of latest R/AL */
+                levState.lastStrongRTL = limit - 1;
+                levState.startON = -1;
+                break;
+
+            case 9:                     /* L after R+ON/EN/AN */
+                /* include possible adjacent number on the left */
+                for (k = start0-1; k >= 0 && ((levels[k] & 1) == 0); k--) {
+                }
+                if (k >= 0) {
+                    addPoint(k, RLM_BEFORE);    /* add RLM before */
+                    insertPoints.confirmed = insertPoints.size; /* confirm it */
+                }
+                levState.startON = start0;
+                break;
+
+            case 10:                    /* AN after L */
+                /* AN numbers between L text on both sides may be trouble. */
+                /* tentatively bracket with LRMs; will be confirmed if followed by L */
+                addPoint(start0, LRM_BEFORE);   /* add LRM before */
+                addPoint(start0, LRM_AFTER);    /* add LRM after  */
+                break;
+
+            case 11:                    /* R after L+ON/EN/AN */
+                /* false alert, infirm LRMs around previous AN */
+                insertPoints.size=insertPoints.confirmed;
+                if (_prop == _S) {          /* add RLM before S */
+                    addPoint(start0, RLM_BEFORE);
+                    insertPoints.confirmed = insertPoints.size;
+                }
+                break;
+
+            case 12:                    /* L after L+ON/AN */
+                level = (byte)(levState.runLevel + addLevel);
+                for (k=levState.startON; k < start0; k++) {
+                    if (levels[k] < level) {
+                        levels[k] = level;
+                    }
+                }
+                insertPoints.confirmed = insertPoints.size;   /* confirm inserts */
+                levState.startON = start0;
+                break;
+
+            case 13:                    /* L after L+ON+EN/AN/ON */
+                level = levState.runLevel;
+                for (k = start0-1; k >= levState.startON; k--) {
+                    if (levels[k] == level+3) {
+                        while (levels[k] == level+3) {
+                            levels[k--] -= 2;
+                        }
+                        while (levels[k] == level) {
+                            k--;
+                        }
+                    }
+                    if (levels[k] == level+2) {
+                        levels[k] = level;
+                        continue;
+                    }
+                    levels[k] = (byte)(level+1);
+                }
+                break;
+
+            case 14:                    /* R after L+ON+EN/AN/ON */
+                level = (byte)(levState.runLevel+1);
+                for (k = start0-1; k >= levState.startON; k--) {
+                    if (levels[k] > level) {
+                        levels[k] -= 2;
+                    }
+                }
+                break;
+
+            default:                        /* we should never get here */
+                throw new IllegalStateException("Internal ICU error in processPropertySeq");
+            }
+        }
+        if ((addLevel) != 0 || (start < start0)) {
+            level = (byte)(levState.runLevel + addLevel);
+            if (start >= levState.runStart) {
+                for (k = start; k < limit; k++) {
+                    levels[k] = level;
+                }
+            } else {
+                setLevelsOutsideIsolates(start, limit, level);
+            }
+        }
+    }
+
+    private void resolveImplicitLevels(int start, int limit, short sor, short eor)
+    {
+        byte dirProp;
+        LevState levState = new LevState();
+        int i, start1, start2;
+        short oldStateImp, stateImp, actionImp;
+        short gprop, resProp, cell;
+        boolean inverseRTL;
+        short nextStrongProp = R;
+        int nextStrongPos = -1;
+
+        /* check for RTL inverse Bidi mode */
+        /* FOOD FOR THOUGHT: in case of RTL inverse Bidi, it would make sense to
+         * loop on the text characters from end to start.
+         * This would need a different properties state table (at least different
+         * actions) and different levels state tables (maybe very similar to the
+         * LTR corresponding ones.
+         */
+        inverseRTL=((start0) &&
+                    (reorderingMode == REORDER_INVERSE_LIKE_DIRECT  ||
+                     reorderingMode == REORDER_INVERSE_FOR_NUMBERS_SPECIAL));
+        /* initialize for property and levels state table */
+        levState.startL2EN = -1;        /* used for INVERSE_LIKE_DIRECT_WITH_MARKS */
+        levState.lastStrongRTL = -1;    /* used for INVERSE_LIKE_DIRECT_WITH_MARKS */
+        levState.runStart = start;
+        levState.runLevel = levels[start];
+        levState.impTab = impTabPair.imptab[levState.runLevel & 1];
+        levState.impAct = impTabPair.impact[levState.runLevel & 1];
+
+        /* The isolates[] entries contain enough information to
+           resume the bidi algorithm in the same state as it was
+           when it was interrupted by an isolate sequence. */
+        if (dirProps[start] == PDI) {
+            levState.startON = isolates[isolateCount].startON;
+            start1 = isolates[isolateCount].start1;
+            stateImp = isolates[isolateCount].stateImp;
+            levState.state = isolates[isolateCount].state;
+            isolateCount--;
+        } else {
+            levState.startON = -1;
+            start1 = start;
+            if (dirProps[start] == NSM)
+              stateImp = (short)(1 + sor);
+            else
+                stateImp = 0;
+            levState.state = 0;
+            processPropertySeq(levState, sor, start, start);
+        }
+        start2 = start;                 /* to make the Java compiler happy */
+
+        for (i = start; i <= limit; i++) {
+            if (i >= limit) {
+                int k;
+                for (k = limit - 1;
+                     k > start &&
+                         (DirPropFlag(dirProps[k]) & MASK_BN_EXPLICIT) != 0;
+                     k--);
+                dirProp = dirProps[k];
+                if (dirProp == LRI || dirProp == RLI)
+                    break;  /* no forced closing for sequence ending with LRI/RLI */
+                gprop = eor;
+            } else {
+                byte prop, prop1;
+                prop = dirProps[i];
+                if (prop == B)
+                    isolateCount = -1;  /* current isolates stack entry == none */
+                if (inverseRTL) {
+                    if (prop == AL) {
+                        /* AL before EN does not make it AN */
+                        prop = R;
+                    } else if (prop == EN) {
+                        if (nextStrongPos <= i) {
+                            /* look for next strong char (L/R/AL) */
+                            int j;
+                            nextStrongProp = R;     /* set default */
+                            nextStrongPos = limit;
+                            for (j = i+1; j < limit; j++) {
+                                prop1 = dirProps[j];
+                                if (prop1 == L || prop1 == R || prop1 == AL) {
+                                    nextStrongProp = prop1;
+                                    nextStrongPos = j;
+                                    break;
+                                }
+                            }
+                        }
+                        if (nextStrongProp == AL) {
+                            prop = AN;
+                        }
+                    }
+                }
+                gprop = groupProp[prop];
+            }
+            oldStateImp = stateImp;
+            cell = impTabProps[oldStateImp][gprop];
+            stateImp = GetStateProps(cell);     /* isolate the new state */
+            actionImp = GetActionProps(cell);   /* isolate the action */
+            if ((i == limit) && (actionImp == 0)) {
+                /* there is an unprocessed sequence if its property == eor   */
+                actionImp = 1;                  /* process the last sequence */
+            }
+            if (actionImp != 0) {
+                resProp = impTabProps[oldStateImp][IMPTABPROPS_RES];
+                switch (actionImp) {
+                case 1:             /* process current seq1, init new seq1 */
+                    processPropertySeq(levState, resProp, start1, i);
+                    start1 = i;
+                    break;
+                case 2:             /* init new seq2 */
+                    start2 = i;
+                    break;
+                case 3:             /* process seq1, process seq2, init new seq1 */
+                    processPropertySeq(levState, resProp, start1, start2);
+                    processPropertySeq(levState, _ON, start2, i);
+                    start1 = i;
+                    break;
+                case 4:             /* process seq1, set seq1=seq2, init new seq2 */
+                    processPropertySeq(levState, resProp, start1, start2);
+                    start1 = start2;
+                    start2 = i;
+                    break;
+                default:            /* we should never get here */
+                    throw new IllegalStateException("Internal ICU error in resolveImplicitLevels");
+                }
+            }
+        }
+
+        /* look for the last char not a BN or LRE/RLE/LRO/RLO/PDF */
+        for (i = limit - 1;
+             i > start &&
+                 (DirPropFlag(dirProps[i]) & MASK_BN_EXPLICIT) != 0;
+             i--);
+        dirProp = dirProps[i];
+        if ((dirProp == LRI || dirProp == RLI) && limit < length) {
+            isolateCount++;
+            if (isolates[isolateCount] == null)
+                isolates[isolateCount] = new Isolate();
+            isolates[isolateCount].stateImp = stateImp;
+            isolates[isolateCount].state = levState.state;
+            isolates[isolateCount].start1 = start1;
+            isolates[isolateCount].startON = levState.startON;
+        }
+        else
+            processPropertySeq(levState, eor, limit, limit);
+    }
+
+    /* perform (L1) and (X9) ---------------------------------------------------- */
+
+    /*
+     * Reset the embedding levels for some non-graphic characters (L1).
+     * This method also sets appropriate levels for BN, and
+     * explicit embedding types that are supposed to have been removed
+     * from the paragraph in (X9).
+     */
+    private void adjustWSLevels() {
+        int i;
+
+        if ((flags & MASK_WS) != 0) {
+            int flag;
+            i = trailingWSStart;
+            while (i > 0) {
+                /* reset a sequence of WS/BN before eop and B/S to the paragraph paraLevel */
+                while (i > 0 && ((flag = DirPropFlag(dirProps[--i])) & MASK_WS) != 0) {
+                    if (orderParagraphsLTR && (flag & DirPropFlag(B)) != 0) {
+                        levels[i] = 0;
+                    } else {
+                        levels[i] = GetParaLevelAt(i);
+                    }
+                }
+
+                /* reset BN to the next character's paraLevel until B/S, which restarts above loop */
+                /* here, i+1 is guaranteed to be  0) {
+                    flag = DirPropFlag(dirProps[--i]);
+                    if ((flag & MASK_BN_EXPLICIT) != 0) {
+                        levels[i] = levels[i + 1];
+                    } else if (orderParagraphsLTR && (flag & DirPropFlag(B)) != 0) {
+                        levels[i] = 0;
+                        break;
+                    } else if ((flag & MASK_B_S) != 0){
+                        levels[i] = GetParaLevelAt(i);
+                        break;
+                    }
+                }
+            }
+        }
+    }
+
+    private void setParaSuccess() {
+        paraBidi = this;                /* mark successful setPara */
+    }
+
+    private int Bidi_Min(int x, int y) {
+        return x < y ? x : y;
+    }
+
+    private int Bidi_Abs(int x) {
+        return x >= 0 ? x : -x;
+    }
+
+    void setParaRunsOnly(char[] parmText, byte parmParaLevel) {
+        int[] visualMap;
+        String visualText;
+        int saveLength, saveTrailingWSStart;
+        byte[] saveLevels;
+        byte saveDirection;
+        int i, j, visualStart, logicalStart,
+            oldRunCount, runLength, addedRuns, insertRemove,
+            start, limit, step, indexOddBit, logicalPos,
+            index, index1;
+        int saveOptions;
+
+        reorderingMode = REORDER_DEFAULT;
+        int parmLength = parmText.length;
+        if (parmLength == 0) {
+            setPara(parmText, parmParaLevel, null);
+            reorderingMode = REORDER_RUNS_ONLY;
+            return;
+        }
+        /* obtain memory for mapping table and visual text */
+        saveOptions = reorderingOptions;
+        if ((saveOptions & OPTION_INSERT_MARKS) > 0) {
+            reorderingOptions &= ~OPTION_INSERT_MARKS;
+            reorderingOptions |= OPTION_REMOVE_CONTROLS;
+        }
+        parmParaLevel &= 1;             /* accept only 0 or 1 */
+        setPara(parmText, parmParaLevel, null);
+        /* we cannot access directly levels since it is not yet set if
+         * direction is not MIXED
+         */
+        saveLevels = new byte[this.length];
+        System.arraycopy(getLevels(), 0, saveLevels, 0, this.length);
+        saveTrailingWSStart = trailingWSStart;
+
+        /* FOOD FOR THOUGHT: instead of writing the visual text, we could use
+         * the visual map and the dirProps array to drive the second call
+         * to setPara (but must make provision for possible removal of
+         * Bidi controls.  Alternatively, only use the dirProps array via
+         * customized classifier callback.
+         */
+        visualText = writeReordered(DO_MIRRORING);
+        visualMap = getVisualMap();
+        this.reorderingOptions = saveOptions;
+        saveLength = this.length;
+        saveDirection=this.direction;
+
+        this.reorderingMode = REORDER_INVERSE_LIKE_DIRECT;
+        parmParaLevel ^= 1;
+        setPara(visualText, parmParaLevel, null);
+        BidiLine.getRuns(this);
+        /* check if some runs must be split, count how many splits */
+        addedRuns = 0;
+        oldRunCount = this.runCount;
+        visualStart = 0;
+        for (i = 0; i < oldRunCount; i++, visualStart += runLength) {
+            runLength = runs[i].limit - visualStart;
+            if (runLength < 2) {
+                continue;
+            }
+            logicalStart = runs[i].start;
+            for (j = logicalStart+1; j < logicalStart+runLength; j++) {
+                index = visualMap[j];
+                index1 = visualMap[j-1];
+                if ((Bidi_Abs(index-index1)!=1) || (saveLevels[index]!=saveLevels[index1])) {
+                    addedRuns++;
+                }
+            }
+        }
+        if (addedRuns > 0) {
+            getRunsMemory(oldRunCount + addedRuns);
+            if (runCount == 1) {
+                /* because we switch from UBiDi.simpleRuns to UBiDi.runs */
+                runsMemory[0] = runs[0];
+            } else {
+                System.arraycopy(runs, 0, runsMemory, 0, runCount);
+            }
+            runs = runsMemory;
+            runCount += addedRuns;
+            for (i = oldRunCount; i < runCount; i++) {
+                if (runs[i] == null) {
+                    runs[i] = new BidiRun(0, 0, (byte)0);
+                }
+            }
+        }
+        /* split runs which are not consecutive in source text */
+        int newI;
+        for (i = oldRunCount-1; i >= 0; i--) {
+            newI = i + addedRuns;
+            runLength = i==0 ? runs[0].limit :
+                               runs[i].limit - runs[i-1].limit;
+            logicalStart = runs[i].start;
+            indexOddBit = runs[i].level & 1;
+            if (runLength < 2) {
+                if (addedRuns > 0) {
+                    runs[newI].copyFrom(runs[i]);
+                }
+                logicalPos = visualMap[logicalStart];
+                runs[newI].start = logicalPos;
+                runs[newI].level = (byte)(saveLevels[logicalPos] ^ indexOddBit);
+                continue;
+            }
+            if (indexOddBit > 0) {
+                start = logicalStart;
+                limit = logicalStart + runLength - 1;
+                step = 1;
+            } else {
+                start = logicalStart + runLength - 1;
+                limit = logicalStart;
+                step = -1;
+            }
+            for (j = start; j != limit; j += step) {
+                index = visualMap[j];
+                index1 = visualMap[j+step];
+                if ((Bidi_Abs(index-index1)!=1) || (saveLevels[index]!=saveLevels[index1])) {
+                    logicalPos = Bidi_Min(visualMap[start], index);
+                    runs[newI].start = logicalPos;
+                    runs[newI].level = (byte)(saveLevels[logicalPos] ^ indexOddBit);
+                    runs[newI].limit = runs[i].limit;
+                    runs[i].limit -= Bidi_Abs(j - start) + 1;
+                    insertRemove = runs[i].insertRemove & (LRM_AFTER|RLM_AFTER);
+                    runs[newI].insertRemove = insertRemove;
+                    runs[i].insertRemove &= ~insertRemove;
+                    start = j + step;
+                    addedRuns--;
+                    newI--;
+                }
+            }
+            if (addedRuns > 0) {
+                runs[newI].copyFrom(runs[i]);
+            }
+            logicalPos = Bidi_Min(visualMap[start], visualMap[limit]);
+            runs[newI].start = logicalPos;
+            runs[newI].level = (byte)(saveLevels[logicalPos] ^ indexOddBit);
+        }
+
+    cleanup1:
+        /* restore initial paraLevel */
+        this.paraLevel ^= 1;
+    cleanup2:
+        /* restore real text */
+        this.text = parmText;
+        this.length = saveLength;
+        this.originalLength = parmLength;
+        this.direction=saveDirection;
+        this.levels = saveLevels;
+        this.trailingWSStart = saveTrailingWSStart;
+        if (runCount > 1) {
+            this.direction = MIXED;
+        }
+    cleanup3:
+        this.reorderingMode = REORDER_RUNS_ONLY;
+    }
+
+    /**
+     * Perform the Unicode Bidi algorithm. It is defined in the
+     * Unicode Standard Annex #9,
+     * version 13,
+     * also described in The Unicode Standard, Version 4.0 .

+     *
+     * This method takes a piece of plain text containing one or more paragraphs,
+     * with or without externally specified embedding levels from styled
+     * text and computes the left-right-directionality of each character.

+     *
+     * If the entire text is all of the same directionality, then
+     * the method may not perform all the steps described by the algorithm,
+     * i.e., some levels may not be the same as if all steps were performed.
+     * This is not relevant for unidirectional text.

+     * For example, in pure LTR text with numbers the numbers would get
+     * a resolved level of 2 higher than the surrounding text according to
+     * the algorithm. This implementation may set all resolved levels to
+     * the same value in such a case.

+     *
+     * The text can be composed of multiple paragraphs. Occurrence of a block
+     * separator in the text terminates a paragraph, and whatever comes next starts
+     * a new paragraph. The exception to this rule is when a Carriage Return (CR)
+     * is followed by a Line Feed (LF). Both CR and LF are block separators, but
+     * in that case, the pair of characters is considered as terminating the
+     * preceding paragraph, and a new paragraph will be started by a character
+     * coming after the LF.
+     *
+     * Although the text is passed here as a String, it is
+     * stored internally as an array of characters. Therefore the
+     * documentation will refer to indexes of the characters in the text.
+     *
+     * @param text contains the text that the Bidi algorithm will be performed
+     *        on. This text can be retrieved with getText() or
+     *        getTextAsString.

+     *
+     * @param paraLevel specifies the default level for the text;
+     *        it is typically 0 (LTR) or 1 (RTL).
+     *        If the method shall determine the paragraph level from the text,
+     *        then paraLevel can be set to
+     *        either LEVEL_DEFAULT_LTR
+     *        or LEVEL_DEFAULT_RTL; if the text contains multiple
+     *        paragraphs, the paragraph level shall be determined separately for
+     *        each paragraph; if a paragraph does not include any strongly typed
+     *        character, then the desired default is used (0 for LTR or 1 for RTL).
+     *        Any other value between 0 and MAX_EXPLICIT_LEVEL
+     *        is also valid, with odd levels indicating RTL.
+     *
+     * @param embeddingLevels (in) may be used to preset the embedding and override levels,
+     *        ignoring characters like LRE and PDF in the text.
+     *        A level overrides the directional property of its corresponding
+     *        (same index) character if the level has the
+     *        LEVEL_OVERRIDE bit set.


+     *        Except for that bit, it must be
+     *        paraLevel<=embeddingLevels[]<=MAX_EXPLICIT_LEVEL,
+     *        with one exception: a level of zero may be specified for a
+     *        paragraph separator even if paraLevel>0 when multiple
+     *        paragraphs are submitted in the same call to setPara().


+     *        Caution: A reference to this array, not a copy
+     *        of the levels, will be stored in the Bidi object;
+     *        the embeddingLevels
+     *        should not be modified to avoid unexpected results on subsequent
+     *        Bidi operations. However, the setPara() and
+     *        setLine() methods may modify some or all of the
+     *        levels.


+     *        Note: the embeddingLevels array must
+     *        have one entry for each character in text.
+     *
+     * @throws IllegalArgumentException if the values in embeddingLevels are
+     *         not within the allowed range
+     *
+     * @see #LEVEL_DEFAULT_LTR
+     * @see #LEVEL_DEFAULT_RTL
+     * @see #LEVEL_OVERRIDE
+     * @see #MAX_EXPLICIT_LEVEL
+     * @stable ICU 3.8
+     */
+    void setPara(String text, byte paraLevel, byte[] embeddingLevels)
+    {
+        if (text == null) {
+            setPara(new char[0], paraLevel, embeddingLevels);
+        } else {
+            setPara(text.toCharArray(), paraLevel, embeddingLevels);
+        }
+    }
+
+    /**
+     * Perform the Unicode Bidi algorithm. It is defined in the
+     * Unicode Standard Annex #9,
+     * version 13,
+     * also described in The Unicode Standard, Version 4.0 .

+     *
+     * This method takes a piece of plain text containing one or more paragraphs,
+     * with or without externally specified embedding levels from styled
+     * text and computes the left-right-directionality of each character.

+     *
+     * If the entire text is all of the same directionality, then
+     * the method may not perform all the steps described by the algorithm,
+     * i.e., some levels may not be the same as if all steps were performed.
+     * This is not relevant for unidirectional text.

+     * For example, in pure LTR text with numbers the numbers would get
+     * a resolved level of 2 higher than the surrounding text according to
+     * the algorithm. This implementation may set all resolved levels to
+     * the same value in such a case.
+     *
+     * The text can be composed of multiple paragraphs. Occurrence of a block
+     * separator in the text terminates a paragraph, and whatever comes next starts
+     * a new paragraph. The exception to this rule is when a Carriage Return (CR)
+     * is followed by a Line Feed (LF). Both CR and LF are block separators, but
+     * in that case, the pair of characters is considered as terminating the
+     * preceding paragraph, and a new paragraph will be started by a character
+     * coming after the LF.
+     *
+     * The text is stored internally as an array of characters. Therefore the
+     * documentation will refer to indexes of the characters in the text.
+     *
+     * @param chars contains the text that the Bidi algorithm will be performed
+     *        on. This text can be retrieved with getText() or
+     *        getTextAsString.

+     *
+     * @param paraLevel specifies the default level for the text;
+     *        it is typically 0 (LTR) or 1 (RTL).
+     *        If the method shall determine the paragraph level from the text,
+     *        then paraLevel can be set to
+     *        either LEVEL_DEFAULT_LTR
+     *        or LEVEL_DEFAULT_RTL; if the text contains multiple
+     *        paragraphs, the paragraph level shall be determined separately for
+     *        each paragraph; if a paragraph does not include any strongly typed
+     *        character, then the desired default is used (0 for LTR or 1 for RTL).
+     *        Any other value between 0 and MAX_EXPLICIT_LEVEL
+     *        is also valid, with odd levels indicating RTL.
+     *
+     * @param embeddingLevels (in) may be used to preset the embedding and
+     *        override levels, ignoring characters like LRE and PDF in the text.
+     *        A level overrides the directional property of its corresponding
+     *        (same index) character if the level has the
+     *        LEVEL_OVERRIDE bit set.


+     *        Except for that bit, it must be
+     *        paraLevel<=embeddingLevels[]<=MAX_EXPLICIT_LEVEL,
+     *        with one exception: a level of zero may be specified for a
+     *        paragraph separator even if paraLevel>0 when multiple
+     *        paragraphs are submitted in the same call to setPara().


+     *        Caution: A reference to this array, not a copy
+     *        of the levels, will be stored in the Bidi object;
+     *        the embeddingLevels
+     *        should not be modified to avoid unexpected results on subsequent
+     *        Bidi operations. However, the setPara() and
+     *        setLine() methods may modify some or all of the
+     *        levels.


+     *        Note: the embeddingLevels array must
+     *        have one entry for each character in text.
+     *
+     * @throws IllegalArgumentException if the values in embeddingLevels are
+     *         not within the allowed range
+     *
+     * @see #LEVEL_DEFAULT_LTR
+     * @see #LEVEL_DEFAULT_RTL
+     * @see #LEVEL_OVERRIDE
+     * @see #MAX_EXPLICIT_LEVEL
+     * @stable ICU 3.8
+     */
+    void setPara(char[] chars, byte paraLevel, byte[] embeddingLevels)
+    {
+        /* check the argument values */
+        if (paraLevel < LEVEL_DEFAULT_LTR) {
+            verifyRange(paraLevel, 0, MAX_EXPLICIT_LEVEL + 1);
+        }
+        if (chars == null) {
+            chars = new char[0];
+        }
+
+        /* special treatment for RUNS_ONLY mode */
+        if (reorderingMode == REORDER_RUNS_ONLY) {
+            setParaRunsOnly(chars, paraLevel);
+            return;
+        }
+
+        /* initialize the Bidi object */
+        this.paraBidi = null;          /* mark unfinished setPara */
+        this.text = chars;
+        this.length = this.originalLength = this.resultLength = text.length;
+        this.paraLevel = paraLevel;
+        this.direction = (byte)(paraLevel & 1);
+        this.paraCount = 1;
+
+        /* Allocate zero-length arrays instead of setting to null here; then
+         * checks for null in various places can be eliminated.
+         */
+        dirProps = new byte[0];
+        levels = new byte[0];
+        runs = new BidiRun[0];
+        isGoodLogicalToVisualRunsMap = false;
+        insertPoints.size = 0;          /* clean up from last call */
+        insertPoints.confirmed = 0;     /* clean up from last call */
+
+        /*
+         * Save the original paraLevel if contextual; otherwise, set to 0.
+         */
+        defaultParaLevel = IsDefaultLevel(paraLevel) ? paraLevel : 0;
+
+        if (length == 0) {
+            /*
+             * For an empty paragraph, create a Bidi object with the paraLevel and
+             * the flags and the direction set but without allocating zero-length arrays.
+             * There is nothing more to do.
+             */
+            if (IsDefaultLevel(paraLevel)) {
+                this.paraLevel &= 1;
+                defaultParaLevel = 0;
+            }
+            flags = DirPropFlagLR(paraLevel);
+            runCount = 0;
+            paraCount = 0;
+            setParaSuccess();
+            return;
+        }
+
+        runCount = -1;
+
+        /*
+         * Get the directional properties,
+         * the flags bit-set, and
+         * determine the paragraph level if necessary.
+         */
+        getDirPropsMemory(length);
+        dirProps = dirPropsMemory;
+        getDirProps();
+        /* the processed length may have changed if OPTION_STREAMING is set */
+        trailingWSStart = length;  /* the levels[] will reflect the WS run */
+
+        /* are explicit levels specified? */
+        if (embeddingLevels == null) {
+            /* no: determine explicit levels according to the (Xn) rules */
+            getLevelsMemory(length);
+            levels = levelsMemory;
+            direction = resolveExplicitLevels();
+        } else {
+            /* set BN for all explicit codes, check that all levels are 0 or paraLevel..MAX_EXPLICIT_LEVEL */
+            levels = embeddingLevels;
+            direction = checkExplicitLevels();
+        }
+
+        /* allocate isolate memory */
+        if (isolateCount > 0) {
+            if (isolates == null || isolates.length < isolateCount)
+                isolates = new Isolate[isolateCount + 3];   /* keep some reserve */
+        }
+        isolateCount = -1;              /* current isolates stack entry == none */
+
+        /*
+         * The steps after (X9) in the Bidi algorithm are performed only if
+         * the paragraph text has mixed directionality!
+         */
+        switch (direction) {
+        case LTR:
+            /* all levels are implicitly at paraLevel (important for getLevels()) */
+            trailingWSStart = 0;
+            break;
+        case RTL:
+            /* all levels are implicitly at paraLevel (important for getLevels()) */
+            trailingWSStart = 0;
+            break;
+        default:
+            /*
+             *  Choose the right implicit state table
+             */
+            switch(reorderingMode) {
+            case REORDER_DEFAULT:
+                this.impTabPair = impTab_DEFAULT;
+                break;
+            case REORDER_NUMBERS_SPECIAL:
+                this.impTabPair = impTab_NUMBERS_SPECIAL;
+                break;
+            case REORDER_GROUP_NUMBERS_WITH_R:
+                this.impTabPair = impTab_GROUP_NUMBERS_WITH_R;
+                break;
+            case REORDER_RUNS_ONLY:
+                /* we should never get here */
+                throw new InternalError("Internal ICU error in setPara");
+                /* break; */
+            case REORDER_INVERSE_NUMBERS_AS_L:
+                this.impTabPair = impTab_INVERSE_NUMBERS_AS_L;
+                break;
+            case REORDER_INVERSE_LIKE_DIRECT:
+                if ((reorderingOptions & OPTION_INSERT_MARKS) != 0) {
+                    this.impTabPair = impTab_INVERSE_LIKE_DIRECT_WITH_MARKS;
+                } else {
+                    this.impTabPair = impTab_INVERSE_LIKE_DIRECT;
+                }
+                break;
+            case REORDER_INVERSE_FOR_NUMBERS_SPECIAL:
+                if ((reorderingOptions & OPTION_INSERT_MARKS) != 0) {
+                    this.impTabPair = impTab_INVERSE_FOR_NUMBERS_SPECIAL_WITH_MARKS;
+                } else {
+                    this.impTabPair = impTab_INVERSE_FOR_NUMBERS_SPECIAL;
+                }
+                break;
+            }
+            /*
+             * If there are no external levels specified and there
+             * are no significant explicit level codes in the text,
+             * then we can treat the entire paragraph as one run.
+             * Otherwise, we need to perform the following rules on runs of
+             * the text with the same embedding levels. (X10)
+             * "Significant" explicit level codes are ones that actually
+             * affect non-BN characters.
+             * Examples for "insignificant" ones are empty embeddings
+             * LRE-PDF, LRE-RLE-PDF-PDF, etc.
+             */
+            if (embeddingLevels == null && paraCount <= 1 &&
+                (flags & DirPropFlagMultiRuns) == 0) {
+                resolveImplicitLevels(0, length,
+                        GetLRFromLevel(GetParaLevelAt(0)),
+                        GetLRFromLevel(GetParaLevelAt(length - 1)));
+            } else {
+                /* sor, eor: start and end types of same-level-run */
+                int start, limit = 0;
+                byte level, nextLevel;
+                short sor, eor;
+
+                /* determine the first sor and set eor to it because of the loop body (sor=eor there) */
+                level = GetParaLevelAt(0);
+                nextLevel = levels[0];
+                if (level < nextLevel) {
+                    eor = GetLRFromLevel(nextLevel);
+                } else {
+                    eor = GetLRFromLevel(level);
+                }
+
+                do {
+                    /* determine start and limit of the run (end points just behind the run) */
+
+                    /* the values for this run's start are the same as for the previous run's end */
+                    start = limit;
+                    level = nextLevel;
+                    if ((start > 0) && (dirProps[start - 1] == B)) {
+                        /* except if this is a new paragraph, then set sor = para level */
+                        sor = GetLRFromLevel(GetParaLevelAt(start));
+                    } else {
+                        sor = eor;
+                    }
+
+                    /* search for the limit of this run */
+                    while ((++limit < length) &&
+                           ((levels[limit] == level) ||
+                            ((DirPropFlag(dirProps[limit]) & MASK_BN_EXPLICIT) != 0))) {}
+
+                    /* get the correct level of the next run */
+                    if (limit < length) {
+                        nextLevel = levels[limit];
+                    } else {
+                        nextLevel = GetParaLevelAt(length - 1);
+                    }
+
+                    /* determine eor from max(level, nextLevel); sor is last run's eor */
+                    if (NoOverride(level) < NoOverride(nextLevel)) {
+                        eor = GetLRFromLevel(nextLevel);
+                    } else {
+                        eor = GetLRFromLevel(level);
+                    }
+
+                    /* if the run consists of overridden directional types, then there
+                       are no implicit types to be resolved */
+                    if ((level & LEVEL_OVERRIDE) == 0) {
+                        resolveImplicitLevels(start, limit, sor, eor);
+                    } else {
+                        /* remove the LEVEL_OVERRIDE flags */
+                        do {
+                            levels[start++] &= ~LEVEL_OVERRIDE;
+                        } while (start < limit);
+                    }
+                } while (limit  < length);
+            }
+
+            /* reset the embedding levels for some non-graphic characters (L1), (X9) */
+            adjustWSLevels();
+
+            break;
+        }
+
+        /* add RLM for inverse Bidi with contextual orientation resolving
+         * to RTL which would not round-trip otherwise
+         */
+        if ((defaultParaLevel > 0) &&
+            ((reorderingOptions & OPTION_INSERT_MARKS) != 0) &&
+            ((reorderingMode == REORDER_INVERSE_LIKE_DIRECT) ||
+             (reorderingMode == REORDER_INVERSE_FOR_NUMBERS_SPECIAL))) {
+            int start, last;
+            byte level;
+            byte dirProp;
+            for (int i = 0; i < paraCount; i++) {
+                last = paras_limit[i] - 1;
+                level = paras_level[i];
+                if (level == 0)
+                    continue;           /* LTR paragraph */
+                start = i == 0 ? 0 : paras_limit[i - 1];
+                for (int j = last; j >= start; j--) {
+                    dirProp = dirProps[j];
+                    if (dirProp == L) {
+                        if (j < last) {
+                            while (dirProps[last] == B) {
+                                last--;
+                            }
+                        }
+                        addPoint(last, RLM_BEFORE);
+                        break;
+                    }
+                    if ((DirPropFlag(dirProp) & MASK_R_AL) != 0) {
+                        break;
+                    }
+                }
+            }
+        }
+
+        if ((reorderingOptions & OPTION_REMOVE_CONTROLS) != 0) {
+            resultLength -= controlCount;
+        } else {
+            resultLength += insertPoints.size;
+        }
+        setParaSuccess();
+    }
+
+    /**
+     * Perform the Unicode Bidi algorithm on a given paragraph, as defined in the
+     * Unicode Standard Annex #9,
+     * version 13,
+     * also described in The Unicode Standard, Version 4.0 .

+     *
+     * This method takes a paragraph of text and computes the
+     * left-right-directionality of each character. The text should not
+     * contain any Unicode block separators.

+     *
+     * The RUN_DIRECTION attribute in the text, if present, determines the base
+     * direction (left-to-right or right-to-left). If not present, the base
+     * direction is computed using the Unicode Bidirectional Algorithm,
+     * defaulting to left-to-right if there are no strong directional characters
+     * in the text. This attribute, if present, must be applied to all the text
+     * in the paragraph.

+     *
+     * The BIDI_EMBEDDING attribute in the text, if present, represents
+     * embedding level information. Negative values from -1 to -62 indicate
+     * overrides at the absolute value of the level. Positive values from 1 to
+     * 62 indicate embeddings. Where values are zero or not defined, the base
+     * embedding level as determined by the base direction is assumed.

+     *
+     * The NUMERIC_SHAPING attribute in the text, if present, converts European
+     * digits to other decimal digits before running the bidi algorithm. This
+     * attribute, if present, must be applied to all the text in the paragraph.
+     *
+     * If the entire text is all of the same directionality, then
+     * the method may not perform all the steps described by the algorithm,
+     * i.e., some levels may not be the same as if all steps were performed.
+     * This is not relevant for unidirectional text.

+     * For example, in pure LTR text with numbers the numbers would get
+     * a resolved level of 2 higher than the surrounding text according to
+     * the algorithm. This implementation may set all resolved levels to
+     * the same value in such a case.

+     *
+     * @param paragraph a paragraph of text with optional character and
+     *        paragraph attribute information
+     * @stable ICU 3.8
+     */
+    public void setPara(AttributedCharacterIterator paragraph)
+    {
+        byte paraLvl;
+        char ch = paragraph.first();
+        Boolean runDirection =
+          (Boolean) paragraph.getAttribute(TextAttributeConstants.RUN_DIRECTION);
+        Object shaper = paragraph.getAttribute(TextAttributeConstants.NUMERIC_SHAPING);
+
+        if (runDirection == null) {
+            paraLvl = LEVEL_DEFAULT_LTR;
+        } else {
+            paraLvl = (runDirection.equals(TextAttributeConstants.RUN_DIRECTION_LTR)) ?
+                        LTR : RTL;
+        }
+
+        byte[] lvls = null;
+        int len = paragraph.getEndIndex() - paragraph.getBeginIndex();
+        byte[] embeddingLevels = new byte[len];
+        char[] txt = new char[len];
+        int i = 0;
+        while (ch != AttributedCharacterIterator.DONE) {
+            txt[i] = ch;
+            Integer embedding =
+                (Integer) paragraph.getAttribute(TextAttributeConstants.BIDI_EMBEDDING);
+            if (embedding != null) {
+                byte level = embedding.byteValue();
+                if (level == 0) {
+                    /* no-op */
+                } else if (level < 0) {
+                    lvls = embeddingLevels;
+                    embeddingLevels[i] = (byte)((0 - level) | LEVEL_OVERRIDE);
+                } else {
+                    lvls = embeddingLevels;
+                    embeddingLevels[i] = level;
+                }
+            }
+            ch = paragraph.next();
+            ++i;
+        }
+
+        if (shaper != null) {
+            NumericShapings.shape(shaper, txt, 0, len);
+        }
+        setPara(txt, paraLvl, lvls);
+    }
+
+    /**
+     * Specify whether block separators must be allocated level zero,
+     * so that successive paragraphs will progress from left to right.
+     * This method must be called before setPara().
+     * Paragraph separators (B) may appear in the text.  Setting them to level zero
+     * means that all paragraph separators (including one possibly appearing
+     * in the last text position) are kept in the reordered text after the text
+     * that they follow in the source text.
+     * When this feature is not enabled, a paragraph separator at the last
+     * position of the text before reordering will go to the first position
+     * of the reordered text when the paragraph level is odd.
+     *
+     * @param ordarParaLTR specifies whether paragraph separators (B) must
+     * receive level 0, so that successive paragraphs progress from left to right.
+     *
+     * @see #setPara
+     * @stable ICU 3.8
+     */
+    public void orderParagraphsLTR(boolean ordarParaLTR) {
+        orderParagraphsLTR = ordarParaLTR;
+    }
+
+    /**
+     * Get the directionality of the text.
+     *
+     * @return a value of LTR, RTL or MIXED
+     *         that indicates if the entire text
+     *         represented by this object is unidirectional,
+     *         and which direction, or if it is mixed-directional.
+     *
+     * @throws IllegalStateException if this call is not preceded by a successful
+     *         call to setPara or setLine
+     *
+     * @see #LTR
+     * @see #RTL
+     * @see #MIXED
+     * @stable ICU 3.8
+     */
+    public byte getDirection()
+    {
+        verifyValidParaOrLine();
+        return direction;
+    }
+
+    /**
+     * Get the length of the text.
+     *
+     * @return The length of the text that the Bidi object was
+     *         created for.
+     *
+     * @throws IllegalStateException if this call is not preceded by a successful
+     *         call to setPara or setLine
+     * @stable ICU 3.8
+     */
+    public int getLength()
+    {
+        verifyValidParaOrLine();
+        return originalLength;
+    }
+
+    /* paragraphs API methods ------------------------------------------------- */
+
+    /**
+     * Get the paragraph level of the text.
+     *
+     * @return The paragraph level. If there are multiple paragraphs, their
+     *         level may vary if the required paraLevel is LEVEL_DEFAULT_LTR or
+     *         LEVEL_DEFAULT_RTL.  In that case, the level of the first paragraph
+     *         is returned.
+     *
+     * @throws IllegalStateException if this call is not preceded by a successful
+     *         call to setPara or setLine
+     *
+     * @see #LEVEL_DEFAULT_LTR
+     * @see #LEVEL_DEFAULT_RTL
+     * @see #getParagraph
+     * @see #getParagraphByIndex
+     * @stable ICU 3.8
+     */
+    public byte getParaLevel()
+    {
+        verifyValidParaOrLine();
+        return paraLevel;
+    }
+
+    /**
+     * Retrieves the Bidi class for a given code point.
+     * 
If a BidiClassifier is defined and returns a value
+     * other than CLASS_DEFAULT, that value is used; otherwise
+     * the default class determination mechanism is invoked.

+     *
+     * @param c The code point to get a Bidi class for.
+     *
+     * @return The Bidi class for the character c that is in effect
+     *         for this Bidi instance.
+     *
+     * @stable ICU 3.8
+     */
+    public int getCustomizedClass(int c) {
+        int dir;
+
+        dir = bdp.getClass(c);
+        if (dir >= CHAR_DIRECTION_COUNT)
+            dir = ON;
+        return dir;
+    }
+
+    /**
+     * setLine() returns a Bidi object to
+     * contain the reordering information, especially the resolved levels,
+     * for all the characters in a line of text. This line of text is
+     * specified by referring to a Bidi object representing
+     * this information for a piece of text containing one or more paragraphs,
+     * and by specifying a range of indexes in this text.

+     * In the new line object, the indexes will range from 0 to limit-start-1.

+     *
+     * This is used after calling setPara()
+     * for a piece of text, and after line-breaking on that text.
+     * It is not necessary if each paragraph is treated as a single line.

+     *
+     * After line-breaking, rules (L1) and (L2) for the treatment of
+     * trailing WS and for reordering are performed on
+     * a Bidi object that represents a line.

+     *
+     * Important: the line Bidi object may
+     * reference data within the global text Bidi object.
+     * You should not alter the content of the global text object until
+     * you are finished using the line object.
+     *
+     * @param start is the line's first index into the text.
+     *
+     * @param limit is just behind the line's last index into the text
+     *        (its last index +1).
+     *
+     * @return a Bidi object that will now represent a line of the text.
+     *
+     * @throws IllegalStateException if this call is not preceded by a successful
+     *         call to setPara
+     * @throws IllegalArgumentException if start and limit are not in the range
+     *         0<=start<limit<=getProcessedLength(),
+     *         or if the specified line crosses a paragraph boundary
+     *
+     * @see #setPara
+     * @see #getProcessedLength
+     * @stable ICU 3.8
+     */
+    public Bidi setLine(Bidi bidi, BidiBase bidiBase, Bidi newBidi, BidiBase newBidiBase, int start, int limit)
+    {
+        verifyValidPara();
+        verifyRange(start, 0, limit);
+        verifyRange(limit, 0, length+1);
+
+        return BidiLine.setLine(this, newBidi, newBidiBase, start, limit);
+    }
+
+    /**
+     * Get the level for one character.
+     *
+     * @param charIndex the index of a character.
+     *
+     * @return The level for the character at charIndex.
+     *
+     * @throws IllegalStateException if this call is not preceded by a successful
+     *         call to setPara or setLine
+     * @throws IllegalArgumentException if charIndex is not in the range
+     *         0<=charIndex<getProcessedLength()
+     *
+     * @see #getProcessedLength
+     * @stable ICU 3.8
+     */
+    public byte getLevelAt(int charIndex)
+    {
+        // for backward compatibility
+        if (charIndex < 0 || charIndex >= length) {
+            return (byte)getBaseLevel();
+        }
+
+        verifyValidParaOrLine();
+        verifyRange(charIndex, 0, length);
+        return BidiLine.getLevelAt(this, charIndex);
+    }
+
+    /**
+     * Get an array of levels for each character.

+     *
+     * Note that this method may allocate memory under some
+     * circumstances, unlike getLevelAt().
+     *
+     * @return The levels array for the text,
+     *         or null if an error occurs.
+     *
+     * @throws IllegalStateException if this call is not preceded by a successful
+     *         call to setPara or setLine
+     * @stable ICU 3.8
+     */
+    byte[] getLevels()
+    {
+        verifyValidParaOrLine();
+        if (length <= 0) {
+            return new byte[0];
+        }
+        return BidiLine.getLevels(this);
+    }
+
+    /**
+     * Get the number of runs.
+     * This method may invoke the actual reordering on the
+     * Bidi object, after setPara()
+     * may have resolved only the levels of the text. Therefore,
+     * countRuns() may have to allocate memory,
+     * and may throw an exception if it fails to do so.
+     *
+     * @return The number of runs.
+     *
+     * @throws IllegalStateException if this call is not preceded by a successful
+     *         call to setPara or setLine
+     * @stable ICU 3.8
+     */
+    public int countRuns()
+    {
+        verifyValidParaOrLine();
+        BidiLine.getRuns(this);
+        return runCount;
+    }
+
+    /**
+     *
+     * Get a BidiRun object according to its index. BidiRun methods
+     * may be used to retrieve the run's logical start, length and level,
+     * which can be even for an LTR run or odd for an RTL run.
+     * In an RTL run, the character at the logical start is
+     * visually on the right of the displayed run.
+     * The length is the number of characters in the run.

+     * countRuns() is normally called
+     * before the runs are retrieved.
+     *
+     * 

+     *  Example:
+     * 
+     *  Bidi bidi = new Bidi();
+     *  String text = "abc 123 DEFG xyz";
+     *  bidi.setPara(text, Bidi.RTL, null);
+     *  int i, count=bidi.countRuns(), logicalStart, visualIndex=0, length;
+     *  BidiRun run;
+     *  for (i = 0; i < count; ++i) {
+     *      run = bidi.getVisualRun(i);
+     *      logicalStart = run.getStart();
+     *      length = run.getLength();
+     *      if (Bidi.LTR == run.getEmbeddingLevel()) {
+     *          do { // LTR
+     *              show_char(text.charAt(logicalStart++), visualIndex++);
+     *          } while (--length > 0);
+     *      } else {
+     *          logicalStart += length;  // logicalLimit
+     *          do { // RTL
+     *              show_char(text.charAt(--logicalStart), visualIndex++);
+     *          } while (--length > 0);
+     *      }
+     *  }
+     * 

+     * 

+     * Note that in right-to-left runs, code like this places
+     * second surrogates before first ones (which is generally a bad idea)
+     * and combining characters before base characters.
+     * 

+     * Use of {@link #writeReordered}, optionally with the
+     * {@link #KEEP_BASE_COMBINING} option, can be considered in
+     * order to avoid these issues.
+     *
+     * @param runIndex is the number of the run in visual order, in the
+     *        range [0..countRuns()-1].
+     *
+     * @return a BidiRun object containing the details of the run. The
+     *         directionality of the run is
+     *         LTR==0 or RTL==1,
+     *         never MIXED.
+     *
+     * @throws IllegalStateException if this call is not preceded by a successful
+     *         call to setPara or setLine
+     * @throws IllegalArgumentException if runIndex is not in
+     *         the range 0<=runIndex<countRuns()
+     *
+     * @see #countRuns()
+     * @see com.ibm.icu.text.BidiRun
+     * @see com.ibm.icu.text.BidiRun#getStart()
+     * @see com.ibm.icu.text.BidiRun#getLength()
+     * @see com.ibm.icu.text.BidiRun#getEmbeddingLevel()
+     * @stable ICU 3.8
+     */
+    BidiRun getVisualRun(int runIndex)
+    {
+        verifyValidParaOrLine();
+        BidiLine.getRuns(this);
+        verifyRange(runIndex, 0, runCount);
+        return BidiLine.getVisualRun(this, runIndex);
+    }
+
+    /**
+     * Get a visual-to-logical index map (array) for the characters in the
+     * Bidi (paragraph or line) object.
+     * 

+     * Some values in the map may be MAP_NOWHERE if the
+     * corresponding text characters are Bidi marks inserted in the visual
+     * output by the option OPTION_INSERT_MARKS.
+     * 

+     * When the visual output is altered by using options of
+     * writeReordered() such as INSERT_LRM_FOR_NUMERIC,
+     * KEEP_BASE_COMBINING, OUTPUT_REVERSE,
+     * REMOVE_BIDI_CONTROLS, the logical positions returned may not
+     * be correct. It is advised to use, when possible, reordering options
+     * such as {@link #OPTION_INSERT_MARKS} and {@link #OPTION_REMOVE_CONTROLS}.
+     *
+     * @return an array of getResultLength()
+     *        indexes which will reflect the reordering of the characters.


+     *        The index map will result in
+     *        indexMap[visualIndex]==logicalIndex, where
+     *        indexMap represents the returned array.
+     *
+     * @throws IllegalStateException if this call is not preceded by a successful
+     *         call to setPara or setLine
+     *
+     * @see #getLogicalMap
+     * @see #getLogicalIndex
+     * @see #getResultLength
+     * @see #MAP_NOWHERE
+     * @see #OPTION_INSERT_MARKS
+     * @see #writeReordered
+     * @stable ICU 3.8
+     */
+    private int[] getVisualMap()
+    {
+        /* countRuns() checks successful call to setPara/setLine */
+        countRuns();
+        if (resultLength <= 0) {
+            return new int[0];
+        }
+        return BidiLine.getVisualMap(this);
+    }
+
+    /**
+     * This is a convenience method that does not use a Bidi object.
+     * It is intended to be used for when an application has determined the levels
+     * of objects (character sequences) and just needs to have them reordered (L2).
+     * This is equivalent to using getVisualMap() on a
+     * Bidi object.
+     *
+     * @param levels is an array of levels that have been determined by
+     *        the application.
+     *
+     * @return an array of levels.length
+     *        indexes which will reflect the reordering of the characters.

+     *        The index map will result in
+     *        indexMap[visualIndex]==logicalIndex, where
+     *        indexMap represents the returned array.
+     *
+     * @stable ICU 3.8
+     */
+    private static int[] reorderVisual(byte[] levels)
+    {
+        return BidiLine.reorderVisual(levels);
+    }
+
+    /**
+     * Constant indicating that the base direction depends on the first strong
+     * directional character in the text according to the Unicode Bidirectional
+     * Algorithm. If no strong directional character is present, the base
+     * direction is right-to-left.
+     * @stable ICU 3.8
+     */
+    public static final int DIRECTION_DEFAULT_RIGHT_TO_LEFT = LEVEL_DEFAULT_RTL;
+
+    /**
+     * Create Bidi from the given text, embedding, and direction information.
+     * The embeddings array may be null. If present, the values represent
+     * embedding level information. Negative values from -1 to -61 indicate
+     * overrides at the absolute value of the level. Positive values from 1 to
+     * 61 indicate embeddings. Where values are zero, the base embedding level
+     * as determined by the base direction is assumed.

+     *
+     * Note: this constructor calls setPara() internally.
+     *
+     * @param text an array containing the paragraph of text to process.
+     * @param textStart the index into the text array of the start of the
+     *        paragraph.
+     * @param embeddings an array containing embedding values for each character
+     *        in the paragraph. This can be null, in which case it is assumed
+     *        that there is no external embedding information.
+     * @param embStart the index into the embedding array of the start of the
+     *        paragraph.
+     * @param paragraphLength the length of the paragraph in the text and
+     *        embeddings arrays.
+     * @param flags a collection of flags that control the algorithm. The
+     *        algorithm understands the flags DIRECTION_LEFT_TO_RIGHT,
+     *        DIRECTION_RIGHT_TO_LEFT, DIRECTION_DEFAULT_LEFT_TO_RIGHT, and
+     *        DIRECTION_DEFAULT_RIGHT_TO_LEFT. Other values are reserved.
+     *
+     * @throws IllegalArgumentException if the values in embeddings are
+     *         not within the allowed range
+     *
+     * @see #DIRECTION_LEFT_TO_RIGHT
+     * @see #DIRECTION_RIGHT_TO_LEFT
+     * @see #DIRECTION_DEFAULT_LEFT_TO_RIGHT
+     * @see #DIRECTION_DEFAULT_RIGHT_TO_LEFT
+     * @stable ICU 3.8
+     */
+    public BidiBase(char[] text,
+            int textStart,
+            byte[] embeddings,
+            int embStart,
+            int paragraphLength,
+            int flags)
+    {
+        this(0, 0);
+        byte paraLvl;
+        switch (flags) {
+        case Bidi.DIRECTION_LEFT_TO_RIGHT:
+        default:
+            paraLvl = LTR;
+            break;
+        case Bidi.DIRECTION_RIGHT_TO_LEFT:
+            paraLvl = RTL;
+            break;
+        case Bidi.DIRECTION_DEFAULT_LEFT_TO_RIGHT:
+            paraLvl = LEVEL_DEFAULT_LTR;
+            break;
+        case Bidi.DIRECTION_DEFAULT_RIGHT_TO_LEFT:
+            paraLvl = LEVEL_DEFAULT_RTL;
+            break;
+        }
+        byte[] paraEmbeddings;
+        if (embeddings == null) {
+            paraEmbeddings = null;
+        } else {
+            paraEmbeddings = new byte[paragraphLength];
+            byte lev;
+            for (int i = 0; i < paragraphLength; i++) {
+                lev = embeddings[i + embStart];
+                if (lev < 0) {
+                    lev = (byte)((- lev) | LEVEL_OVERRIDE);
+                } else if (lev == 0) {
+                    lev = paraLvl;
+                    if (paraLvl > MAX_EXPLICIT_LEVEL) {
+                        lev &= 1;
+                    }
+                }
+                paraEmbeddings[i] = lev;
+            }
+        }
+
+        char[] paraText = new char[paragraphLength];
+        System.arraycopy(text, textStart, paraText, 0, paragraphLength);
+        setPara(paraText, paraLvl, paraEmbeddings);
+    }
+
+    /**
+     * Return true if the line is not left-to-right or right-to-left. This means
+     * it either has mixed runs of left-to-right and right-to-left text, or the
+     * base direction differs from the direction of the only run of text.
+     *
+     * @return true if the line is not left-to-right or right-to-left.
+     *
+     * @throws IllegalStateException if this call is not preceded by a successful
+     *         call to setPara
+     * @stable ICU 3.8
+     */
+    public boolean isMixed()
+    {
+        return (!isLeftToRight() && !isRightToLeft());
+    }
+
+    /**
+     * Return true if the line is all left-to-right text and the base direction
+     * is left-to-right.
+     *
+     * @return true if the line is all left-to-right text and the base direction
+     *         is left-to-right.
+     *
+     * @throws IllegalStateException if this call is not preceded by a successful
+     *         call to setPara
+     * @stable ICU 3.8
+     */
+    public boolean isLeftToRight()
+    {
+        return (getDirection() == LTR && (paraLevel & 1) == 0);
+    }
+
+    /**
+     * Return true if the line is all right-to-left text, and the base direction
+     * is right-to-left
+     *
+     * @return true if the line is all right-to-left text, and the base
+     *         direction is right-to-left
+     *
+     * @throws IllegalStateException if this call is not preceded by a successful
+     *         call to setPara
+     * @stable ICU 3.8
+     */
+    public boolean isRightToLeft()
+    {
+        return (getDirection() == RTL && (paraLevel & 1) == 1);
+    }
+
+    /**
+     * Return true if the base direction is left-to-right
+     *
+     * @return true if the base direction is left-to-right
+     *
+     * @throws IllegalStateException if this call is not preceded by a successful
+     *         call to setPara or setLine
+     *
+     * @stable ICU 3.8
+     */
+    public boolean baseIsLeftToRight()
+    {
+        return (getParaLevel() == LTR);
+    }
+
+    /**
+     * Return the base level (0 if left-to-right, 1 if right-to-left).
+     *
+     * @return the base level
+     *
+     * @throws IllegalStateException if this call is not preceded by a successful
+     *         call to setPara or setLine
+     *
+     * @stable ICU 3.8
+     */
+    public int getBaseLevel()
+    {
+        return getParaLevel();
+    }
+
+    /**
+     * Compute the logical to visual run mapping
+     */
+     void getLogicalToVisualRunsMap()
+     {
+        if (isGoodLogicalToVisualRunsMap) {
+            return;
+        }
+        int count = countRuns();
+        if ((logicalToVisualRunsMap == null) ||
+            (logicalToVisualRunsMap.length < count)) {
+            logicalToVisualRunsMap = new int[count];
+        }
+        int i;
+        long[] keys = new long[count];
+        for (i = 0; i < count; i++) {
+            keys[i] = ((long)(runs[i].start)<<32) + i;
+        }
+        Arrays.sort(keys);
+        for (i = 0; i < count; i++) {
+            logicalToVisualRunsMap[i] = (int)(keys[i] & 0x00000000FFFFFFFF);
+        }
+        isGoodLogicalToVisualRunsMap = true;
+     }
+
+    /**
+     * Return the level of the nth logical run in this line.
+     *
+     * @param run the index of the run, between 0 and countRuns()-1
+     *
+     * @return the level of the run
+     *
+     * @throws IllegalStateException if this call is not preceded by a successful
+     *         call to setPara or setLine
+     * @throws IllegalArgumentException if run is not in
+     *         the range 0<=run<countRuns()
+     * @stable ICU 3.8
+     */
+    public int getRunLevel(int run)
+    {
+        verifyValidParaOrLine();
+        BidiLine.getRuns(this);
+
+        // for backward compatibility
+        if (run < 0 || run >= runCount) {
+            return getParaLevel();
+        }
+
+        getLogicalToVisualRunsMap();
+        return runs[logicalToVisualRunsMap[run]].level;
+    }
+
+    /**
+     * Return the index of the character at the start of the nth logical run in
+     * this line, as an offset from the start of the line.
+     *
+     * @param run the index of the run, between 0 and countRuns()
+     *
+     * @return the start of the run
+     *
+     * @throws IllegalStateException if this call is not preceded by a successful
+     *         call to setPara or setLine
+     * @throws IllegalArgumentException if run is not in
+     *         the range 0<=run<countRuns()
+     * @stable ICU 3.8
+     */
+    public int getRunStart(int run)
+    {
+        verifyValidParaOrLine();
+        BidiLine.getRuns(this);
+
+        // for backward compatibility
+        if (runCount == 1) {
+            return 0;
+        } else if (run == runCount) {
+            return length;
+        }
+
+        getLogicalToVisualRunsMap();
+        return runs[logicalToVisualRunsMap[run]].start;
+    }
+
+    /**
+     * Return the index of the character past the end of the nth logical run in
+     * this line, as an offset from the start of the line. For example, this
+     * will return the length of the line for the last run on the line.
+     *
+     * @param run the index of the run, between 0 and countRuns()
+     *
+     * @return the limit of the run
+     *
+     * @throws IllegalStateException if this call is not preceded by a successful
+     *         call to setPara or setLine
+     * @throws IllegalArgumentException if run is not in
+     *         the range 0<=run<countRuns()
+     * @stable ICU 3.8
+     */
+    public int getRunLimit(int run)
+    {
+        verifyValidParaOrLine();
+        BidiLine.getRuns(this);
+
+        // for backward compatibility
+        if (runCount == 1) {
+            return length;
+        }
+
+        getLogicalToVisualRunsMap();
+        int idx = logicalToVisualRunsMap[run];
+        int len = idx == 0 ? runs[idx].limit :
+                                runs[idx].limit - runs[idx-1].limit;
+        return runs[idx].start + len;
+    }
+
+    /**
+     * Return true if the specified text requires bidi analysis. If this returns
+     * false, the text will display left-to-right. Clients can then avoid
+     * constructing a Bidi object. Text in the Arabic Presentation Forms area of
+     * Unicode is presumed to already be shaped and ordered for display, and so
+     * will not cause this method to return true.
+     *
+     * @param text the text containing the characters to test
+     * @param start the start of the range of characters to test
+     * @param limit the limit of the range of characters to test
+     *
+     * @return true if the range of characters requires bidi analysis
+     *
+     * @stable ICU 3.8
+     */
+    public static boolean requiresBidi(char[] text,
+            int start,
+            int limit)
+    {
+        final int RTLMask = (1 << R |
+                1 << AL |
+                1 << RLE |
+                1 << RLO |
+                1 << AN);
+
+        if (0 > start || start > limit || limit > text.length) {
+            throw new IllegalArgumentException("Value start " + start +
+                      " is out of range 0 to " + limit);
+        }
+
+        for (int i = start; i < limit; ++i) {
+            if (Character.isHighSurrogate(text[i]) && i < (limit-1) &&
+                Character.isLowSurrogate(text[i+1])) {
+                if (((1 << UCharacter.getDirection(Character.codePointAt(text, i))) & RTLMask) != 0) {
+                    return true;
+                }
+            } else if (((1 << UCharacter.getDirection(text[i])) & RTLMask) != 0) {
+                return true;
+            }
+        }
+
+        return false;
+    }
+
+    /**
+     * Reorder the objects in the array into visual order based on their levels.
+     * This is a utility method to use when you have a collection of objects
+     * representing runs of text in logical order, each run containing text at a
+     * single level. The elements at index from
+     * objectStart up to objectStart + count in the
+     * objects array will be reordered into visual order assuming
+     * each run of text has the level indicated by the corresponding element in
+     * the levels array (at index - objectStart + levelStart).
+     *
+     * @param levels an array representing the bidi level of each object
+     * @param levelStart the start position in the levels array
+     * @param objects the array of objects to be reordered into visual order
+     * @param objectStart the start position in the objects array
+     * @param count the number of objects to reorder
+     * @stable ICU 3.8
+     */
+    public static void reorderVisually(byte[] levels,
+            int levelStart,
+            Object[] objects,
+            int objectStart,
+            int count)
+    {
+        // for backward compatibility
+        if (0 > levelStart || levels.length <= levelStart) {
+          throw new IllegalArgumentException("Value levelStart " +
+                      levelStart + " is out of range 0 to " +
+                      (levels.length-1));
+        }
+        if (0 > objectStart || objects.length <= objectStart) {
+            throw new IllegalArgumentException("Value objectStart " +
+                      levelStart + " is out of range 0 to " +
+                      (objects.length-1));
+        }
+        if (0 > count || objects.length < (objectStart+count)) {
+            throw new IllegalArgumentException("Value count " +
+                      levelStart + " is out of range 0 to " +
+                      (objects.length - objectStart));
+        }
+
+        byte[] reorderLevels = new byte[count];
+        System.arraycopy(levels, levelStart, reorderLevels, 0, count);
+        int[] indexMap = reorderVisual(reorderLevels);
+        Object[] temp = new Object[count];
+        System.arraycopy(objects, objectStart, temp, 0, count);
+        for (int i = 0; i < count; ++i) {
+            objects[objectStart + i] = temp[indexMap[i]];
+        }
+    }
+
+    /**
+     * Take a Bidi object containing the reordering
+     * information for a piece of text (one or more paragraphs) set by
+     * setPara() or for a line of text set by setLine()
+     * and return a string containing the reordered text.
+     *
+     * 
The text may have been aliased (only a reference was stored
+     * without copying the contents), thus it must not have been modified
+     * since the setPara() call.

+     *
+     * This method preserves the integrity of characters with multiple
+     * code units and (optionally) combining characters.
+     * Characters in RTL runs can be replaced by mirror-image characters
+     * in the returned string. Note that "real" mirroring has to be done in a
+     * rendering engine by glyph selection and that for many "mirrored"
+     * characters there are no Unicode characters as mirror-image equivalents.
+     * There are also options to insert or remove Bidi control
+     * characters; see the descriptions of the return value and the
+     * options parameter, and of the option bit flags.
+     *
+     * @param options A bit set of options for the reordering that control
+     *                how the reordered text is written.
+     *                The options include mirroring the characters on a code
+     *                point basis and inserting LRM characters, which is used
+     *                especially for transforming visually stored text
+     *                to logically stored text (although this is still an
+     *                imperfect implementation of an "inverse Bidi" algorithm
+     *                because it uses the "forward Bidi" algorithm at its core).
+     *                The available options are:
+     *                DO_MIRRORING,
+     *                INSERT_LRM_FOR_NUMERIC,
+     *                KEEP_BASE_COMBINING,
+     *                OUTPUT_REVERSE,
+     *                REMOVE_BIDI_CONTROLS,
+     *                STREAMING
+     *
+     * @return The reordered text.
+     *         If the INSERT_LRM_FOR_NUMERIC option is set, then
+     *         the length of the returned string could be as large as
+     *         getLength()+2*countRuns().

+     *         If the REMOVE_BIDI_CONTROLS option is set, then the
+     *         length of the returned string may be less than
+     *         getLength().

+     *         If none of these options is set, then the length of the returned
+     *         string will be exactly getProcessedLength().
+     *
+     * @throws IllegalStateException if this call is not preceded by a successful
+     *         call to setPara or setLine
+     *
+     * @see #DO_MIRRORING
+     * @see #INSERT_LRM_FOR_NUMERIC
+     * @see #KEEP_BASE_COMBINING
+     * @see #OUTPUT_REVERSE
+     * @see #REMOVE_BIDI_CONTROLS
+     * @see #OPTION_STREAMING
+     * @see #getProcessedLength
+     * @stable ICU 3.8
+     */
+    public String writeReordered(int options)
+    {
+        verifyValidParaOrLine();
+        if (length == 0) {
+            /* nothing to do */
+            return "";
+        }
+        return BidiWriter.writeReordered(this, options);
+    }
+
+    /**
+     * Display the bidi internal state, used in debugging.
+     */
+    public String toString() {
+        StringBuilder buf = new StringBuilder(getClass().getName());
+
+        buf.append("[dir: ");
+        buf.append(direction);
+        buf.append(" baselevel: ");
+        buf.append(paraLevel);
+        buf.append(" length: ");
+        buf.append(length);
+        buf.append(" runs: ");
+        if (levels == null) {
+            buf.append("none");
+        } else {
+            buf.append('[');
+            buf.append(levels[0]);
+            for (int i = 1; i < levels.length; i++) {
+                buf.append(' ');
+                buf.append(levels[i]);
+            }
+            buf.append(']');
+        }
+        buf.append(" text: [0x");
+        buf.append(Integer.toHexString(text[0]));
+        for (int i = 1; i < text.length; i++) {
+            buf.append(" 0x");
+            buf.append(Integer.toHexString(text[i]));
+        }
+        buf.append("]]");
+
+        return buf.toString();
+    }
+
+    /**
+     * A class that provides access to constants defined by
+     * java.awt.font.TextAttribute without creating a static dependency.
+     */
+    private static class TextAttributeConstants {
+        // Make sure to load the AWT's TextAttribute class before using the constants, if any.
+        static {
+            try {
+                Class.forName("java.awt.font.TextAttribute", true, null);
+            } catch (ClassNotFoundException e) {}
+        }
+        static final JavaAWTFontAccess jafa = SharedSecrets.getJavaAWTFontAccess();
+
+        /**
+         * TextAttribute instances (or a fake Attribute type if
+         * java.awt.font.TextAttribute is not present)
+         */
+        static final AttributedCharacterIterator.Attribute RUN_DIRECTION =
+            getTextAttribute("RUN_DIRECTION");
+        static final AttributedCharacterIterator.Attribute NUMERIC_SHAPING =
+            getTextAttribute("NUMERIC_SHAPING");
+        static final AttributedCharacterIterator.Attribute BIDI_EMBEDDING =
+            getTextAttribute("BIDI_EMBEDDING");
+
+        /**
+         * TextAttribute.RUN_DIRECTION_LTR
+         */
+        static final Boolean RUN_DIRECTION_LTR = (jafa == null) ?
+            Boolean.FALSE : (Boolean)jafa.getTextAttributeConstant("RUN_DIRECTION_LTR");
+
+        @SuppressWarnings("serial")
+        private static AttributedCharacterIterator.Attribute
+            getTextAttribute(String name)
+        {
+            if (jafa == null) {
+                // fake attribute
+                return new AttributedCharacterIterator.Attribute(name) { };
+            } else {
+                return (AttributedCharacterIterator.Attribute)jafa.getTextAttributeConstant(name);
+            }
+        }
+    }
+
+    /**
+     * A class that provides access to java.awt.font.NumericShaper without
+     * creating a static dependency.
+     */
+    private static class NumericShapings {
+        // Make sure to load the AWT's NumericShaper class before calling shape, if any.
+        static {
+            try {
+                Class.forName("java.awt.font.NumericShaper", true, null);
+            } catch (ClassNotFoundException e) {}
+        }
+        static final JavaAWTFontAccess jafa = SharedSecrets.getJavaAWTFontAccess();
+
+        /**
+         * Invokes NumericShaping shape(text,start,count) method.
+         */
+        static void shape(Object shaper, char[] text, int start, int count) {
+            if (jafa != null) {
+                jafa.shape(shaper, text, start, count);
+            }
+        }
+    }
+
+}
--- old/src/java.base/share/classes/sun/text/bidi/BidiLine.java	2020-01-10 15:57:54.000000000 -0800
+++ /dev/null	2020-01-10 15:57:54.000000000 -0800
@@ -1,835 +0,0 @@
-/*
- * Copyright (c) 2009, 2015, Oracle and/or its affiliates. All rights reserved.
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.  Oracle designates this
- * particular file as subject to the "Classpath" exception as provided
- * by Oracle in the LICENSE file that accompanied this code.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- */
-
-/*
-*******************************************************************************
-*   Copyright (C) 2001-2014, International Business Machines
-*   Corporation and others.  All Rights Reserved.
-*******************************************************************************
-*/
-/* Written by Simon Montagu, Matitiahu Allouche
- * (ported from C code written by Markus W. Scherer)
- */
-
-package sun.text.bidi;
-
-import java.text.Bidi;
-import java.util.Arrays;
-
-final class BidiLine {
-
-    /*
-     * General remarks about the functions in this file:
-     *
-     * These functions deal with the aspects of potentially mixed-directional
-     * text in a single paragraph or in a line of a single paragraph
-     * which has already been processed according to
-     * the Unicode 3.0 Bidi algorithm as defined in
-     * http://www.unicode.org/unicode/reports/tr9/ , version 13,
-     * also described in The Unicode Standard, Version 4.0.1 .
-     *
-     * This means that there is a Bidi object with a levels
-     * and a dirProps array.
-     * paraLevel and direction are also set.
-     * Only if the length of the text is zero, then levels==dirProps==NULL.
-     *
-     * The overall directionality of the paragraph
-     * or line is used to bypass the reordering steps if possible.
-     * Even purely RTL text does not need reordering there because
-     * the getLogical/VisualIndex() methods can compute the
-     * index on the fly in such a case.
-     *
-     * The implementation of the access to same-level-runs and of the reordering
-     * do attempt to provide better performance and less memory usage compared to
-     * a direct implementation of especially rule (L2) with an array of
-     * one (32-bit) integer per text character.
-     *
-     * Here, the levels array is scanned as soon as necessary, and a vector of
-     * same-level-runs is created. Reordering then is done on this vector.
-     * For each run of text positions that were resolved to the same level,
-     * only 8 bytes are stored: the first text position of the run and the visual
-     * position behind the run after reordering.
-     * One sign bit is used to hold the directionality of the run.
-     * This is inefficient if there are many very short runs. If the average run
-     * length is <2, then this uses more memory.
-     *
-     * In a further attempt to save memory, the levels array is never changed
-     * after all the resolution rules (Xn, Wn, Nn, In).
-     * Many methods have to consider the field trailingWSStart:
-     * if it is less than length, then there is an implicit trailing run
-     * at the paraLevel,
-     * which is not reflected in the levels array.
-     * This allows a line Bidi object to use the same levels array as
-     * its paragraph parent object.
-     *
-     * When a Bidi object is created for a line of a paragraph, then the
-     * paragraph's levels and dirProps arrays are reused by way of setting
-     * a pointer into them, not by copying. This again saves memory and forbids to
-     * change the now shared levels for (L1).
-     */
-
-    /* handle trailing WS (L1) -------------------------------------------------- */
-
-    /*
-     * setTrailingWSStart() sets the start index for a trailing
-     * run of WS in the line. This is necessary because we do not modify
-     * the paragraph's levels array that we just point into.
-     * Using trailingWSStart is another form of performing (L1).
-     *
-     * To make subsequent operations easier, we also include the run
-     * before the WS if it is at the paraLevel - we merge the two here.
-     *
-     * This method is called only from setLine(), so paraLevel is
-     * set correctly for the line even when contextual multiple paragraphs.
-     */
-
-    static void setTrailingWSStart(BidiBase bidiBase)
-    {
-        byte[] dirProps = bidiBase.dirProps;
-        byte[] levels = bidiBase.levels;
-        int start = bidiBase.length;
-        byte paraLevel = bidiBase.paraLevel;
-
-        /* If the line is terminated by a block separator, all preceding WS etc...
-           are already set to paragraph level.
-           Setting trailingWSStart to pBidi->length will avoid changing the
-           level of B chars from 0 to paraLevel in getLevels when
-           orderParagraphsLTR==TRUE
-        */
-        if (dirProps[start - 1] == BidiBase.B) {
-            bidiBase.trailingWSStart = start;   /* currently == bidiBase.length */
-            return;
-        }
-        /* go backwards across all WS, BN, explicit codes */
-        while (start > 0 &&
-                (BidiBase.DirPropFlag(dirProps[start - 1]) & BidiBase.MASK_WS) != 0) {
-            --start;
-        }
-
-        /* if the WS run can be merged with the previous run then do so here */
-        while (start > 0 && levels[start - 1] == paraLevel) {
-            --start;
-        }
-
-        bidiBase.trailingWSStart=start;
-    }
-
-    static Bidi setLine(BidiBase paraBidi,
-                              Bidi newBidi, BidiBase lineBidi,
-                              int start, int limit) {
-        int length;
-
-        /* set the values in lineBidi from its paraBidi parent */
-        /* class members are already initialized to 0 */
-        // lineBidi.paraBidi = null;        /* mark unfinished setLine */
-        // lineBidi.flags = 0;
-        // lineBidi.controlCount = 0;
-
-        length = lineBidi.length = lineBidi.originalLength =
-                lineBidi.resultLength = limit - start;
-
-        lineBidi.text = new char[length];
-        System.arraycopy(paraBidi.text, start, lineBidi.text, 0, length);
-        lineBidi.paraLevel = paraBidi.GetParaLevelAt(start);
-        lineBidi.paraCount = paraBidi.paraCount;
-        lineBidi.runs = new BidiRun[0];
-        lineBidi.reorderingMode = paraBidi.reorderingMode;
-        lineBidi.reorderingOptions = paraBidi.reorderingOptions;
-        if (paraBidi.controlCount > 0) {
-            int j;
-            for (j = start; j < limit; j++) {
-                if (BidiBase.IsBidiControlChar(paraBidi.text[j])) {
-                    lineBidi.controlCount++;
-                }
-            }
-            lineBidi.resultLength -= lineBidi.controlCount;
-        }
-        /* copy proper subset of DirProps */
-        lineBidi.getDirPropsMemory(length);
-        lineBidi.dirProps = lineBidi.dirPropsMemory;
-        System.arraycopy(paraBidi.dirProps, start, lineBidi.dirProps, 0,
-                         length);
-        /* copy proper subset of Levels */
-        lineBidi.getLevelsMemory(length);
-        lineBidi.levels = lineBidi.levelsMemory;
-        System.arraycopy(paraBidi.levels, start, lineBidi.levels, 0,
-                         length);
-        lineBidi.runCount = -1;
-
-        if (paraBidi.direction != BidiBase.MIXED) {
-            /* the parent is already trivial */
-            lineBidi.direction = paraBidi.direction;
-
-            /*
-             * The parent's levels are all either
-             * implicitly or explicitly ==paraLevel;
-             * do the same here.
-             */
-            if (paraBidi.trailingWSStart <= start) {
-                lineBidi.trailingWSStart = 0;
-            } else if (paraBidi.trailingWSStart < limit) {
-                lineBidi.trailingWSStart = paraBidi.trailingWSStart - start;
-            } else {
-                lineBidi.trailingWSStart = length;
-            }
-        } else {
-            byte[] levels = lineBidi.levels;
-            int i, trailingWSStart;
-            byte level;
-
-            setTrailingWSStart(lineBidi);
-            trailingWSStart = lineBidi.trailingWSStart;
-
-            /* recalculate lineBidiBase.direction */
-            if (trailingWSStart == 0) {
-                /* all levels are at paraLevel */
-                lineBidi.direction = (byte)(lineBidi.paraLevel & 1);
-            } else {
-                /* get the level of the first character */
-                level = (byte)(levels[0] & 1);
-
-                /* if there is anything of a different level, then the line
-                   is mixed */
-                if (trailingWSStart < length &&
-                    (lineBidi.paraLevel & 1) != level) {
-                    /* the trailing WS is at paraLevel, which differs from
-                       levels[0] */
-                    lineBidi.direction = BidiBase.MIXED;
-                } else {
-                    /* see if levels[1..trailingWSStart-1] have the same
-                       direction as levels[0] and paraLevel */
-                    for (i = 1; ; i++) {
-                        if (i == trailingWSStart) {
-                            /* the direction values match those in level */
-                            lineBidi.direction = level;
-                            break;
-                        } else if ((levels[i] & 1) != level) {
-                            lineBidi.direction = BidiBase.MIXED;
-                            break;
-                        }
-                    }
-                }
-            }
-
-            switch(lineBidi.direction) {
-                case Bidi.DIRECTION_LEFT_TO_RIGHT:
-                    /* make sure paraLevel is even */
-                    lineBidi.paraLevel = (byte)
-                        ((lineBidi.paraLevel + 1) & ~1);
-
-                    /* all levels are implicitly at paraLevel (important for
-                       getLevels()) */
-                    lineBidi.trailingWSStart = 0;
-                    break;
-                case Bidi.DIRECTION_RIGHT_TO_LEFT:
-                    /* make sure paraLevel is odd */
-                    lineBidi.paraLevel |= 1;
-
-                    /* all levels are implicitly at paraLevel (important for
-                       getLevels()) */
-                    lineBidi.trailingWSStart = 0;
-                    break;
-                default:
-                    break;
-            }
-        }
-
-        lineBidi.paraBidi = paraBidi;     /* mark successful setLine */
-
-        return newBidi;
-    }
-
-    static byte getLevelAt(BidiBase bidiBase, int charIndex)
-    {
-        /* return paraLevel if in the trailing WS run, otherwise the real level */
-        if (bidiBase.direction != BidiBase.MIXED || charIndex >= bidiBase.trailingWSStart) {
-            return bidiBase.GetParaLevelAt(charIndex);
-        } else {
-            return bidiBase.levels[charIndex];
-        }
-    }
-
-    static byte[] getLevels(BidiBase bidiBase)
-    {
-        int start = bidiBase.trailingWSStart;
-        int length = bidiBase.length;
-
-        if (start != length) {
-            /* the current levels array does not reflect the WS run */
-            /*
-             * After the previous if(), we know that the levels array
-             * has an implicit trailing WS run and therefore does not fully
-             * reflect itself all the levels.
-             * This must be a Bidi object for a line, and
-             * we need to create a new levels array.
-             */
-            /* bidiBase.paraLevel is ok even if contextual multiple paragraphs,
-               since bidiBase is a line object                                     */
-            Arrays.fill(bidiBase.levels, start, length, bidiBase.paraLevel);
-
-            /* this new levels array is set for the line and reflects the WS run */
-            bidiBase.trailingWSStart = length;
-        }
-        if (length < bidiBase.levels.length) {
-            byte[] levels = new byte[length];
-            System.arraycopy(bidiBase.levels, 0, levels, 0, length);
-            return levels;
-        }
-        return bidiBase.levels;
-    }
-
-    static BidiRun getVisualRun(BidiBase bidiBase, int runIndex) {
-        int start = bidiBase.runs[runIndex].start;
-        int limit;
-        byte level = bidiBase.runs[runIndex].level;
-
-        if (runIndex > 0) {
-            limit = start +
-                    bidiBase.runs[runIndex].limit -
-                    bidiBase.runs[runIndex - 1].limit;
-        } else {
-            limit = start + bidiBase.runs[0].limit;
-        }
-        return new BidiRun(start, limit, level);
-    }
-
-    /* in trivial cases there is only one trivial run; called by getRuns() */
-    private static void getSingleRun(BidiBase bidiBase, byte level) {
-        /* simple, single-run case */
-        bidiBase.runs = bidiBase.simpleRuns;
-        bidiBase.runCount = 1;
-
-        /* fill and reorder the single run */
-        bidiBase.runs[0] = new BidiRun(0, bidiBase.length, level);
-    }
-
-    /* reorder the runs array (L2) ---------------------------------------------- */
-
-    /*
-     * Reorder the same-level runs in the runs array.
-     * Here, runCount>1 and maxLevel>=minLevel>=paraLevel.
-     * All the visualStart fields=logical start before reordering.
-     * The "odd" bits are not set yet.
-     *
-     * Reordering with this data structure lends itself to some handy shortcuts:
-     *
-     * Since each run is moved but not modified, and since at the initial maxLevel
-     * each sequence of same-level runs consists of only one run each, we
-     * don't need to do anything there and can predecrement maxLevel.
-     * In many simple cases, the reordering is thus done entirely in the
-     * index mapping.
-     * Also, reordering occurs only down to the lowest odd level that occurs,
-     * which is minLevel|1. However, if the lowest level itself is odd, then
-     * in the last reordering the sequence of the runs at this level or higher
-     * will be all runs, and we don't need the elaborate loop to search for them.
-     * This is covered by ++minLevel instead of minLevel|=1 followed
-     * by an extra reorder-all after the reorder-some loop.
-     * About a trailing WS run:
-     * Such a run would need special treatment because its level is not
-     * reflected in levels[] if this is not a paragraph object.
-     * Instead, all characters from trailingWSStart on are implicitly at
-     * paraLevel.
-     * However, for all maxLevel>paraLevel, this run will never be reordered
-     * and does not need to be taken into account. maxLevel==paraLevel is only reordered
-     * if minLevel==paraLevel is odd, which is done in the extra segment.
-     * This means that for the main reordering loop we don't need to consider
-     * this run and can --runCount. If it is later part of the all-runs
-     * reordering, then runCount is adjusted accordingly.
-     */
-    private static void reorderLine(BidiBase bidiBase, byte minLevel, byte maxLevel) {
-
-        /* nothing to do? */
-        if (maxLevel<=(minLevel|1)) {
-            return;
-        }
-
-        BidiRun[] runs;
-        BidiRun tempRun;
-        byte[] levels;
-        int firstRun, endRun, limitRun, runCount;
-
-        /*
-         * Reorder only down to the lowest odd level
-         * and reorder at an odd minLevel in a separate, simpler loop.
-         * See comments above for why minLevel is always incremented.
-         */
-        ++minLevel;
-
-        runs = bidiBase.runs;
-        levels = bidiBase.levels;
-        runCount = bidiBase.runCount;
-
-        /* do not include the WS run at paraLevel<=old minLevel except in the simple loop */
-        if (bidiBase.trailingWSStart < bidiBase.length) {
-            --runCount;
-        }
-
-        while (--maxLevel >= minLevel) {
-            firstRun = 0;
-
-            /* loop for all sequences of runs */
-            for ( ; ; ) {
-                /* look for a sequence of runs that are all at >=maxLevel */
-                /* look for the first run of such a sequence */
-                while (firstRun < runCount && levels[runs[firstRun].start] < maxLevel) {
-                    ++firstRun;
-                }
-                if (firstRun >= runCount) {
-                    break;  /* no more such runs */
-                }
-
-                /* look for the limit run of such a sequence (the run behind it) */
-                for (limitRun = firstRun; ++limitRun < runCount &&
-                      levels[runs[limitRun].start]>=maxLevel; ) {}
-
-                /* Swap the entire sequence of runs from firstRun to limitRun-1. */
-                endRun = limitRun - 1;
-                while (firstRun < endRun) {
-                    tempRun = runs[firstRun];
-                    runs[firstRun] = runs[endRun];
-                    runs[endRun] = tempRun;
-                    ++firstRun;
-                    --endRun;
-                }
-
-                if (limitRun == runCount) {
-                    break;  /* no more such runs */
-                } else {
-                    firstRun = limitRun + 1;
-                }
-            }
-        }
-
-        /* now do maxLevel==old minLevel (==odd!), see above */
-        if ((minLevel & 1) == 0) {
-            firstRun = 0;
-
-            /* include the trailing WS run in this complete reordering */
-            if (bidiBase.trailingWSStart == bidiBase.length) {
-                --runCount;
-            }
-
-            /* Swap the entire sequence of all runs. (endRun==runCount) */
-            while (firstRun < runCount) {
-                tempRun = runs[firstRun];
-                runs[firstRun] = runs[runCount];
-                runs[runCount] = tempRun;
-                ++firstRun;
-                --runCount;
-            }
-        }
-    }
-
-    /* compute the runs array --------------------------------------------------- */
-
-    static int getRunFromLogicalIndex(BidiBase bidiBase, int logicalIndex) {
-        BidiRun[] runs = bidiBase.runs;
-        int runCount = bidiBase.runCount, visualStart = 0, i, length, logicalStart;
-
-        for (i = 0; i < runCount; i++) {
-            length = runs[i].limit - visualStart;
-            logicalStart = runs[i].start;
-            if ((logicalIndex >= logicalStart) && (logicalIndex < (logicalStart+length))) {
-                return i;
-            }
-            visualStart += length;
-        }
-        /* we should never get here */
-        throw new IllegalStateException("Internal ICU error in getRunFromLogicalIndex");
-    }
-
-    /*
-     * Compute the runs array from the levels array.
-     * After getRuns() returns true, runCount is guaranteed to be >0
-     * and the runs are reordered.
-     * Odd-level runs have visualStart on their visual right edge and
-     * they progress visually to the left.
-     * If option OPTION_INSERT_MARKS is set, insertRemove will contain the
-     * sum of appropriate LRM/RLM_BEFORE/AFTER flags.
-     * If option OPTION_REMOVE_CONTROLS is set, insertRemove will contain the
-     * negative number of BiDi control characters within this run.
-     */
-    static void getRuns(BidiBase bidiBase) {
-        /*
-         * This method returns immediately if the runs are already set. This
-         * includes the case of length==0 (handled in setPara)..
-         */
-        if (bidiBase.runCount >= 0) {
-            return;
-        }
-        if (bidiBase.direction != BidiBase.MIXED) {
-            /* simple, single-run case - this covers length==0 */
-            /* bidiBase.paraLevel is ok even for contextual multiple paragraphs */
-            getSingleRun(bidiBase, bidiBase.paraLevel);
-        } else /* BidiBase.MIXED, length>0 */ {
-            /* mixed directionality */
-            int length = bidiBase.length, limit;
-            byte[] levels = bidiBase.levels;
-            int i, runCount;
-            byte level = -1;    /* initialize with no valid level */
-            /*
-             * If there are WS characters at the end of the line
-             * and the run preceding them has a level different from
-             * paraLevel, then they will form their own run at paraLevel (L1).
-             * Count them separately.
-             * We need some special treatment for this in order to not
-             * modify the levels array which a line Bidi object shares
-             * with its paragraph parent and its other line siblings.
-             * In other words, for the trailing WS, it may be
-             * levels[]!=paraLevel but we have to treat it like it were so.
-             */
-            limit = bidiBase.trailingWSStart;
-            /* count the runs, there is at least one non-WS run, and limit>0 */
-            runCount = 0;
-            for (i = 0; i < limit; ++i) {
-                /* increment runCount at the start of each run */
-                if (levels[i] != level) {
-                    ++runCount;
-                    level = levels[i];
-                }
-            }
-
-            /*
-             * We don't need to see if the last run can be merged with a trailing
-             * WS run because setTrailingWSStart() would have done that.
-             */
-            if (runCount == 1 && limit == length) {
-                /* There is only one non-WS run and no trailing WS-run. */
-                getSingleRun(bidiBase, levels[0]);
-            } else /* runCount>1 || limit 1 */
-                bidiBase.getRunsMemory(runCount);
-                runs = bidiBase.runsMemory;
-
-                /* set the runs */
-                /* FOOD FOR THOUGHT: this could be optimized, e.g.:
-                 * 464->444, 484->444, 575->555, 595->555
-                 * However, that would take longer. Check also how it would
-                 * interact with BiDi control removal and inserting Marks.
-                 */
-                runIndex = 0;
-
-                /* search for the run limits and initialize visualLimit values with the run lengths */
-                i = 0;
-                do {
-                    /* prepare this run */
-                    start = i;
-                    level = levels[i];
-                    if (level < minLevel) {
-                        minLevel = level;
-                    }
-                    if (level > maxLevel) {
-                        maxLevel = level;
-                    }
-
-                    /* look for the run limit */
-                    while (++i < limit && levels[i] == level) {}
-
-                    /* i is another run limit */
-                    runs[runIndex] = new BidiRun(start, i - start, level);
-                    ++runIndex;
-                } while (i < limit);
-
-                if (limit < length) {
-                    /* there is a separate WS run */
-                    runs[runIndex] = new BidiRun(limit, length - limit, bidiBase.paraLevel);
-                    /* For the trailing WS run, bidiBase.paraLevel is ok even
-                       if contextual multiple paragraphs.                   */
-                    if (bidiBase.paraLevel < minLevel) {
-                        minLevel = bidiBase.paraLevel;
-                    }
-                }
-
-                /* set the object fields */
-                bidiBase.runs = runs;
-                bidiBase.runCount = runCount;
-
-                reorderLine(bidiBase, minLevel, maxLevel);
-
-                /* now add the direction flags and adjust the visualLimit's to be just that */
-                /* this loop will also handle the trailing WS run */
-                limit = 0;
-                for (i = 0; i < runCount; ++i) {
-                    runs[i].level = levels[runs[i].start];
-                    limit = (runs[i].limit += limit);
-                }
-
-                /* Set the embedding level for the trailing WS run. */
-                /* For a RTL paragraph, it will be the *first* run in visual order. */
-                /* For the trailing WS run, bidiBase.paraLevel is ok even if
-                   contextual multiple paragraphs.                          */
-                if (runIndex < runCount) {
-                    int trailingRun = ((bidiBase.paraLevel & 1) != 0)? 0 : runIndex;
-                    runs[trailingRun].level = bidiBase.paraLevel;
-                }
-            }
-        }
-
-        /* handle insert LRM/RLM BEFORE/AFTER run */
-        if (bidiBase.insertPoints.size > 0) {
-            BidiBase.Point point;
-            int runIndex, ip;
-            for (ip = 0; ip < bidiBase.insertPoints.size; ip++) {
-                point = bidiBase.insertPoints.points[ip];
-                runIndex = getRunFromLogicalIndex(bidiBase, point.pos);
-                bidiBase.runs[runIndex].insertRemove |= point.flag;
-            }
-        }
-
-        /* handle remove BiDi control characters */
-        if (bidiBase.controlCount > 0) {
-            int runIndex, ic;
-            char c;
-            for (ic = 0; ic < bidiBase.length; ic++) {
-                c = bidiBase.text[ic];
-                if (BidiBase.IsBidiControlChar(c)) {
-                    runIndex = getRunFromLogicalIndex(bidiBase, ic);
-                    bidiBase.runs[runIndex].insertRemove--;
-                }
-            }
-        }
-    }
-
-    static int[] prepareReorder(byte[] levels, byte[] pMinLevel, byte[] pMaxLevel)
-    {
-        int start;
-        byte level, minLevel, maxLevel;
-
-        if (levels == null || levels.length <= 0) {
-            return null;
-        }
-
-        /* determine minLevel and maxLevel */
-        minLevel = BidiBase.MAX_EXPLICIT_LEVEL + 1;
-        maxLevel = 0;
-        for (start = levels.length; start>0; ) {
-            level = levels[--start];
-            if (level < 0 || level > (BidiBase.MAX_EXPLICIT_LEVEL + 1)) {
-                return null;
-            }
-            if (level < minLevel) {
-                minLevel = level;
-            }
-            if (level > maxLevel) {
-                maxLevel = level;
-            }
-        }
-        pMinLevel[0] = minLevel;
-        pMaxLevel[0] = maxLevel;
-
-        /* initialize the index map */
-        int[] indexMap = new int[levels.length];
-        for (start = levels.length; start > 0; ) {
-            --start;
-            indexMap[start] = start;
-        }
-
-        return indexMap;
-    }
-
-    static int[] reorderVisual(byte[] levels)
-    {
-        byte[] aMinLevel = new byte[1];
-        byte[] aMaxLevel = new byte[1];
-        int start, end, limit, temp;
-        byte minLevel, maxLevel;
-
-        int[] indexMap = prepareReorder(levels, aMinLevel, aMaxLevel);
-        if (indexMap == null) {
-            return null;
-        }
-
-        minLevel = aMinLevel[0];
-        maxLevel = aMaxLevel[0];
-
-        /* nothing to do? */
-        if (minLevel == maxLevel && (minLevel & 1) == 0) {
-            return indexMap;
-        }
-
-        /* reorder only down to the lowest odd level */
-        minLevel |= 1;
-
-        /* loop maxLevel..minLevel */
-        do {
-            start = 0;
-
-            /* loop for all sequences of levels to reorder at the current maxLevel */
-            for ( ; ; ) {
-                /* look for a sequence of levels that are all at >=maxLevel */
-                /* look for the first index of such a sequence */
-                while (start < levels.length && levels[start] < maxLevel) {
-                    ++start;
-                }
-                if (start >= levels.length) {
-                    break;  /* no more such runs */
-                }
-
-                /* look for the limit of such a sequence (the index behind it) */
-                for (limit = start; ++limit < levels.length && levels[limit] >= maxLevel; ) {}
-
-                /*
-                 * Swap the entire interval of indexes from start to limit-1.
-                 * We don't need to swap the levels for the purpose of this
-                 * algorithm: the sequence of levels that we look at does not
-                 * move anyway.
-                 */
-                end = limit - 1;
-                while (start < end) {
-                    temp = indexMap[start];
-                    indexMap[start] = indexMap[end];
-                    indexMap[end] = temp;
-
-                    ++start;
-                    --end;
-                }
-
-                if (limit == levels.length) {
-                    break;  /* no more such sequences */
-                } else {
-                    start = limit + 1;
-                }
-            }
-        } while (--maxLevel >= minLevel);
-
-        return indexMap;
-    }
-
-    static int[] getVisualMap(BidiBase bidiBase)
-    {
-        /* fill a visual-to-logical index map using the runs[] */
-        BidiRun[] runs = bidiBase.runs;
-        int logicalStart, visualStart, visualLimit;
-        int allocLength = bidiBase.length > bidiBase.resultLength ? bidiBase.length
-                                                          : bidiBase.resultLength;
-        int[] indexMap = new int[allocLength];
-
-        visualStart = 0;
-        int idx = 0;
-        for (int j = 0; j < bidiBase.runCount; ++j) {
-            logicalStart = runs[j].start;
-            visualLimit = runs[j].limit;
-            if (runs[j].isEvenRun()) {
-                do { /* LTR */
-                    indexMap[idx++] = logicalStart++;
-                } while (++visualStart < visualLimit);
-            } else {
-                logicalStart += visualLimit - visualStart;  /* logicalLimit */
-                do { /* RTL */
-                    indexMap[idx++] = --logicalStart;
-                } while (++visualStart < visualLimit);
-            }
-            /* visualStart==visualLimit; */
-        }
-
-        if (bidiBase.insertPoints.size > 0) {
-            int markFound = 0, runCount = bidiBase.runCount;
-            int insertRemove, i, j, k;
-            runs = bidiBase.runs;
-            /* count all inserted marks */
-            for (i = 0; i < runCount; i++) {
-                insertRemove = runs[i].insertRemove;
-                if ((insertRemove & (BidiBase.LRM_BEFORE|BidiBase.RLM_BEFORE)) > 0) {
-                    markFound++;
-                }
-                if ((insertRemove & (BidiBase.LRM_AFTER|BidiBase.RLM_AFTER)) > 0) {
-                    markFound++;
-                }
-            }
-            /* move back indexes by number of preceding marks */
-            k = bidiBase.resultLength;
-            for (i = runCount - 1; i >= 0 && markFound > 0; i--) {
-                insertRemove = runs[i].insertRemove;
-                if ((insertRemove & (BidiBase.LRM_AFTER|BidiBase.RLM_AFTER)) > 0) {
-                    indexMap[--k] = BidiBase.MAP_NOWHERE;
-                    markFound--;
-                }
-                visualStart = i > 0 ? runs[i-1].limit : 0;
-                for (j = runs[i].limit - 1; j >= visualStart && markFound > 0; j--) {
-                    indexMap[--k] = indexMap[j];
-                }
-                if ((insertRemove & (BidiBase.LRM_BEFORE|BidiBase.RLM_BEFORE)) > 0) {
-                    indexMap[--k] = BidiBase.MAP_NOWHERE;
-                    markFound--;
-                }
-            }
-        }
-        else if (bidiBase.controlCount > 0) {
-            int runCount = bidiBase.runCount, logicalEnd;
-            int insertRemove, length, i, j, k, m;
-            char uchar;
-            boolean evenRun;
-            runs = bidiBase.runs;
-            visualStart = 0;
-            /* move forward indexes by number of preceding controls */
-            k = 0;
-            for (i = 0; i < runCount; i++, visualStart += length) {
-                length = runs[i].limit - visualStart;
-                insertRemove = runs[i].insertRemove;
-                /* if no control found yet, nothing to do in this run */
-                if ((insertRemove == 0) && (k == visualStart)) {
-                    k += length;
-                    continue;
-                }
-                /* if no control in this run */
-                if (insertRemove == 0) {
-                    visualLimit = runs[i].limit;
-                    for (j = visualStart; j < visualLimit; j++) {
-                        indexMap[k++] = indexMap[j];
-                    }
-                    continue;
-                }
-                logicalStart = runs[i].start;
-                evenRun = runs[i].isEvenRun();
-                logicalEnd = logicalStart + length - 1;
-                for (j = 0; j < length; j++) {
-                    m = evenRun ? logicalStart + j : logicalEnd - j;
-                    uchar = bidiBase.text[m];
-                    if (!BidiBase.IsBidiControlChar(uchar)) {
-                        indexMap[k++] = m;
-                    }
-                }
-            }
-        }
-        if (allocLength == bidiBase.resultLength) {
-            return indexMap;
-        }
-        int[] newMap = new int[bidiBase.resultLength];
-        System.arraycopy(indexMap, 0, newMap, 0, bidiBase.resultLength);
-        return newMap;
-    }
-
-}
--- /dev/null	2020-01-10 15:57:54.000000000 -0800
+++ new/src/java.base/share/classes/jdk/internal/icu/text/BidiLine.java	2020-01-10 15:57:53.000000000 -0800
@@ -0,0 +1,835 @@
+/*
+ * Copyright (c) 2009, 2020, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.  Oracle designates this
+ * particular file as subject to the "Classpath" exception as provided
+ * by Oracle in the LICENSE file that accompanied this code.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+/*
+*******************************************************************************
+*   Copyright (C) 2001-2014, International Business Machines
+*   Corporation and others.  All Rights Reserved.
+*******************************************************************************
+*/
+/* Written by Simon Montagu, Matitiahu Allouche
+ * (ported from C code written by Markus W. Scherer)
+ */
+
+package jdk.internal.icu.text;
+
+import java.text.Bidi;
+import java.util.Arrays;
+
+final class BidiLine {
+
+    /*
+     * General remarks about the functions in this file:
+     *
+     * These functions deal with the aspects of potentially mixed-directional
+     * text in a single paragraph or in a line of a single paragraph
+     * which has already been processed according to
+     * the Unicode 3.0 Bidi algorithm as defined in
+     * http://www.unicode.org/unicode/reports/tr9/ , version 13,
+     * also described in The Unicode Standard, Version 4.0.1 .
+     *
+     * This means that there is a Bidi object with a levels
+     * and a dirProps array.
+     * paraLevel and direction are also set.
+     * Only if the length of the text is zero, then levels==dirProps==NULL.
+     *
+     * The overall directionality of the paragraph
+     * or line is used to bypass the reordering steps if possible.
+     * Even purely RTL text does not need reordering there because
+     * the getLogical/VisualIndex() methods can compute the
+     * index on the fly in such a case.
+     *
+     * The implementation of the access to same-level-runs and of the reordering
+     * do attempt to provide better performance and less memory usage compared to
+     * a direct implementation of especially rule (L2) with an array of
+     * one (32-bit) integer per text character.
+     *
+     * Here, the levels array is scanned as soon as necessary, and a vector of
+     * same-level-runs is created. Reordering then is done on this vector.
+     * For each run of text positions that were resolved to the same level,
+     * only 8 bytes are stored: the first text position of the run and the visual
+     * position behind the run after reordering.
+     * One sign bit is used to hold the directionality of the run.
+     * This is inefficient if there are many very short runs. If the average run
+     * length is <2, then this uses more memory.
+     *
+     * In a further attempt to save memory, the levels array is never changed
+     * after all the resolution rules (Xn, Wn, Nn, In).
+     * Many methods have to consider the field trailingWSStart:
+     * if it is less than length, then there is an implicit trailing run
+     * at the paraLevel,
+     * which is not reflected in the levels array.
+     * This allows a line Bidi object to use the same levels array as
+     * its paragraph parent object.
+     *
+     * When a Bidi object is created for a line of a paragraph, then the
+     * paragraph's levels and dirProps arrays are reused by way of setting
+     * a pointer into them, not by copying. This again saves memory and forbids to
+     * change the now shared levels for (L1).
+     */
+
+    /* handle trailing WS (L1) -------------------------------------------------- */
+
+    /*
+     * setTrailingWSStart() sets the start index for a trailing
+     * run of WS in the line. This is necessary because we do not modify
+     * the paragraph's levels array that we just point into.
+     * Using trailingWSStart is another form of performing (L1).
+     *
+     * To make subsequent operations easier, we also include the run
+     * before the WS if it is at the paraLevel - we merge the two here.
+     *
+     * This method is called only from setLine(), so paraLevel is
+     * set correctly for the line even when contextual multiple paragraphs.
+     */
+
+    static void setTrailingWSStart(BidiBase bidiBase)
+    {
+        byte[] dirProps = bidiBase.dirProps;
+        byte[] levels = bidiBase.levels;
+        int start = bidiBase.length;
+        byte paraLevel = bidiBase.paraLevel;
+
+        /* If the line is terminated by a block separator, all preceding WS etc...
+           are already set to paragraph level.
+           Setting trailingWSStart to pBidi->length will avoid changing the
+           level of B chars from 0 to paraLevel in getLevels when
+           orderParagraphsLTR==TRUE
+        */
+        if (dirProps[start - 1] == BidiBase.B) {
+            bidiBase.trailingWSStart = start;   /* currently == bidiBase.length */
+            return;
+        }
+        /* go backwards across all WS, BN, explicit codes */
+        while (start > 0 &&
+                (BidiBase.DirPropFlag(dirProps[start - 1]) & BidiBase.MASK_WS) != 0) {
+            --start;
+        }
+
+        /* if the WS run can be merged with the previous run then do so here */
+        while (start > 0 && levels[start - 1] == paraLevel) {
+            --start;
+        }
+
+        bidiBase.trailingWSStart=start;
+    }
+
+    static Bidi setLine(BidiBase paraBidi,
+                              Bidi newBidi, BidiBase lineBidi,
+                              int start, int limit) {
+        int length;
+
+        /* set the values in lineBidi from its paraBidi parent */
+        /* class members are already initialized to 0 */
+        // lineBidi.paraBidi = null;        /* mark unfinished setLine */
+        // lineBidi.flags = 0;
+        // lineBidi.controlCount = 0;
+
+        length = lineBidi.length = lineBidi.originalLength =
+                lineBidi.resultLength = limit - start;
+
+        lineBidi.text = new char[length];
+        System.arraycopy(paraBidi.text, start, lineBidi.text, 0, length);
+        lineBidi.paraLevel = paraBidi.GetParaLevelAt(start);
+        lineBidi.paraCount = paraBidi.paraCount;
+        lineBidi.runs = new BidiRun[0];
+        lineBidi.reorderingMode = paraBidi.reorderingMode;
+        lineBidi.reorderingOptions = paraBidi.reorderingOptions;
+        if (paraBidi.controlCount > 0) {
+            int j;
+            for (j = start; j < limit; j++) {
+                if (BidiBase.IsBidiControlChar(paraBidi.text[j])) {
+                    lineBidi.controlCount++;
+                }
+            }
+            lineBidi.resultLength -= lineBidi.controlCount;
+        }
+        /* copy proper subset of DirProps */
+        lineBidi.getDirPropsMemory(length);
+        lineBidi.dirProps = lineBidi.dirPropsMemory;
+        System.arraycopy(paraBidi.dirProps, start, lineBidi.dirProps, 0,
+                         length);
+        /* copy proper subset of Levels */
+        lineBidi.getLevelsMemory(length);
+        lineBidi.levels = lineBidi.levelsMemory;
+        System.arraycopy(paraBidi.levels, start, lineBidi.levels, 0,
+                         length);
+        lineBidi.runCount = -1;
+
+        if (paraBidi.direction != BidiBase.MIXED) {
+            /* the parent is already trivial */
+            lineBidi.direction = paraBidi.direction;
+
+            /*
+             * The parent's levels are all either
+             * implicitly or explicitly ==paraLevel;
+             * do the same here.
+             */
+            if (paraBidi.trailingWSStart <= start) {
+                lineBidi.trailingWSStart = 0;
+            } else if (paraBidi.trailingWSStart < limit) {
+                lineBidi.trailingWSStart = paraBidi.trailingWSStart - start;
+            } else {
+                lineBidi.trailingWSStart = length;
+            }
+        } else {
+            byte[] levels = lineBidi.levels;
+            int i, trailingWSStart;
+            byte level;
+
+            setTrailingWSStart(lineBidi);
+            trailingWSStart = lineBidi.trailingWSStart;
+
+            /* recalculate lineBidiBase.direction */
+            if (trailingWSStart == 0) {
+                /* all levels are at paraLevel */
+                lineBidi.direction = (byte)(lineBidi.paraLevel & 1);
+            } else {
+                /* get the level of the first character */
+                level = (byte)(levels[0] & 1);
+
+                /* if there is anything of a different level, then the line
+                   is mixed */
+                if (trailingWSStart < length &&
+                    (lineBidi.paraLevel & 1) != level) {
+                    /* the trailing WS is at paraLevel, which differs from
+                       levels[0] */
+                    lineBidi.direction = BidiBase.MIXED;
+                } else {
+                    /* see if levels[1..trailingWSStart-1] have the same
+                       direction as levels[0] and paraLevel */
+                    for (i = 1; ; i++) {
+                        if (i == trailingWSStart) {
+                            /* the direction values match those in level */
+                            lineBidi.direction = level;
+                            break;
+                        } else if ((levels[i] & 1) != level) {
+                            lineBidi.direction = BidiBase.MIXED;
+                            break;
+                        }
+                    }
+                }
+            }
+
+            switch(lineBidi.direction) {
+                case Bidi.DIRECTION_LEFT_TO_RIGHT:
+                    /* make sure paraLevel is even */
+                    lineBidi.paraLevel = (byte)
+                        ((lineBidi.paraLevel + 1) & ~1);
+
+                    /* all levels are implicitly at paraLevel (important for
+                       getLevels()) */
+                    lineBidi.trailingWSStart = 0;
+                    break;
+                case Bidi.DIRECTION_RIGHT_TO_LEFT:
+                    /* make sure paraLevel is odd */
+                    lineBidi.paraLevel |= 1;
+
+                    /* all levels are implicitly at paraLevel (important for
+                       getLevels()) */
+                    lineBidi.trailingWSStart = 0;
+                    break;
+                default:
+                    break;
+            }
+        }
+
+        lineBidi.paraBidi = paraBidi;     /* mark successful setLine */
+
+        return newBidi;
+    }
+
+    static byte getLevelAt(BidiBase bidiBase, int charIndex)
+    {
+        /* return paraLevel if in the trailing WS run, otherwise the real level */
+        if (bidiBase.direction != BidiBase.MIXED || charIndex >= bidiBase.trailingWSStart) {
+            return bidiBase.GetParaLevelAt(charIndex);
+        } else {
+            return bidiBase.levels[charIndex];
+        }
+    }
+
+    static byte[] getLevels(BidiBase bidiBase)
+    {
+        int start = bidiBase.trailingWSStart;
+        int length = bidiBase.length;
+
+        if (start != length) {
+            /* the current levels array does not reflect the WS run */
+            /*
+             * After the previous if(), we know that the levels array
+             * has an implicit trailing WS run and therefore does not fully
+             * reflect itself all the levels.
+             * This must be a Bidi object for a line, and
+             * we need to create a new levels array.
+             */
+            /* bidiBase.paraLevel is ok even if contextual multiple paragraphs,
+               since bidiBase is a line object                                     */
+            Arrays.fill(bidiBase.levels, start, length, bidiBase.paraLevel);
+
+            /* this new levels array is set for the line and reflects the WS run */
+            bidiBase.trailingWSStart = length;
+        }
+        if (length < bidiBase.levels.length) {
+            byte[] levels = new byte[length];
+            System.arraycopy(bidiBase.levels, 0, levels, 0, length);
+            return levels;
+        }
+        return bidiBase.levels;
+    }
+
+    static BidiRun getVisualRun(BidiBase bidiBase, int runIndex) {
+        int start = bidiBase.runs[runIndex].start;
+        int limit;
+        byte level = bidiBase.runs[runIndex].level;
+
+        if (runIndex > 0) {
+            limit = start +
+                    bidiBase.runs[runIndex].limit -
+                    bidiBase.runs[runIndex - 1].limit;
+        } else {
+            limit = start + bidiBase.runs[0].limit;
+        }
+        return new BidiRun(start, limit, level);
+    }
+
+    /* in trivial cases there is only one trivial run; called by getRuns() */
+    private static void getSingleRun(BidiBase bidiBase, byte level) {
+        /* simple, single-run case */
+        bidiBase.runs = bidiBase.simpleRuns;
+        bidiBase.runCount = 1;
+
+        /* fill and reorder the single run */
+        bidiBase.runs[0] = new BidiRun(0, bidiBase.length, level);
+    }
+
+    /* reorder the runs array (L2) ---------------------------------------------- */
+
+    /*
+     * Reorder the same-level runs in the runs array.
+     * Here, runCount>1 and maxLevel>=minLevel>=paraLevel.
+     * All the visualStart fields=logical start before reordering.
+     * The "odd" bits are not set yet.
+     *
+     * Reordering with this data structure lends itself to some handy shortcuts:
+     *
+     * Since each run is moved but not modified, and since at the initial maxLevel
+     * each sequence of same-level runs consists of only one run each, we
+     * don't need to do anything there and can predecrement maxLevel.
+     * In many simple cases, the reordering is thus done entirely in the
+     * index mapping.
+     * Also, reordering occurs only down to the lowest odd level that occurs,
+     * which is minLevel|1. However, if the lowest level itself is odd, then
+     * in the last reordering the sequence of the runs at this level or higher
+     * will be all runs, and we don't need the elaborate loop to search for them.
+     * This is covered by ++minLevel instead of minLevel|=1 followed
+     * by an extra reorder-all after the reorder-some loop.
+     * About a trailing WS run:
+     * Such a run would need special treatment because its level is not
+     * reflected in levels[] if this is not a paragraph object.
+     * Instead, all characters from trailingWSStart on are implicitly at
+     * paraLevel.
+     * However, for all maxLevel>paraLevel, this run will never be reordered
+     * and does not need to be taken into account. maxLevel==paraLevel is only reordered
+     * if minLevel==paraLevel is odd, which is done in the extra segment.
+     * This means that for the main reordering loop we don't need to consider
+     * this run and can --runCount. If it is later part of the all-runs
+     * reordering, then runCount is adjusted accordingly.
+     */
+    private static void reorderLine(BidiBase bidiBase, byte minLevel, byte maxLevel) {
+
+        /* nothing to do? */
+        if (maxLevel<=(minLevel|1)) {
+            return;
+        }
+
+        BidiRun[] runs;
+        BidiRun tempRun;
+        byte[] levels;
+        int firstRun, endRun, limitRun, runCount;
+
+        /*
+         * Reorder only down to the lowest odd level
+         * and reorder at an odd minLevel in a separate, simpler loop.
+         * See comments above for why minLevel is always incremented.
+         */
+        ++minLevel;
+
+        runs = bidiBase.runs;
+        levels = bidiBase.levels;
+        runCount = bidiBase.runCount;
+
+        /* do not include the WS run at paraLevel<=old minLevel except in the simple loop */
+        if (bidiBase.trailingWSStart < bidiBase.length) {
+            --runCount;
+        }
+
+        while (--maxLevel >= minLevel) {
+            firstRun = 0;
+
+            /* loop for all sequences of runs */
+            for ( ; ; ) {
+                /* look for a sequence of runs that are all at >=maxLevel */
+                /* look for the first run of such a sequence */
+                while (firstRun < runCount && levels[runs[firstRun].start] < maxLevel) {
+                    ++firstRun;
+                }
+                if (firstRun >= runCount) {
+                    break;  /* no more such runs */
+                }
+
+                /* look for the limit run of such a sequence (the run behind it) */
+                for (limitRun = firstRun; ++limitRun < runCount &&
+                      levels[runs[limitRun].start]>=maxLevel; ) {}
+
+                /* Swap the entire sequence of runs from firstRun to limitRun-1. */
+                endRun = limitRun - 1;
+                while (firstRun < endRun) {
+                    tempRun = runs[firstRun];
+                    runs[firstRun] = runs[endRun];
+                    runs[endRun] = tempRun;
+                    ++firstRun;
+                    --endRun;
+                }
+
+                if (limitRun == runCount) {
+                    break;  /* no more such runs */
+                } else {
+                    firstRun = limitRun + 1;
+                }
+            }
+        }
+
+        /* now do maxLevel==old minLevel (==odd!), see above */
+        if ((minLevel & 1) == 0) {
+            firstRun = 0;
+
+            /* include the trailing WS run in this complete reordering */
+            if (bidiBase.trailingWSStart == bidiBase.length) {
+                --runCount;
+            }
+
+            /* Swap the entire sequence of all runs. (endRun==runCount) */
+            while (firstRun < runCount) {
+                tempRun = runs[firstRun];
+                runs[firstRun] = runs[runCount];
+                runs[runCount] = tempRun;
+                ++firstRun;
+                --runCount;
+            }
+        }
+    }
+
+    /* compute the runs array --------------------------------------------------- */
+
+    static int getRunFromLogicalIndex(BidiBase bidiBase, int logicalIndex) {
+        BidiRun[] runs = bidiBase.runs;
+        int runCount = bidiBase.runCount, visualStart = 0, i, length, logicalStart;
+
+        for (i = 0; i < runCount; i++) {
+            length = runs[i].limit - visualStart;
+            logicalStart = runs[i].start;
+            if ((logicalIndex >= logicalStart) && (logicalIndex < (logicalStart+length))) {
+                return i;
+            }
+            visualStart += length;
+        }
+        /* we should never get here */
+        throw new IllegalStateException("Internal ICU error in getRunFromLogicalIndex");
+    }
+
+    /*
+     * Compute the runs array from the levels array.
+     * After getRuns() returns true, runCount is guaranteed to be >0
+     * and the runs are reordered.
+     * Odd-level runs have visualStart on their visual right edge and
+     * they progress visually to the left.
+     * If option OPTION_INSERT_MARKS is set, insertRemove will contain the
+     * sum of appropriate LRM/RLM_BEFORE/AFTER flags.
+     * If option OPTION_REMOVE_CONTROLS is set, insertRemove will contain the
+     * negative number of BiDi control characters within this run.
+     */
+    static void getRuns(BidiBase bidiBase) {
+        /*
+         * This method returns immediately if the runs are already set. This
+         * includes the case of length==0 (handled in setPara)..
+         */
+        if (bidiBase.runCount >= 0) {
+            return;
+        }
+        if (bidiBase.direction != BidiBase.MIXED) {
+            /* simple, single-run case - this covers length==0 */
+            /* bidiBase.paraLevel is ok even for contextual multiple paragraphs */
+            getSingleRun(bidiBase, bidiBase.paraLevel);
+        } else /* BidiBase.MIXED, length>0 */ {
+            /* mixed directionality */
+            int length = bidiBase.length, limit;
+            byte[] levels = bidiBase.levels;
+            int i, runCount;
+            byte level = -1;    /* initialize with no valid level */
+            /*
+             * If there are WS characters at the end of the line
+             * and the run preceding them has a level different from
+             * paraLevel, then they will form their own run at paraLevel (L1).
+             * Count them separately.
+             * We need some special treatment for this in order to not
+             * modify the levels array which a line Bidi object shares
+             * with its paragraph parent and its other line siblings.
+             * In other words, for the trailing WS, it may be
+             * levels[]!=paraLevel but we have to treat it like it were so.
+             */
+            limit = bidiBase.trailingWSStart;
+            /* count the runs, there is at least one non-WS run, and limit>0 */
+            runCount = 0;
+            for (i = 0; i < limit; ++i) {
+                /* increment runCount at the start of each run */
+                if (levels[i] != level) {
+                    ++runCount;
+                    level = levels[i];
+                }
+            }
+
+            /*
+             * We don't need to see if the last run can be merged with a trailing
+             * WS run because setTrailingWSStart() would have done that.
+             */
+            if (runCount == 1 && limit == length) {
+                /* There is only one non-WS run and no trailing WS-run. */
+                getSingleRun(bidiBase, levels[0]);
+            } else /* runCount>1 || limit 1 */
+                bidiBase.getRunsMemory(runCount);
+                runs = bidiBase.runsMemory;
+
+                /* set the runs */
+                /* FOOD FOR THOUGHT: this could be optimized, e.g.:
+                 * 464->444, 484->444, 575->555, 595->555
+                 * However, that would take longer. Check also how it would
+                 * interact with BiDi control removal and inserting Marks.
+                 */
+                runIndex = 0;
+
+                /* search for the run limits and initialize visualLimit values with the run lengths */
+                i = 0;
+                do {
+                    /* prepare this run */
+                    start = i;
+                    level = levels[i];
+                    if (level < minLevel) {
+                        minLevel = level;
+                    }
+                    if (level > maxLevel) {
+                        maxLevel = level;
+                    }
+
+                    /* look for the run limit */
+                    while (++i < limit && levels[i] == level) {}
+
+                    /* i is another run limit */
+                    runs[runIndex] = new BidiRun(start, i - start, level);
+                    ++runIndex;
+                } while (i < limit);
+
+                if (limit < length) {
+                    /* there is a separate WS run */
+                    runs[runIndex] = new BidiRun(limit, length - limit, bidiBase.paraLevel);
+                    /* For the trailing WS run, bidiBase.paraLevel is ok even
+                       if contextual multiple paragraphs.                   */
+                    if (bidiBase.paraLevel < minLevel) {
+                        minLevel = bidiBase.paraLevel;
+                    }
+                }
+
+                /* set the object fields */
+                bidiBase.runs = runs;
+                bidiBase.runCount = runCount;
+
+                reorderLine(bidiBase, minLevel, maxLevel);
+
+                /* now add the direction flags and adjust the visualLimit's to be just that */
+                /* this loop will also handle the trailing WS run */
+                limit = 0;
+                for (i = 0; i < runCount; ++i) {
+                    runs[i].level = levels[runs[i].start];
+                    limit = (runs[i].limit += limit);
+                }
+
+                /* Set the embedding level for the trailing WS run. */
+                /* For a RTL paragraph, it will be the *first* run in visual order. */
+                /* For the trailing WS run, bidiBase.paraLevel is ok even if
+                   contextual multiple paragraphs.                          */
+                if (runIndex < runCount) {
+                    int trailingRun = ((bidiBase.paraLevel & 1) != 0)? 0 : runIndex;
+                    runs[trailingRun].level = bidiBase.paraLevel;
+                }
+            }
+        }
+
+        /* handle insert LRM/RLM BEFORE/AFTER run */
+        if (bidiBase.insertPoints.size > 0) {
+            BidiBase.Point point;
+            int runIndex, ip;
+            for (ip = 0; ip < bidiBase.insertPoints.size; ip++) {
+                point = bidiBase.insertPoints.points[ip];
+                runIndex = getRunFromLogicalIndex(bidiBase, point.pos);
+                bidiBase.runs[runIndex].insertRemove |= point.flag;
+            }
+        }
+
+        /* handle remove BiDi control characters */
+        if (bidiBase.controlCount > 0) {
+            int runIndex, ic;
+            char c;
+            for (ic = 0; ic < bidiBase.length; ic++) {
+                c = bidiBase.text[ic];
+                if (BidiBase.IsBidiControlChar(c)) {
+                    runIndex = getRunFromLogicalIndex(bidiBase, ic);
+                    bidiBase.runs[runIndex].insertRemove--;
+                }
+            }
+        }
+    }
+
+    static int[] prepareReorder(byte[] levels, byte[] pMinLevel, byte[] pMaxLevel)
+    {
+        int start;
+        byte level, minLevel, maxLevel;
+
+        if (levels == null || levels.length <= 0) {
+            return null;
+        }
+
+        /* determine minLevel and maxLevel */
+        minLevel = BidiBase.MAX_EXPLICIT_LEVEL + 1;
+        maxLevel = 0;
+        for (start = levels.length; start>0; ) {
+            level = levels[--start];
+            if (level < 0 || level > (BidiBase.MAX_EXPLICIT_LEVEL + 1)) {
+                return null;
+            }
+            if (level < minLevel) {
+                minLevel = level;
+            }
+            if (level > maxLevel) {
+                maxLevel = level;
+            }
+        }
+        pMinLevel[0] = minLevel;
+        pMaxLevel[0] = maxLevel;
+
+        /* initialize the index map */
+        int[] indexMap = new int[levels.length];
+        for (start = levels.length; start > 0; ) {
+            --start;
+            indexMap[start] = start;
+        }
+
+        return indexMap;
+    }
+
+    static int[] reorderVisual(byte[] levels)
+    {
+        byte[] aMinLevel = new byte[1];
+        byte[] aMaxLevel = new byte[1];
+        int start, end, limit, temp;
+        byte minLevel, maxLevel;
+
+        int[] indexMap = prepareReorder(levels, aMinLevel, aMaxLevel);
+        if (indexMap == null) {
+            return null;
+        }
+
+        minLevel = aMinLevel[0];
+        maxLevel = aMaxLevel[0];
+
+        /* nothing to do? */
+        if (minLevel == maxLevel && (minLevel & 1) == 0) {
+            return indexMap;
+        }
+
+        /* reorder only down to the lowest odd level */
+        minLevel |= 1;
+
+        /* loop maxLevel..minLevel */
+        do {
+            start = 0;
+
+            /* loop for all sequences of levels to reorder at the current maxLevel */
+            for ( ; ; ) {
+                /* look for a sequence of levels that are all at >=maxLevel */
+                /* look for the first index of such a sequence */
+                while (start < levels.length && levels[start] < maxLevel) {
+                    ++start;
+                }
+                if (start >= levels.length) {
+                    break;  /* no more such runs */
+                }
+
+                /* look for the limit of such a sequence (the index behind it) */
+                for (limit = start; ++limit < levels.length && levels[limit] >= maxLevel; ) {}
+
+                /*
+                 * Swap the entire interval of indexes from start to limit-1.
+                 * We don't need to swap the levels for the purpose of this
+                 * algorithm: the sequence of levels that we look at does not
+                 * move anyway.
+                 */
+                end = limit - 1;
+                while (start < end) {
+                    temp = indexMap[start];
+                    indexMap[start] = indexMap[end];
+                    indexMap[end] = temp;
+
+                    ++start;
+                    --end;
+                }
+
+                if (limit == levels.length) {
+                    break;  /* no more such sequences */
+                } else {
+                    start = limit + 1;
+                }
+            }
+        } while (--maxLevel >= minLevel);
+
+        return indexMap;
+    }
+
+    static int[] getVisualMap(BidiBase bidiBase)
+    {
+        /* fill a visual-to-logical index map using the runs[] */
+        BidiRun[] runs = bidiBase.runs;
+        int logicalStart, visualStart, visualLimit;
+        int allocLength = bidiBase.length > bidiBase.resultLength ? bidiBase.length
+                                                          : bidiBase.resultLength;
+        int[] indexMap = new int[allocLength];
+
+        visualStart = 0;
+        int idx = 0;
+        for (int j = 0; j < bidiBase.runCount; ++j) {
+            logicalStart = runs[j].start;
+            visualLimit = runs[j].limit;
+            if (runs[j].isEvenRun()) {
+                do { /* LTR */
+                    indexMap[idx++] = logicalStart++;
+                } while (++visualStart < visualLimit);
+            } else {
+                logicalStart += visualLimit - visualStart;  /* logicalLimit */
+                do { /* RTL */
+                    indexMap[idx++] = --logicalStart;
+                } while (++visualStart < visualLimit);
+            }
+            /* visualStart==visualLimit; */
+        }
+
+        if (bidiBase.insertPoints.size > 0) {
+            int markFound = 0, runCount = bidiBase.runCount;
+            int insertRemove, i, j, k;
+            runs = bidiBase.runs;
+            /* count all inserted marks */
+            for (i = 0; i < runCount; i++) {
+                insertRemove = runs[i].insertRemove;
+                if ((insertRemove & (BidiBase.LRM_BEFORE|BidiBase.RLM_BEFORE)) > 0) {
+                    markFound++;
+                }
+                if ((insertRemove & (BidiBase.LRM_AFTER|BidiBase.RLM_AFTER)) > 0) {
+                    markFound++;
+                }
+            }
+            /* move back indexes by number of preceding marks */
+            k = bidiBase.resultLength;
+            for (i = runCount - 1; i >= 0 && markFound > 0; i--) {
+                insertRemove = runs[i].insertRemove;
+                if ((insertRemove & (BidiBase.LRM_AFTER|BidiBase.RLM_AFTER)) > 0) {
+                    indexMap[--k] = BidiBase.MAP_NOWHERE;
+                    markFound--;
+                }
+                visualStart = i > 0 ? runs[i-1].limit : 0;
+                for (j = runs[i].limit - 1; j >= visualStart && markFound > 0; j--) {
+                    indexMap[--k] = indexMap[j];
+                }
+                if ((insertRemove & (BidiBase.LRM_BEFORE|BidiBase.RLM_BEFORE)) > 0) {
+                    indexMap[--k] = BidiBase.MAP_NOWHERE;
+                    markFound--;
+                }
+            }
+        }
+        else if (bidiBase.controlCount > 0) {
+            int runCount = bidiBase.runCount, logicalEnd;
+            int insertRemove, length, i, j, k, m;
+            char uchar;
+            boolean evenRun;
+            runs = bidiBase.runs;
+            visualStart = 0;
+            /* move forward indexes by number of preceding controls */
+            k = 0;
+            for (i = 0; i < runCount; i++, visualStart += length) {
+                length = runs[i].limit - visualStart;
+                insertRemove = runs[i].insertRemove;
+                /* if no control found yet, nothing to do in this run */
+                if ((insertRemove == 0) && (k == visualStart)) {
+                    k += length;
+                    continue;
+                }
+                /* if no control in this run */
+                if (insertRemove == 0) {
+                    visualLimit = runs[i].limit;
+                    for (j = visualStart; j < visualLimit; j++) {
+                        indexMap[k++] = indexMap[j];
+                    }
+                    continue;
+                }
+                logicalStart = runs[i].start;
+                evenRun = runs[i].isEvenRun();
+                logicalEnd = logicalStart + length - 1;
+                for (j = 0; j < length; j++) {
+                    m = evenRun ? logicalStart + j : logicalEnd - j;
+                    uchar = bidiBase.text[m];
+                    if (!BidiBase.IsBidiControlChar(uchar)) {
+                        indexMap[k++] = m;
+                    }
+                }
+            }
+        }
+        if (allocLength == bidiBase.resultLength) {
+            return indexMap;
+        }
+        int[] newMap = new int[bidiBase.resultLength];
+        System.arraycopy(indexMap, 0, newMap, 0, bidiBase.resultLength);
+        return newMap;
+    }
+
+}
--- old/src/java.base/share/classes/sun/text/bidi/BidiRun.java	2020-01-10 15:57:55.000000000 -0800
+++ /dev/null	2020-01-10 15:57:55.000000000 -0800
@@ -1,124 +0,0 @@
-/*
- * Copyright (c) 2009, Oracle and/or its affiliates. All rights reserved.
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.  Oracle designates this
- * particular file as subject to the "Classpath" exception as provided
- * by Oracle in the LICENSE file that accompanied this code.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- */
-/*
- *******************************************************************************
- * (C) Copyright IBM Corp. and others, 1996-2009 - All Rights Reserved         *
- *                                                                             *
- * The original version of this source code and documentation is copyrighted   *
- * and owned by IBM, These materials are provided under terms of a License     *
- * Agreement between IBM and Sun. This technology is protected by multiple     *
- * US and International patents. This notice and attribution to IBM may not    *
- * to removed.                                                                 *
- *******************************************************************************
- */
-/* Written by Simon Montagu, Matitiahu Allouche
- * (ported from C code written by Markus W. Scherer)
- */
-
-package sun.text.bidi;
-
-/**
- * A BidiRun represents a sequence of characters at the same embedding level.
- * The Bidi algorithm decomposes a piece of text into sequences of characters
- * at the same embedding level, each such sequence is called a "run".
- *
- * 
A BidiRun represents such a run by storing its essential properties,
- * but does not duplicate the characters which form the run.
- *
- * 
The "limit" of the run is the position just after the
- * last character, i.e., one more than that position.
- *
- * 
This class has no public constructor, and its members cannot be
- * modified by users.
- *
- * @see com.ibm.icu.text.Bidi
- */
-class BidiRun {
-
-    int start;              /* first logical position of the run */
-    int limit;              /* last visual position of the run +1 */
-    int insertRemove;       /* if >0, flags for inserting LRM/RLM before/after run,
-                               if <0, count of bidi controls within run            */
-    byte level;
-
-    /*
-     * Default constructor
-     *
-     * Note that members start and limit of a run instance have different
-     * meanings depending whether the run is part of the runs array of a Bidi
-     * object, or if it is a reference returned by getVisualRun() or
-     * getLogicalRun().
-     * For a member of the runs array of a Bidi object,
-     *   - start is the first logical position of the run in the source text.
-     *   - limit is one after the last visual position of the run.
-     * For a reference returned by getLogicalRun() or getVisualRun(),
-     *   - start is the first logical position of the run in the source text.
-     *   - limit is one after the last logical position of the run.
-     */
-    BidiRun()
-    {
-        this(0, 0, (byte)0);
-    }
-
-    /*
-     * Constructor
-     */
-    BidiRun(int start, int limit, byte embeddingLevel)
-    {
-        this.start = start;
-        this.limit = limit;
-        this.level = embeddingLevel;
-    }
-
-    /*
-     * Copy the content of a BidiRun instance
-     */
-    void copyFrom(BidiRun run)
-    {
-        this.start = run.start;
-        this.limit = run.limit;
-        this.level = run.level;
-        this.insertRemove = run.insertRemove;
-    }
-
-    /**
-     * Get level of run
-     */
-    byte getEmbeddingLevel()
-    {
-        return level;
-    }
-
-    /**
-     * Check if run level is even
-     * @return true if the embedding level of this run is even, i.e. it is a
-     *  left-to-right run.
-     */
-    boolean isEvenRun()
-    {
-        return (level & 1) == 0;
-    }
-
-}
--- /dev/null	2020-01-10 15:57:55.000000000 -0800
+++ new/src/java.base/share/classes/jdk/internal/icu/text/BidiRun.java	2020-01-10 15:57:55.000000000 -0800
@@ -0,0 +1,124 @@
+/*
+ * Copyright (c) 2009, 2020, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.  Oracle designates this
+ * particular file as subject to the "Classpath" exception as provided
+ * by Oracle in the LICENSE file that accompanied this code.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+/*
+ *******************************************************************************
+ * (C) Copyright IBM Corp. and others, 1996-2009 - All Rights Reserved         *
+ *                                                                             *
+ * The original version of this source code and documentation is copyrighted   *
+ * and owned by IBM, These materials are provided under terms of a License     *
+ * Agreement between IBM and Sun. This technology is protected by multiple     *
+ * US and International patents. This notice and attribution to IBM may not    *
+ * to removed.                                                                 *
+ *******************************************************************************
+ */
+/* Written by Simon Montagu, Matitiahu Allouche
+ * (ported from C code written by Markus W. Scherer)
+ */
+
+package jdk.internal.icu.text;
+
+/**
+ * A BidiRun represents a sequence of characters at the same embedding level.
+ * The Bidi algorithm decomposes a piece of text into sequences of characters
+ * at the same embedding level, each such sequence is called a "run".
+ *
+ * 
A BidiRun represents such a run by storing its essential properties,
+ * but does not duplicate the characters which form the run.
+ *
+ * 
The "limit" of the run is the position just after the
+ * last character, i.e., one more than that position.
+ *
+ * 
This class has no public constructor, and its members cannot be
+ * modified by users.
+ *
+ * @see com.ibm.icu.text.Bidi
+ */
+class BidiRun {
+
+    int start;              /* first logical position of the run */
+    int limit;              /* last visual position of the run +1 */
+    int insertRemove;       /* if >0, flags for inserting LRM/RLM before/after run,
+                               if <0, count of bidi controls within run            */
+    byte level;
+
+    /*
+     * Default constructor
+     *
+     * Note that members start and limit of a run instance have different
+     * meanings depending whether the run is part of the runs array of a Bidi
+     * object, or if it is a reference returned by getVisualRun() or
+     * getLogicalRun().
+     * For a member of the runs array of a Bidi object,
+     *   - start is the first logical position of the run in the source text.
+     *   - limit is one after the last visual position of the run.
+     * For a reference returned by getLogicalRun() or getVisualRun(),
+     *   - start is the first logical position of the run in the source text.
+     *   - limit is one after the last logical position of the run.
+     */
+    BidiRun()
+    {
+        this(0, 0, (byte)0);
+    }
+
+    /*
+     * Constructor
+     */
+    BidiRun(int start, int limit, byte embeddingLevel)
+    {
+        this.start = start;
+        this.limit = limit;
+        this.level = embeddingLevel;
+    }
+
+    /*
+     * Copy the content of a BidiRun instance
+     */
+    void copyFrom(BidiRun run)
+    {
+        this.start = run.start;
+        this.limit = run.limit;
+        this.level = run.level;
+        this.insertRemove = run.insertRemove;
+    }
+
+    /**
+     * Get level of run
+     */
+    byte getEmbeddingLevel()
+    {
+        return level;
+    }
+
+    /**
+     * Check if run level is even
+     * @return true if the embedding level of this run is even, i.e. it is a
+     *  left-to-right run.
+     */
+    boolean isEvenRun()
+    {
+        return (level & 1) == 0;
+    }
+
+}
--- old/src/java.base/share/classes/sun/text/bidi/BidiWriter.java	2020-01-10 15:57:56.000000000 -0800
+++ /dev/null	2020-01-10 15:57:56.000000000 -0800
@@ -1,452 +0,0 @@
-/*
- * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.  Oracle designates this
- * particular file as subject to the "Classpath" exception as provided
- * by Oracle in the LICENSE file that accompanied this code.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- */
-
-/*
-*******************************************************************************
-*   Copyright (C) 2001-2010, International Business Machines
-*   Corporation and others.  All Rights Reserved.
-*******************************************************************************
-*/
-/* Written by Simon Montagu, Matitiahu Allouche
- * (ported from C code written by Markus W. Scherer)
- */
-
-package sun.text.bidi;
-
-import sun.text.normalizer.UCharacter;
-import sun.text.normalizer.UTF16;
-
-final class BidiWriter {
-
-    /** Bidi control code points */
-    static final char LRM_CHAR = 0x200e;
-    static final char RLM_CHAR = 0x200f;
-    static final int MASK_R_AL = (1 << UCharacter.RIGHT_TO_LEFT |
-                                  1 << UCharacter.RIGHT_TO_LEFT_ARABIC);
-
-    private static boolean IsCombining(int type) {
-        return ((1< 0);
-            break;
-
-        case BidiBase.KEEP_BASE_COMBINING:
-            /*
-             * Here, too, the destination
-             * run will have the same length as the source run,
-             * and there is no mirroring.
-             * We do need to keep combining characters with their base
-             * characters.
-             */
-            srcLength = src.length();
-
-            /* preserve character integrity */
-            do {
-                /* i is always after the last code unit known to need to be kept
-                 *  in this segment */
-                int c;
-                int i = srcLength;
-
-                /* collect code units and modifier letters for one base
-                 * character */
-                do {
-                    c = UTF16.charAt(src, srcLength - 1);
-                    srcLength -= UTF16.getCharCount(c);
-                } while(srcLength > 0 && IsCombining(UCharacter.getType(c)));
-
-                /* copy this "user character" */
-                dest.append(src.substring(srcLength, i));
-            } while(srcLength > 0);
-            break;
-
-        default:
-            /*
-             * With several "complicated" options set, this is the most
-             * general and the slowest copying of an RTL run.
-             * We will do mirroring, remove Bidi controls, and
-             * keep combining characters with their base characters
-             * as requested.
-             */
-            srcLength = src.length();
-
-            /* preserve character integrity */
-            do {
-                /* i is always after the last code unit known to need to be kept
-                 *  in this segment */
-                int i = srcLength;
-
-                /* collect code units for one base character */
-                int c = UTF16.charAt(src, srcLength - 1);
-                srcLength -= UTF16.getCharCount(c);
-                if ((options & BidiBase.KEEP_BASE_COMBINING) != 0) {
-                    /* collect modifier letters for this base character */
-                    while(srcLength > 0 && IsCombining(UCharacter.getType(c))) {
-                        c = UTF16.charAt(src, srcLength - 1);
-                        srcLength -= UTF16.getCharCount(c);
-                    }
-                }
-
-                if ((options & BidiBase.REMOVE_BIDI_CONTROLS) != 0 &&
-                    BidiBase.IsBidiControlChar(c)) {
-                    /* do not copy this Bidi control character */
-                    continue;
-                }
-
-                /* copy this "user character" */
-                int j = srcLength;
-                if((options & BidiBase.DO_MIRRORING) != 0) {
-                    /* mirror only the base character */
-                    c = UCharacter.getMirror(c);
-                    UTF16.append(dest, c);
-                    j += UTF16.getCharCount(c);
-                }
-                dest.append(src.substring(j, i));
-            } while(srcLength > 0);
-            break;
-        } /* end of switch */
-
-        return dest.toString();
-    }
-
-    static String doWriteReverse(char[] text, int start, int limit, int options) {
-        return writeReverse(new String(text, start, limit - start), options);
-    }
-
-    static String writeReordered(BidiBase bidi, int options) {
-        int run, runCount;
-        StringBuilder dest;
-        char[] text = bidi.text;
-        runCount = bidi.countRuns();
-
-        /*
-         * Option "insert marks" implies BidiBase.INSERT_LRM_FOR_NUMERIC if the
-         * reordering mode (checked below) is appropriate.
-         */
-        if ((bidi.reorderingOptions & BidiBase.OPTION_INSERT_MARKS) != 0) {
-            options |= BidiBase.INSERT_LRM_FOR_NUMERIC;
-            options &= ~BidiBase.REMOVE_BIDI_CONTROLS;
-        }
-        /*
-         * Option "remove controls" implies BidiBase.REMOVE_BIDI_CONTROLS
-         * and cancels BidiBase.INSERT_LRM_FOR_NUMERIC.
-         */
-        if ((bidi.reorderingOptions & BidiBase.OPTION_REMOVE_CONTROLS) != 0) {
-            options |= BidiBase.REMOVE_BIDI_CONTROLS;
-            options &= ~BidiBase.INSERT_LRM_FOR_NUMERIC;
-        }
-        /*
-         * If we do not perform the "inverse Bidi" algorithm, then we
-         * don't need to insert any LRMs, and don't need to test for it.
-         */
-        if ((bidi.reorderingMode != BidiBase.REORDER_INVERSE_NUMBERS_AS_L) &&
-            (bidi.reorderingMode != BidiBase.REORDER_INVERSE_LIKE_DIRECT)  &&
-            (bidi.reorderingMode != BidiBase.REORDER_INVERSE_FOR_NUMBERS_SPECIAL) &&
-            (bidi.reorderingMode != BidiBase.REORDER_RUNS_ONLY)) {
-            options &= ~BidiBase.INSERT_LRM_FOR_NUMERIC;
-        }
-        dest = new StringBuilder((options & BidiBase.INSERT_LRM_FOR_NUMERIC) != 0 ?
-                                 bidi.length * 2 : bidi.length);
-        /*
-         * Iterate through all visual runs and copy the run text segments to
-         * the destination, according to the options.
-         *
-         * The tests for where to insert LRMs ignore the fact that there may be
-         * BN codes or non-BMP code points at the beginning and end of a run;
-         * they may insert LRMs unnecessarily but the tests are faster this way
-         * (this would have to be improved for UTF-8).
-         */
-        if ((options & BidiBase.OUTPUT_REVERSE) == 0) {
-            /* forward output */
-            if ((options & BidiBase.INSERT_LRM_FOR_NUMERIC) == 0) {
-                /* do not insert Bidi controls */
-                for (run = 0; run < runCount; ++run) {
-                    BidiRun bidiRun = bidi.getVisualRun(run);
-                    if (bidiRun.isEvenRun()) {
-                        dest.append(doWriteForward(text, bidiRun.start,
-                                                   bidiRun.limit,
-                                                   options & ~BidiBase.DO_MIRRORING));
-                     } else {
-                        dest.append(doWriteReverse(text, bidiRun.start,
-                                                   bidiRun.limit, options));
-                     }
-                }
-            } else {
-                /* insert Bidi controls for "inverse Bidi" */
-                byte[] dirProps = bidi.dirProps;
-                char uc;
-                int markFlag;
-
-                for (run = 0; run < runCount; ++run) {
-                    BidiRun bidiRun = bidi.getVisualRun(run);
-                    markFlag=0;
-                    /* check if something relevant in insertPoints */
-                    markFlag = bidi.runs[run].insertRemove;
-                    if (markFlag < 0) { /* bidi controls count */
-                        markFlag = 0;
-                    }
-                    if (bidiRun.isEvenRun()) {
-                        if (bidi.isInverse() &&
-                                dirProps[bidiRun.start] != BidiBase.L) {
-                            markFlag |= BidiBase.LRM_BEFORE;
-                        }
-                        if ((markFlag & BidiBase.LRM_BEFORE) != 0) {
-                            uc = LRM_CHAR;
-                        } else if ((markFlag & BidiBase.RLM_BEFORE) != 0) {
-                            uc = RLM_CHAR;
-                        } else {
-                            uc = 0;
-                        }
-                        if (uc != 0) {
-                            dest.append(uc);
-                        }
-                        dest.append(doWriteForward(text,
-                                                   bidiRun.start, bidiRun.limit,
-                                                   options & ~BidiBase.DO_MIRRORING));
-
-                        if (bidi.isInverse() &&
-                             dirProps[bidiRun.limit - 1] != BidiBase.L) {
-                            markFlag |= BidiBase.LRM_AFTER;
-                        }
-                        if ((markFlag & BidiBase.LRM_AFTER) != 0) {
-                            uc = LRM_CHAR;
-                        } else if ((markFlag & BidiBase.RLM_AFTER) != 0) {
-                            uc = RLM_CHAR;
-                        } else {
-                            uc = 0;
-                        }
-                        if (uc != 0) {
-                            dest.append(uc);
-                        }
-                    } else { /* RTL run */
-                        if (bidi.isInverse() &&
-                            !bidi.testDirPropFlagAt(MASK_R_AL,
-                                                    bidiRun.limit - 1)) {
-                            markFlag |= BidiBase.RLM_BEFORE;
-                        }
-                        if ((markFlag & BidiBase.LRM_BEFORE) != 0) {
-                            uc = LRM_CHAR;
-                        } else if ((markFlag & BidiBase.RLM_BEFORE) != 0) {
-                            uc = RLM_CHAR;
-                        } else {
-                            uc = 0;
-                        }
-                        if (uc != 0) {
-                            dest.append(uc);
-                        }
-                        dest.append(doWriteReverse(text, bidiRun.start,
-                                                   bidiRun.limit, options));
-
-                        if(bidi.isInverse() &&
-                                (MASK_R_AL & BidiBase.DirPropFlag(dirProps[bidiRun.start])) == 0) {
-                            markFlag |= BidiBase.RLM_AFTER;
-                        }
-                        if ((markFlag & BidiBase.LRM_AFTER) != 0) {
-                            uc = LRM_CHAR;
-                        } else if ((markFlag & BidiBase.RLM_AFTER) != 0) {
-                            uc = RLM_CHAR;
-                        } else {
-                            uc = 0;
-                        }
-                        if (uc != 0) {
-                            dest.append(uc);
-                        }
-                    }
-                }
-            }
-        } else {
-            /* reverse output */
-            if((options & BidiBase.INSERT_LRM_FOR_NUMERIC) == 0) {
-                /* do not insert Bidi controls */
-                for(run = runCount; --run >= 0; ) {
-                    BidiRun bidiRun = bidi.getVisualRun(run);
-                    if (bidiRun.isEvenRun()) {
-                        dest.append(doWriteReverse(text,
-                                                   bidiRun.start, bidiRun.limit,
-                                                   options & ~BidiBase.DO_MIRRORING));
-                    } else {
-                        dest.append(doWriteForward(text, bidiRun.start,
-                                                   bidiRun.limit, options));
-                    }
-                }
-            } else {
-                /* insert Bidi controls for "inverse Bidi" */
-
-                byte[] dirProps = bidi.dirProps;
-
-                for (run = runCount; --run >= 0; ) {
-                    /* reverse output */
-                    BidiRun bidiRun = bidi.getVisualRun(run);
-                    if (bidiRun.isEvenRun()) {
-                        if (dirProps[bidiRun.limit - 1] != BidiBase.L) {
-                            dest.append(LRM_CHAR);
-                        }
-
-                        dest.append(doWriteReverse(text, bidiRun.start,
-                                bidiRun.limit, options & ~BidiBase.DO_MIRRORING));
-
-                        if (dirProps[bidiRun.start] != BidiBase.L) {
-                            dest.append(LRM_CHAR);
-                        }
-                    } else {
-                        if ((MASK_R_AL & BidiBase.DirPropFlag(dirProps[bidiRun.start])) == 0) {
-                            dest.append(RLM_CHAR);
-                        }
-
-                        dest.append(doWriteForward(text, bidiRun.start,
-                                                   bidiRun.limit, options));
-
-                        if ((MASK_R_AL & BidiBase.DirPropFlag(dirProps[bidiRun.limit - 1])) == 0) {
-                            dest.append(RLM_CHAR);
-                        }
-                    }
-                }
-            }
-        }
-
-        return dest.toString();
-    }
-}
--- /dev/null	2020-01-10 15:57:56.000000000 -0800
+++ new/src/java.base/share/classes/jdk/internal/icu/text/BidiWriter.java	2020-01-10 15:57:56.000000000 -0800
@@ -0,0 +1,451 @@
+/*
+ * Copyright (c) 2015, 2020, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.  Oracle designates this
+ * particular file as subject to the "Classpath" exception as provided
+ * by Oracle in the LICENSE file that accompanied this code.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+/*
+*******************************************************************************
+*   Copyright (C) 2001-2010, International Business Machines
+*   Corporation and others.  All Rights Reserved.
+*******************************************************************************
+*/
+/* Written by Simon Montagu, Matitiahu Allouche
+ * (ported from C code written by Markus W. Scherer)
+ */
+
+package jdk.internal.icu.text;
+
+import jdk.internal.icu.lang.UCharacter;
+
+final class BidiWriter {
+
+    /** Bidi control code points */
+    static final char LRM_CHAR = 0x200e;
+    static final char RLM_CHAR = 0x200f;
+    static final int MASK_R_AL = (1 << UCharacter.RIGHT_TO_LEFT |
+                                  1 << UCharacter.RIGHT_TO_LEFT_ARABIC);
+
+    private static boolean IsCombining(int type) {
+        return ((1< 0);
+            break;
+
+        case BidiBase.KEEP_BASE_COMBINING:
+            /*
+             * Here, too, the destination
+             * run will have the same length as the source run,
+             * and there is no mirroring.
+             * We do need to keep combining characters with their base
+             * characters.
+             */
+            srcLength = src.length();
+
+            /* preserve character integrity */
+            do {
+                /* i is always after the last code unit known to need to be kept
+                 *  in this segment */
+                int c;
+                int i = srcLength;
+
+                /* collect code units and modifier letters for one base
+                 * character */
+                do {
+                    c = UTF16.charAt(src, srcLength - 1);
+                    srcLength -= UTF16.getCharCount(c);
+                } while(srcLength > 0 && IsCombining(UCharacter.getType(c)));
+
+                /* copy this "user character" */
+                dest.append(src.substring(srcLength, i));
+            } while(srcLength > 0);
+            break;
+
+        default:
+            /*
+             * With several "complicated" options set, this is the most
+             * general and the slowest copying of an RTL run.
+             * We will do mirroring, remove Bidi controls, and
+             * keep combining characters with their base characters
+             * as requested.
+             */
+            srcLength = src.length();
+
+            /* preserve character integrity */
+            do {
+                /* i is always after the last code unit known to need to be kept
+                 *  in this segment */
+                int i = srcLength;
+
+                /* collect code units for one base character */
+                int c = UTF16.charAt(src, srcLength - 1);
+                srcLength -= UTF16.getCharCount(c);
+                if ((options & BidiBase.KEEP_BASE_COMBINING) != 0) {
+                    /* collect modifier letters for this base character */
+                    while(srcLength > 0 && IsCombining(UCharacter.getType(c))) {
+                        c = UTF16.charAt(src, srcLength - 1);
+                        srcLength -= UTF16.getCharCount(c);
+                    }
+                }
+
+                if ((options & BidiBase.REMOVE_BIDI_CONTROLS) != 0 &&
+                    BidiBase.IsBidiControlChar(c)) {
+                    /* do not copy this Bidi control character */
+                    continue;
+                }
+
+                /* copy this "user character" */
+                int j = srcLength;
+                if((options & BidiBase.DO_MIRRORING) != 0) {
+                    /* mirror only the base character */
+                    c = UCharacter.getMirror(c);
+                    UTF16.append(dest, c);
+                    j += UTF16.getCharCount(c);
+                }
+                dest.append(src.substring(j, i));
+            } while(srcLength > 0);
+            break;
+        } /* end of switch */
+
+        return dest.toString();
+    }
+
+    static String doWriteReverse(char[] text, int start, int limit, int options) {
+        return writeReverse(new String(text, start, limit - start), options);
+    }
+
+    static String writeReordered(BidiBase bidi, int options) {
+        int run, runCount;
+        StringBuilder dest;
+        char[] text = bidi.text;
+        runCount = bidi.countRuns();
+
+        /*
+         * Option "insert marks" implies BidiBase.INSERT_LRM_FOR_NUMERIC if the
+         * reordering mode (checked below) is appropriate.
+         */
+        if ((bidi.reorderingOptions & BidiBase.OPTION_INSERT_MARKS) != 0) {
+            options |= BidiBase.INSERT_LRM_FOR_NUMERIC;
+            options &= ~BidiBase.REMOVE_BIDI_CONTROLS;
+        }
+        /*
+         * Option "remove controls" implies BidiBase.REMOVE_BIDI_CONTROLS
+         * and cancels BidiBase.INSERT_LRM_FOR_NUMERIC.
+         */
+        if ((bidi.reorderingOptions & BidiBase.OPTION_REMOVE_CONTROLS) != 0) {
+            options |= BidiBase.REMOVE_BIDI_CONTROLS;
+            options &= ~BidiBase.INSERT_LRM_FOR_NUMERIC;
+        }
+        /*
+         * If we do not perform the "inverse Bidi" algorithm, then we
+         * don't need to insert any LRMs, and don't need to test for it.
+         */
+        if ((bidi.reorderingMode != BidiBase.REORDER_INVERSE_NUMBERS_AS_L) &&
+            (bidi.reorderingMode != BidiBase.REORDER_INVERSE_LIKE_DIRECT)  &&
+            (bidi.reorderingMode != BidiBase.REORDER_INVERSE_FOR_NUMBERS_SPECIAL) &&
+            (bidi.reorderingMode != BidiBase.REORDER_RUNS_ONLY)) {
+            options &= ~BidiBase.INSERT_LRM_FOR_NUMERIC;
+        }
+        dest = new StringBuilder((options & BidiBase.INSERT_LRM_FOR_NUMERIC) != 0 ?
+                                 bidi.length * 2 : bidi.length);
+        /*
+         * Iterate through all visual runs and copy the run text segments to
+         * the destination, according to the options.
+         *
+         * The tests for where to insert LRMs ignore the fact that there may be
+         * BN codes or non-BMP code points at the beginning and end of a run;
+         * they may insert LRMs unnecessarily but the tests are faster this way
+         * (this would have to be improved for UTF-8).
+         */
+        if ((options & BidiBase.OUTPUT_REVERSE) == 0) {
+            /* forward output */
+            if ((options & BidiBase.INSERT_LRM_FOR_NUMERIC) == 0) {
+                /* do not insert Bidi controls */
+                for (run = 0; run < runCount; ++run) {
+                    BidiRun bidiRun = bidi.getVisualRun(run);
+                    if (bidiRun.isEvenRun()) {
+                        dest.append(doWriteForward(text, bidiRun.start,
+                                                   bidiRun.limit,
+                                                   options & ~BidiBase.DO_MIRRORING));
+                     } else {
+                        dest.append(doWriteReverse(text, bidiRun.start,
+                                                   bidiRun.limit, options));
+                     }
+                }
+            } else {
+                /* insert Bidi controls for "inverse Bidi" */
+                byte[] dirProps = bidi.dirProps;
+                char uc;
+                int markFlag;
+
+                for (run = 0; run < runCount; ++run) {
+                    BidiRun bidiRun = bidi.getVisualRun(run);
+                    markFlag=0;
+                    /* check if something relevant in insertPoints */
+                    markFlag = bidi.runs[run].insertRemove;
+                    if (markFlag < 0) { /* bidi controls count */
+                        markFlag = 0;
+                    }
+                    if (bidiRun.isEvenRun()) {
+                        if (bidi.isInverse() &&
+                                dirProps[bidiRun.start] != BidiBase.L) {
+                            markFlag |= BidiBase.LRM_BEFORE;
+                        }
+                        if ((markFlag & BidiBase.LRM_BEFORE) != 0) {
+                            uc = LRM_CHAR;
+                        } else if ((markFlag & BidiBase.RLM_BEFORE) != 0) {
+                            uc = RLM_CHAR;
+                        } else {
+                            uc = 0;
+                        }
+                        if (uc != 0) {
+                            dest.append(uc);
+                        }
+                        dest.append(doWriteForward(text,
+                                                   bidiRun.start, bidiRun.limit,
+                                                   options & ~BidiBase.DO_MIRRORING));
+
+                        if (bidi.isInverse() &&
+                             dirProps[bidiRun.limit - 1] != BidiBase.L) {
+                            markFlag |= BidiBase.LRM_AFTER;
+                        }
+                        if ((markFlag & BidiBase.LRM_AFTER) != 0) {
+                            uc = LRM_CHAR;
+                        } else if ((markFlag & BidiBase.RLM_AFTER) != 0) {
+                            uc = RLM_CHAR;
+                        } else {
+                            uc = 0;
+                        }
+                        if (uc != 0) {
+                            dest.append(uc);
+                        }
+                    } else { /* RTL run */
+                        if (bidi.isInverse() &&
+                            !bidi.testDirPropFlagAt(MASK_R_AL,
+                                                    bidiRun.limit - 1)) {
+                            markFlag |= BidiBase.RLM_BEFORE;
+                        }
+                        if ((markFlag & BidiBase.LRM_BEFORE) != 0) {
+                            uc = LRM_CHAR;
+                        } else if ((markFlag & BidiBase.RLM_BEFORE) != 0) {
+                            uc = RLM_CHAR;
+                        } else {
+                            uc = 0;
+                        }
+                        if (uc != 0) {
+                            dest.append(uc);
+                        }
+                        dest.append(doWriteReverse(text, bidiRun.start,
+                                                   bidiRun.limit, options));
+
+                        if(bidi.isInverse() &&
+                                (MASK_R_AL & BidiBase.DirPropFlag(dirProps[bidiRun.start])) == 0) {
+                            markFlag |= BidiBase.RLM_AFTER;
+                        }
+                        if ((markFlag & BidiBase.LRM_AFTER) != 0) {
+                            uc = LRM_CHAR;
+                        } else if ((markFlag & BidiBase.RLM_AFTER) != 0) {
+                            uc = RLM_CHAR;
+                        } else {
+                            uc = 0;
+                        }
+                        if (uc != 0) {
+                            dest.append(uc);
+                        }
+                    }
+                }
+            }
+        } else {
+            /* reverse output */
+            if((options & BidiBase.INSERT_LRM_FOR_NUMERIC) == 0) {
+                /* do not insert Bidi controls */
+                for(run = runCount; --run >= 0; ) {
+                    BidiRun bidiRun = bidi.getVisualRun(run);
+                    if (bidiRun.isEvenRun()) {
+                        dest.append(doWriteReverse(text,
+                                                   bidiRun.start, bidiRun.limit,
+                                                   options & ~BidiBase.DO_MIRRORING));
+                    } else {
+                        dest.append(doWriteForward(text, bidiRun.start,
+                                                   bidiRun.limit, options));
+                    }
+                }
+            } else {
+                /* insert Bidi controls for "inverse Bidi" */
+
+                byte[] dirProps = bidi.dirProps;
+
+                for (run = runCount; --run >= 0; ) {
+                    /* reverse output */
+                    BidiRun bidiRun = bidi.getVisualRun(run);
+                    if (bidiRun.isEvenRun()) {
+                        if (dirProps[bidiRun.limit - 1] != BidiBase.L) {
+                            dest.append(LRM_CHAR);
+                        }
+
+                        dest.append(doWriteReverse(text, bidiRun.start,
+                                bidiRun.limit, options & ~BidiBase.DO_MIRRORING));
+
+                        if (dirProps[bidiRun.start] != BidiBase.L) {
+                            dest.append(LRM_CHAR);
+                        }
+                    } else {
+                        if ((MASK_R_AL & BidiBase.DirPropFlag(dirProps[bidiRun.start])) == 0) {
+                            dest.append(RLM_CHAR);
+                        }
+
+                        dest.append(doWriteForward(text, bidiRun.start,
+                                                   bidiRun.limit, options));
+
+                        if ((MASK_R_AL & BidiBase.DirPropFlag(dirProps[bidiRun.limit - 1])) == 0) {
+                            dest.append(RLM_CHAR);
+                        }
+                    }
+                }
+            }
+        }
+
+        return dest.toString();
+    }
+}
--- old/src/java.base/share/classes/sun/text/normalizer/FilteredNormalizer2.java	2020-01-10 15:57:58.000000000 -0800
+++ /dev/null	2020-01-10 15:57:58.000000000 -0800
@@ -1,266 +0,0 @@
-/*
- * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.  Oracle designates this
- * particular file as subject to the "Classpath" exception as provided
- * by Oracle in the LICENSE file that accompanied this code.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- */
-
-/*
-*******************************************************************************
-*   Copyright (C) 2009-2014, International Business Machines
-*   Corporation and others.  All Rights Reserved.
-*******************************************************************************
-*/
-package sun.text.normalizer;
-
-import java.io.IOException;
-
-/**
- * Normalization filtered by a UnicodeSet.
- * Normalizes portions of the text contained in the filter set and leaves
- * portions not contained in the filter set unchanged.
- * Filtering is done via UnicodeSet.span(..., UnicodeSet.SpanCondition.SIMPLE).
- * Not-in-the-filter text is treated as "is normalized" and "quick check yes".
- * This class implements all of (and only) the Normalizer2 API.
- * An instance of this class is unmodifiable/immutable.
- * @stable ICU 4.4
- * @author Markus W. Scherer
- */
-class FilteredNormalizer2 extends Normalizer2 {
-
-    /**
-     * Constructs a filtered normalizer wrapping any Normalizer2 instance
-     * and a filter set.
-     * Both are aliased and must not be modified or deleted while this object
-     * is used.
-     * The filter set should be frozen; otherwise the performance will suffer greatly.
-     * @param n2 wrapped Normalizer2 instance
-     * @param filterSet UnicodeSet which determines the characters to be normalized
-     * @stable ICU 4.4
-     */
-    public FilteredNormalizer2(Normalizer2 n2, UnicodeSet filterSet) {
-        norm2=n2;
-        set=filterSet;
-    }
-
-    /**
-     * {@inheritDoc}
-     * @stable ICU 4.4
-     */
-    @Override
-    public StringBuilder normalize(CharSequence src, StringBuilder dest) {
-        if(dest==src) {
-            throw new IllegalArgumentException();
-        }
-        dest.setLength(0);
-        normalize(src, dest, UnicodeSet.SpanCondition.SIMPLE);
-        return dest;
-    }
-
-    /**
-     * {@inheritDoc}
-     * @stable ICU 4.6
-     */
-    @Override
-    public Appendable normalize(CharSequence src, Appendable dest) {
-        if(dest==src) {
-            throw new IllegalArgumentException();
-        }
-        return normalize(src, dest, UnicodeSet.SpanCondition.SIMPLE);
-    }
-
-    /**
-     * {@inheritDoc}
-     * @stable ICU 4.4
-     */
-    @Override
-    public StringBuilder normalizeSecondAndAppend(
-            StringBuilder first, CharSequence second) {
-        return normalizeSecondAndAppend(first, second, true);
-    }
-
-    /**
-     * {@inheritDoc}
-     * @stable ICU 4.4
-     */
-    @Override
-    public StringBuilder append(StringBuilder first, CharSequence second) {
-        return normalizeSecondAndAppend(first, second, false);
-    }
-
-    /**
-     * {@inheritDoc}
-     * @stable ICU 4.6
-     */
-    @Override
-    public String getDecomposition(int c) {
-        return set.contains(c) ? norm2.getDecomposition(c) : null;
-    }
-
-    /**
-     * {@inheritDoc}
-     * @stable ICU 49
-     */
-    @Override
-    public int getCombiningClass(int c) {
-        return set.contains(c) ? norm2.getCombiningClass(c) : 0;
-    }
-
-    /**
-     * {@inheritDoc}
-     * @stable ICU 4.4
-     */
-    @Override
-    public boolean isNormalized(CharSequence s) {
-        UnicodeSet.SpanCondition spanCondition=UnicodeSet.SpanCondition.SIMPLE;
-        for(int prevSpanLimit=0; prevSpanLimit
- * The primary functions are to produce a normalized string and to detect whether
- * a string is already normalized.
- * The most commonly used normalization forms are those defined in
- * http://www.unicode.org/unicode/reports/tr15/
- * However, this API supports additional normalization forms for specialized purposes.
- * For example, NFKC_Casefold is provided via getInstance("nfkc_cf", COMPOSE)
- * and can be used in implementations of UTS #46.
- * 

- * Not only are the standard compose and decompose modes supplied,
- * but additional modes are provided as documented in the Mode enum.
- * 

- * Some of the functions in this class identify normalization boundaries.
- * At a normalization boundary, the portions of the string
- * before it and starting from it do not interact and can be handled independently.
- * 

- * The spanQuickCheckYes() stops at a normalization boundary.
- * When the goal is a normalized string, then the text before the boundary
- * can be copied, and the remainder can be processed with normalizeSecondAndAppend().
- * 

- * The hasBoundaryBefore(), hasBoundaryAfter() and isInert() functions test whether
- * a character is guaranteed to be at a normalization boundary,
- * regardless of context.
- * This is used for moving from one normalization boundary to the next
- * or preceding boundary, and for performing iterative normalization.
- * 

- * Iterative normalization is useful when only a small portion of a
- * longer string needs to be processed.
- * For example, in ICU, iterative normalization is used by the NormalizationTransliterator
- * (to avoid replacing already-normalized text) and ucol_nextSortKeyPart()
- * (to process only the substring for which sort key bytes are computed).
- * 

- * The set of normalization boundaries returned by these functions may not be
- * complete: There may be more boundaries that could be returned.
- * Different functions may return different boundaries.
- * @stable ICU 4.4
- * @author Markus W. Scherer
- */
-abstract class Normalizer2 {
-
-    /**
-     * Returns a Normalizer2 instance for Unicode NFC normalization.
-     * Same as getInstance(null, "nfc", Mode.COMPOSE).
-     * Returns an unmodifiable singleton instance.
-     * @return the requested Normalizer2, if successful
-     * @stable ICU 49
-     */
-    public static Normalizer2 getNFCInstance() {
-        return Norm2AllModes.getNFCInstance().comp;
-    }
-
-    /**
-     * Returns a Normalizer2 instance for Unicode NFD normalization.
-     * Same as getInstance(null, "nfc", Mode.DECOMPOSE).
-     * Returns an unmodifiable singleton instance.
-     * @return the requested Normalizer2, if successful
-     * @stable ICU 49
-     */
-    public static Normalizer2 getNFDInstance() {
-        return Norm2AllModes.getNFCInstance().decomp;
-    }
-
-    /**
-     * Returns a Normalizer2 instance for Unicode NFKC normalization.
-     * Same as getInstance(null, "nfkc", Mode.COMPOSE).
-     * Returns an unmodifiable singleton instance.
-     * @return the requested Normalizer2, if successful
-     * @stable ICU 49
-     */
-    public static Normalizer2 getNFKCInstance() {
-        return Norm2AllModes.getNFKCInstance().comp;
-    }
-
-    /**
-     * Returns a Normalizer2 instance for Unicode NFKD normalization.
-     * Same as getInstance(null, "nfkc", Mode.DECOMPOSE).
-     * Returns an unmodifiable singleton instance.
-     * @return the requested Normalizer2, if successful
-     * @stable ICU 49
-     */
-    public static Normalizer2 getNFKDInstance() {
-        return Norm2AllModes.getNFKCInstance().decomp;
-    }
-
-    /**
-     * Returns the normalized form of the source string.
-     * @param src source string
-     * @return normalized src
-     * @stable ICU 4.4
-     */
-    public String normalize(CharSequence src) {
-        if(src instanceof String) {
-            // Fastpath: Do not construct a new String if the src is a String
-            // and is already normalized.
-            int spanLength=spanQuickCheckYes(src);
-            if(spanLength==src.length()) {
-                return (String)src;
-            }
-            if (spanLength != 0) {
-                StringBuilder sb=new StringBuilder(src.length()).append(src, 0, spanLength);
-                return normalizeSecondAndAppend(sb, src.subSequence(spanLength, src.length())).toString();
-            }
-        }
-        return normalize(src, new StringBuilder(src.length())).toString();
-    }
-
-    /**
-     * Writes the normalized form of the source string to the destination string
-     * (replacing its contents) and returns the destination string.
-     * The source and destination strings must be different objects.
-     * @param src source string
-     * @param dest destination string; its contents is replaced with normalized src
-     * @return dest
-     * @stable ICU 4.4
-     */
-    public abstract StringBuilder normalize(CharSequence src, StringBuilder dest);
-
-    /**
-     * Writes the normalized form of the source string to the destination Appendable
-     * and returns the destination Appendable.
-     * The source and destination strings must be different objects.
-     *
-     * 
Any {@link java.io.IOException} is wrapped into a {@link com.ibm.icu.util.ICUUncheckedIOException}.
-     *
-     * @param src source string
-     * @param dest destination Appendable; gets normalized src appended
-     * @return dest
-     * @stable ICU 4.6
-     */
-    public abstract Appendable normalize(CharSequence src, Appendable dest);
-
-    /**
-     * Appends the normalized form of the second string to the first string
-     * (merging them at the boundary) and returns the first string.
-     * The result is normalized if the first string was normalized.
-     * The first and second strings must be different objects.
-     * @param first string, should be normalized
-     * @param second string, will be normalized
-     * @return first
-     * @stable ICU 4.4
-     */
-    public abstract StringBuilder normalizeSecondAndAppend(
-            StringBuilder first, CharSequence second);
-
-    /**
-     * Appends the second string to the first string
-     * (merging them at the boundary) and returns the first string.
-     * The result is normalized if both the strings were normalized.
-     * The first and second strings must be different objects.
-     * @param first string, should be normalized
-     * @param second string, should be normalized
-     * @return first
-     * @stable ICU 4.4
-     */
-    public abstract StringBuilder append(StringBuilder first, CharSequence second);
-
-    /**
-     * Gets the decomposition mapping of c.
-     * Roughly equivalent to normalizing the String form of c
-     * on a DECOMPOSE Normalizer2 instance, but much faster, and except that this function
-     * returns null if c does not have a decomposition mapping in this instance's data.
-     * This function is independent of the mode of the Normalizer2.
-     * @param c code point
-     * @return c's decomposition mapping, if any; otherwise null
-     * @stable ICU 4.6
-     */
-    public abstract String getDecomposition(int c);
-
-    /**
-     * Gets the combining class of c.
-     * The default implementation returns 0
-     * but all standard implementations return the Unicode Canonical_Combining_Class value.
-     * @param c code point
-     * @return c's combining class
-     * @stable ICU 49
-     */
-    public int getCombiningClass(int c) { return 0; }
-
-    /**
-     * Tests if the string is normalized.
-     * Internally, in cases where the quickCheck() method would return "maybe"
-     * (which is only possible for the two COMPOSE modes) this method
-     * resolves to "yes" or "no" to provide a definitive result,
-     * at the cost of doing more work in those cases.
-     * @param s input string
-     * @return true if s is normalized
-     * @stable ICU 4.4
-     */
-    public abstract boolean isNormalized(CharSequence s);
-
-    /**
-     * Returns the end of the normalized substring of the input string.
-     * In other words, with end=spanQuickCheckYes(s);
-     * the substring s.subSequence(0, end)
-     * will pass the quick check with a "yes" result.
-     * 

-     * The returned end index is usually one or more characters before the
-     * "no" or "maybe" character: The end index is at a normalization boundary.
-     * (See the class documentation for more about normalization boundaries.)
-     * 

-     * When the goal is a normalized string and most input strings are expected
-     * to be normalized already, then call this method,
-     * and if it returns a prefix shorter than the input string,
-     * copy that prefix and use normalizeSecondAndAppend() for the remainder.
-     * @param s input string
-     * @return "yes" span end index
-     * @stable ICU 4.4
-     */
-    public abstract int spanQuickCheckYes(CharSequence s);
-
-    /**
-     * Tests if the character always has a normalization boundary before it,
-     * regardless of context.
-     * If true, then the character does not normalization-interact with
-     * preceding characters.
-     * In other words, a string containing this character can be normalized
-     * by processing portions before this character and starting from this
-     * character independently.
-     * This is used for iterative normalization. See the class documentation for details.
-     * @param c character to test
-     * @return true if c has a normalization boundary before it
-     * @stable ICU 4.4
-     */
-    public abstract boolean hasBoundaryBefore(int c);
-
-    /**
-     * Sole constructor.  (For invocation by subclass constructors,
-     * typically implicit.)
-     * @internal
-     * deprecated This API is ICU internal only.
-     */
-    protected Normalizer2() {
-    }
-}
--- /dev/null	2020-01-10 15:57:59.000000000 -0800
+++ new/src/java.base/share/classes/jdk/internal/icu/text/Normalizer2.java	2020-01-10 15:57:59.000000000 -0800
@@ -0,0 +1,275 @@
+/*
+ * Copyright (c) 2015, 2020, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.  Oracle designates this
+ * particular file as subject to the "Classpath" exception as provided
+ * by Oracle in the LICENSE file that accompanied this code.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+/*
+ *******************************************************************************
+ *   Copyright (C) 2009-2014, International Business Machines
+ *   Corporation and others.  All Rights Reserved.
+ *******************************************************************************
+ */
+
+package jdk.internal.icu.text;
+
+import jdk.internal.icu.impl.Norm2AllModes;
+
+/**
+ * Unicode normalization functionality for standard Unicode normalization or
+ * for using custom mapping tables.
+ * All instances of this class are unmodifiable/immutable.
+ * The Normalizer2 class is not intended for public subclassing.
+ * 

+ * The primary functions are to produce a normalized string and to detect whether
+ * a string is already normalized.
+ * The most commonly used normalization forms are those defined in
+ * http://www.unicode.org/unicode/reports/tr15/
+ * However, this API supports additional normalization forms for specialized purposes.
+ * For example, NFKC_Casefold is provided via getInstance("nfkc_cf", COMPOSE)
+ * and can be used in implementations of UTS #46.
+ * 

+ * Not only are the standard compose and decompose modes supplied,
+ * but additional modes are provided as documented in the Mode enum.
+ * 

+ * Some of the functions in this class identify normalization boundaries.
+ * At a normalization boundary, the portions of the string
+ * before it and starting from it do not interact and can be handled independently.
+ * 

+ * The spanQuickCheckYes() stops at a normalization boundary.
+ * When the goal is a normalized string, then the text before the boundary
+ * can be copied, and the remainder can be processed with normalizeSecondAndAppend().
+ * 

+ * The hasBoundaryBefore(), hasBoundaryAfter() and isInert() functions test whether
+ * a character is guaranteed to be at a normalization boundary,
+ * regardless of context.
+ * This is used for moving from one normalization boundary to the next
+ * or preceding boundary, and for performing iterative normalization.
+ * 

+ * Iterative normalization is useful when only a small portion of a
+ * longer string needs to be processed.
+ * For example, in ICU, iterative normalization is used by the NormalizationTransliterator
+ * (to avoid replacing already-normalized text) and ucol_nextSortKeyPart()
+ * (to process only the substring for which sort key bytes are computed).
+ * 

+ * The set of normalization boundaries returned by these functions may not be
+ * complete: There may be more boundaries that could be returned.
+ * Different functions may return different boundaries.
+ * @stable ICU 4.4
+ * @author Markus W. Scherer
+ */
+public abstract class Normalizer2 {
+
+    /**
+     * Returns a Normalizer2 instance for Unicode NFC normalization.
+     * Same as getInstance(null, "nfc", Mode.COMPOSE).
+     * Returns an unmodifiable singleton instance.
+     * @return the requested Normalizer2, if successful
+     * @stable ICU 49
+     */
+    public static Normalizer2 getNFCInstance() {
+        return Norm2AllModes.getNFCInstance().comp;
+    }
+
+    /**
+     * Returns a Normalizer2 instance for Unicode NFD normalization.
+     * Same as getInstance(null, "nfc", Mode.DECOMPOSE).
+     * Returns an unmodifiable singleton instance.
+     * @return the requested Normalizer2, if successful
+     * @stable ICU 49
+     */
+    public static Normalizer2 getNFDInstance() {
+        return Norm2AllModes.getNFCInstance().decomp;
+    }
+
+    /**
+     * Returns a Normalizer2 instance for Unicode NFKC normalization.
+     * Same as getInstance(null, "nfkc", Mode.COMPOSE).
+     * Returns an unmodifiable singleton instance.
+     * @return the requested Normalizer2, if successful
+     * @stable ICU 49
+     */
+    public static Normalizer2 getNFKCInstance() {
+        return Norm2AllModes.getNFKCInstance().comp;
+    }
+
+    /**
+     * Returns a Normalizer2 instance for Unicode NFKD normalization.
+     * Same as getInstance(null, "nfkc", Mode.DECOMPOSE).
+     * Returns an unmodifiable singleton instance.
+     * @return the requested Normalizer2, if successful
+     * @stable ICU 49
+     */
+    public static Normalizer2 getNFKDInstance() {
+        return Norm2AllModes.getNFKCInstance().decomp;
+    }
+
+    /**
+     * Returns the normalized form of the source string.
+     * @param src source string
+     * @return normalized src
+     * @stable ICU 4.4
+     */
+    public String normalize(CharSequence src) {
+        if(src instanceof String) {
+            // Fastpath: Do not construct a new String if the src is a String
+            // and is already normalized.
+            int spanLength=spanQuickCheckYes(src);
+            if(spanLength==src.length()) {
+                return (String)src;
+            }
+            if (spanLength != 0) {
+                StringBuilder sb=new StringBuilder(src.length()).append(src, 0, spanLength);
+                return normalizeSecondAndAppend(sb, src.subSequence(spanLength, src.length())).toString();
+            }
+        }
+        return normalize(src, new StringBuilder(src.length())).toString();
+    }
+
+    /**
+     * Writes the normalized form of the source string to the destination string
+     * (replacing its contents) and returns the destination string.
+     * The source and destination strings must be different objects.
+     * @param src source string
+     * @param dest destination string; its contents is replaced with normalized src
+     * @return dest
+     * @stable ICU 4.4
+     */
+    public abstract StringBuilder normalize(CharSequence src, StringBuilder dest);
+
+    /**
+     * Writes the normalized form of the source string to the destination Appendable
+     * and returns the destination Appendable.
+     * The source and destination strings must be different objects.
+     *
+     * 
Any {@link java.io.IOException} is wrapped into a {@link com.ibm.icu.util.ICUUncheckedIOException}.
+     *
+     * @param src source string
+     * @param dest destination Appendable; gets normalized src appended
+     * @return dest
+     * @stable ICU 4.6
+     */
+    public abstract Appendable normalize(CharSequence src, Appendable dest);
+
+    /**
+     * Appends the normalized form of the second string to the first string
+     * (merging them at the boundary) and returns the first string.
+     * The result is normalized if the first string was normalized.
+     * The first and second strings must be different objects.
+     * @param first string, should be normalized
+     * @param second string, will be normalized
+     * @return first
+     * @stable ICU 4.4
+     */
+    public abstract StringBuilder normalizeSecondAndAppend(
+            StringBuilder first, CharSequence second);
+
+    /**
+     * Appends the second string to the first string
+     * (merging them at the boundary) and returns the first string.
+     * The result is normalized if both the strings were normalized.
+     * The first and second strings must be different objects.
+     * @param first string, should be normalized
+     * @param second string, should be normalized
+     * @return first
+     * @stable ICU 4.4
+     */
+    public abstract StringBuilder append(StringBuilder first, CharSequence second);
+
+    /**
+     * Gets the decomposition mapping of c.
+     * Roughly equivalent to normalizing the String form of c
+     * on a DECOMPOSE Normalizer2 instance, but much faster, and except that this function
+     * returns null if c does not have a decomposition mapping in this instance's data.
+     * This function is independent of the mode of the Normalizer2.
+     * @param c code point
+     * @return c's decomposition mapping, if any; otherwise null
+     * @stable ICU 4.6
+     */
+    public abstract String getDecomposition(int c);
+
+    /**
+     * Gets the combining class of c.
+     * The default implementation returns 0
+     * but all standard implementations return the Unicode Canonical_Combining_Class value.
+     * @param c code point
+     * @return c's combining class
+     * @stable ICU 49
+     */
+    public int getCombiningClass(int c) { return 0; }
+
+    /**
+     * Tests if the string is normalized.
+     * Internally, in cases where the quickCheck() method would return "maybe"
+     * (which is only possible for the two COMPOSE modes) this method
+     * resolves to "yes" or "no" to provide a definitive result,
+     * at the cost of doing more work in those cases.
+     * @param s input string
+     * @return true if s is normalized
+     * @stable ICU 4.4
+     */
+    public abstract boolean isNormalized(CharSequence s);
+
+    /**
+     * Returns the end of the normalized substring of the input string.
+     * In other words, with end=spanQuickCheckYes(s);
+     * the substring s.subSequence(0, end)
+     * will pass the quick check with a "yes" result.
+     * 

+     * The returned end index is usually one or more characters before the
+     * "no" or "maybe" character: The end index is at a normalization boundary.
+     * (See the class documentation for more about normalization boundaries.)
+     * 

+     * When the goal is a normalized string and most input strings are expected
+     * to be normalized already, then call this method,
+     * and if it returns a prefix shorter than the input string,
+     * copy that prefix and use normalizeSecondAndAppend() for the remainder.
+     * @param s input string
+     * @return "yes" span end index
+     * @stable ICU 4.4
+     */
+    public abstract int spanQuickCheckYes(CharSequence s);
+
+    /**
+     * Tests if the character always has a normalization boundary before it,
+     * regardless of context.
+     * If true, then the character does not normalization-interact with
+     * preceding characters.
+     * In other words, a string containing this character can be normalized
+     * by processing portions before this character and starting from this
+     * character independently.
+     * This is used for iterative normalization. See the class documentation for details.
+     * @param c character to test
+     * @return true if c has a normalization boundary before it
+     * @stable ICU 4.4
+     */
+    public abstract boolean hasBoundaryBefore(int c);
+
+    /**
+     * Sole constructor.  (For invocation by subclass constructors,
+     * typically implicit.)
+     * @internal
+     * deprecated This API is ICU internal only.
+     */
+    protected Normalizer2() {
+    }
+}
--- old/src/java.base/share/classes/sun/text/normalizer/NormalizerBase.java	2020-01-10 15:58:00.000000000 -0800
+++ /dev/null	2020-01-10 15:58:00.000000000 -0800
@@ -1,782 +0,0 @@
-/*
- * Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved.
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.  Oracle designates this
- * particular file as subject to the "Classpath" exception as provided
- * by Oracle in the LICENSE file that accompanied this code.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- */
-
-/*
- *******************************************************************************
- * Copyright (C) 2000-2014, International Business Machines Corporation and
- * others. All Rights Reserved.
- *******************************************************************************
- */
-package sun.text.normalizer;
-
-import java.text.CharacterIterator;
-import java.text.Normalizer;
-
-/**
- * Unicode Normalization
- *
- * 
Unicode normalization API

- *
- * normalize transforms Unicode text into an equivalent composed or
- * decomposed form, allowing for easier sorting and searching of text.
- * normalize supports the standard normalization forms described in
- * 
- * Unicode Standard Annex #15 — Unicode Normalization Forms.
- *
- * Characters with accents or other adornments can be encoded in
- * several different ways in Unicode.  For example, take the character A-acute.
- * In Unicode, this can be encoded as a single character (the
- * "composed" form):
- *
- * 
- *      00C1    LATIN CAPITAL LETTER A WITH ACUTE
- * 

- *
- * or as two separate characters (the "decomposed" form):
- *
- * 
- *      0041    LATIN CAPITAL LETTER A
- *      0301    COMBINING ACUTE ACCENT
- * 

- *
- * To a user of your program, however, both of these sequences should be
- * treated as the same "user-level" character "A with acute accent".  When you
- * are searching or comparing text, you must ensure that these two sequences are
- * treated equivalently.  In addition, you must handle characters with more than
- * one accent.  Sometimes the order of a character's combining accents is
- * significant, while in other cases accent sequences in different orders are
- * really equivalent.
- *
- * Similarly, the string "ffi" can be encoded as three separate letters:
- *
- * 
- *      0066    LATIN SMALL LETTER F
- *      0066    LATIN SMALL LETTER F
- *      0069    LATIN SMALL LETTER I
- * 

- *
- * or as the single character
- *
- * 
- *      FB03    LATIN SMALL LIGATURE FFI
- * 

- *
- * The ffi ligature is not a distinct semantic character, and strictly speaking
- * it shouldn't be in Unicode at all, but it was included for compatibility
- * with existing character sets that already provided it.  The Unicode standard
- * identifies such characters by giving them "compatibility" decompositions
- * into the corresponding semantic characters.  When sorting and searching, you
- * will often want to use these mappings.
- *
- * normalize helps solve these problems by transforming text into
- * the canonical composed and decomposed forms as shown in the first example
- * above. In addition, you can have it perform compatibility decompositions so
- * that you can treat compatibility characters the same as their equivalents.
- * Finally, normalize rearranges accents into the proper canonical
- * order, so that you do not have to worry about accent rearrangement on your
- * own.
- *
- * Form FCD, "Fast C or D", is also designed for collation.
- * It allows to work on strings that are not necessarily normalized
- * with an algorithm (like in collation) that works under "canonical closure",
- * i.e., it treats precomposed characters and their decomposed equivalents the
- * same.
- *
- * It is not a normalization form because it does not provide for uniqueness of
- * representation. Multiple strings may be canonically equivalent (their NFDs
- * are identical) and may all conform to FCD without being identical themselves.
- *
- * The form is defined such that the "raw decomposition", the recursive
- * canonical decomposition of each character, results in a string that is
- * canonically ordered. This means that precomposed characters are allowed for
- * as long as their decompositions do not need canonical reordering.
- *
- * Its advantage for a process like collation is that all NFD and most NFC texts
- * - and many unnormalized texts - already conform to FCD and do not need to be
- * normalized (NFD) for such a process. The FCD quick check will return YES for
- * most strings in practice.
- *
- * normalize(FCD) may be implemented with NFD.
- *
- * For more details on FCD see Unicode Technical Note #5 (Canonical Equivalence in Applications):
- * http://www.unicode.org/notes/tn5/#FCD
- *
- * ICU collation performs either NFD or FCD normalization automatically if
- * normalization is turned on for the collator object. Beyond collation and
- * string search, normalized strings may be useful for string equivalence
- * comparisons, transliteration/transcription, unique representations, etc.
- *
- * The W3C generally recommends to exchange texts in NFC.
- * Note also that most legacy character encodings use only precomposed forms and
- * often do not encode any combining marks by themselves. For conversion to such
- * character encodings the Unicode text needs to be normalized to NFC.
- * For more usage examples, see the Unicode Standard Annex.
- *
- * Note: The Normalizer class also provides API for iterative normalization.
- * While the setIndex() and getIndex() refer to indices in the
- * underlying Unicode input text, the next() and previous() methods
- * iterate through characters in the normalized output.
- * This means that there is not necessarily a one-to-one correspondence
- * between characters returned by next() and previous() and the indices
- * passed to and returned from setIndex() and getIndex().
- * It is for this reason that Normalizer does not implement the CharacterIterator interface.
- *
- * @stable ICU 2.8
- */
-// Original filename in ICU4J: Normalizer.java
-public final class NormalizerBase implements Cloneable {
-
-    // The input text and our position in it
-    private UCharacterIterator  text;
-    private Normalizer2         norm2;
-    private Mode                mode;
-    private int                 options;
-
-    // The normalization buffer is the result of normalization
-    // of the source in [currentIndex..nextIndex] .
-    private int                 currentIndex;
-    private int                 nextIndex;
-
-    // A buffer for holding intermediate results
-    private StringBuilder       buffer;
-    private int                 bufferPos;
-
-    // Helper classes to defer loading of normalization data.
-    private static final class ModeImpl {
-        private ModeImpl(Normalizer2 n2) {
-            normalizer2 = n2;
-        }
-        private final Normalizer2 normalizer2;
-    }
-
-    private static final class NFDModeImpl {
-        private static final ModeImpl INSTANCE = new ModeImpl(Normalizer2.getNFDInstance());
-    }
-
-    private static final class NFKDModeImpl {
-        private static final ModeImpl INSTANCE = new ModeImpl(Normalizer2.getNFKDInstance());
-    }
-
-    private static final class NFCModeImpl {
-        private static final ModeImpl INSTANCE = new ModeImpl(Normalizer2.getNFCInstance());
-    }
-
-    private static final class NFKCModeImpl {
-        private static final ModeImpl INSTANCE = new ModeImpl(Normalizer2.getNFKCInstance());
-    }
-
-    private static final class Unicode32 {
-        private static final UnicodeSet INSTANCE = new UnicodeSet("[:age=3.2:]").freeze();
-    }
-
-    private static final class NFD32ModeImpl {
-        private static final ModeImpl INSTANCE =
-            new ModeImpl(new FilteredNormalizer2(Normalizer2.getNFDInstance(),
-                                                 Unicode32.INSTANCE));
-    }
-
-    private static final class NFKD32ModeImpl {
-        private static final ModeImpl INSTANCE =
-            new ModeImpl(new FilteredNormalizer2(Normalizer2.getNFKDInstance(),
-                                                 Unicode32.INSTANCE));
-    }
-
-    private static final class NFC32ModeImpl {
-        private static final ModeImpl INSTANCE =
-            new ModeImpl(new FilteredNormalizer2(Normalizer2.getNFCInstance(),
-                                                 Unicode32.INSTANCE));
-    }
-
-    private static final class NFKC32ModeImpl {
-        private static final ModeImpl INSTANCE =
-            new ModeImpl(new FilteredNormalizer2(Normalizer2.getNFKCInstance(),
-                                                 Unicode32.INSTANCE));
-    }
-
-    /**
-     * Options bit set value to select Unicode 3.2 normalization
-     * (except NormalizationCorrections).
-     * At most one Unicode version can be selected at a time.
-     * @stable ICU 2.6
-     */
-    public static final int UNICODE_3_2=0x20;
-
-    public static final int UNICODE_3_2_0_ORIGINAL=UNICODE_3_2;
-
-    /*
-     * Default option for the latest Unicode normalization. This option is
-     * provided mainly for testing.
-     * The value zero means that normalization is done with the fixes for
-     *   - Corrigendum 4 (Five CJK Canonical Mapping Errors)
-     *   - Corrigendum 5 (Normalization Idempotency)
-     */
-    public static final int UNICODE_LATEST = 0x00;
-
-    /**
-     * Constant indicating that the end of the iteration has been reached.
-     * This is guaranteed to have the same value as {@link UCharacterIterator#DONE}.
-     * @stable ICU 2.8
-     */
-    public static final int DONE = UCharacterIterator.DONE;
-
-    /**
-     * Constants for normalization modes.
-     * 

-     * The Mode class is not intended for public subclassing.
-     * Only the Mode constants provided by the Normalizer class should be used,
-     * and any fields or methods should not be called or overridden by users.
-     * @stable ICU 2.8
-     */
-    public abstract static class Mode {
-
-        /**
-         * Sole constructor
-         * @internal
-         * @deprecated This API is ICU internal only.
-         */
-        @Deprecated
-        protected Mode() {
-        }
-
-        /**
-         * @internal
-         * @deprecated This API is ICU internal only.
-         */
-        @Deprecated
-        protected abstract Normalizer2 getNormalizer2(int options);
-    }
-
-    private static Mode toMode(Normalizer.Form form) {
-        switch (form) {
-        case NFC :
-            return NFC;
-        case NFD :
-            return NFD;
-        case NFKC :
-            return NFKC;
-        case NFKD :
-            return NFKD;
-        }
-
-        throw new IllegalArgumentException("Unexpected normalization form: " +
-                                           form);
-    }
-
-    private static final class NONEMode extends Mode {
-        protected Normalizer2 getNormalizer2(int options) { return Norm2AllModes.NOOP_NORMALIZER2; }
-    }
-
-    private static final class NFDMode extends Mode {
-        protected Normalizer2 getNormalizer2(int options) {
-            return (options&UNICODE_3_2) != 0 ?
-                    NFD32ModeImpl.INSTANCE.normalizer2 :
-                    NFDModeImpl.INSTANCE.normalizer2;
-        }
-    }
-
-    private static final class NFKDMode extends Mode {
-        protected Normalizer2 getNormalizer2(int options) {
-            return (options&UNICODE_3_2) != 0 ?
-                    NFKD32ModeImpl.INSTANCE.normalizer2 :
-                    NFKDModeImpl.INSTANCE.normalizer2;
-        }
-    }
-
-    private static final class NFCMode extends Mode {
-        protected Normalizer2 getNormalizer2(int options) {
-            return (options&UNICODE_3_2) != 0 ?
-                    NFC32ModeImpl.INSTANCE.normalizer2 :
-                    NFCModeImpl.INSTANCE.normalizer2;
-        }
-    }
-
-    private static final class NFKCMode extends Mode {
-        protected Normalizer2 getNormalizer2(int options) {
-            return (options&UNICODE_3_2) != 0 ?
-                    NFKC32ModeImpl.INSTANCE.normalizer2 :
-                    NFKCModeImpl.INSTANCE.normalizer2;
-        }
-    }
-
-    /**
-     * No decomposition/composition.
-     * @stable ICU 2.8
-     */
-    public static final Mode NONE = new NONEMode();
-
-    /**
-     * Canonical decomposition.
-     * @stable ICU 2.8
-     */
-    public static final Mode NFD = new NFDMode();
-
-    /**
-     * Compatibility decomposition.
-     * @stable ICU 2.8
-     */
-    public static final Mode NFKD = new NFKDMode();
-
-    /**
-     * Canonical decomposition followed by canonical composition.
-     * @stable ICU 2.8
-     */
-    public static final Mode NFC = new NFCMode();
-
-    public static final Mode NFKC =new NFKCMode();
-
-    //-------------------------------------------------------------------------
-    // Iterator constructors
-    //-------------------------------------------------------------------------
-
-    /**
-     * Creates a new {@code NormalizerBase} object for iterating over the
-     * normalized form of a given string.
-     * 

-     * The {@code options} parameter specifies which optional
-     * {@code NormalizerBase} features are to be enabled for this object.
-     * 

-     * @param str  The string to be normalized.  The normalization
-     *              will start at the beginning of the string.
-     *
-     * @param mode The normalization mode.
-     *
-     * @param opt Any optional features to be enabled.
-     *            Currently the only available option is {@link #UNICODE_3_2}.
-     *            If you want the default behavior corresponding to one of the
-     *            standard Unicode Normalization Forms, use 0 for this argument.
-     * @stable ICU 2.6
-     */
-    public NormalizerBase(String str, Mode mode, int opt) {
-        this.text = UCharacterIterator.getInstance(str);
-        this.mode = mode;
-        this.options=opt;
-        norm2 = mode.getNormalizer2(opt);
-        buffer = new StringBuilder();
-    }
-
-    public NormalizerBase(String str, Mode mode) {
-       this(str, mode, 0);
-    }
-
-
-    /**
-     * Creates a new {@code NormalizerBase} object for iterating over the
-     * normalized form of the given text.
-     * 

-     * @param iter  The input text to be normalized.  The normalization
-     *              will start at the beginning of the string.
-     *
-     * @param mode  The normalization mode.
-     *
-     * @param opt Any optional features to be enabled.
-     *            Currently the only available option is {@link #UNICODE_3_2}.
-     *            If you want the default behavior corresponding to one of the
-     *            standard Unicode Normalization Forms, use 0 for this argument.
-     * @stable ICU 2.6
-     */
-    public NormalizerBase(CharacterIterator iter, Mode mode, int opt) {
-        this.text = UCharacterIterator.getInstance((CharacterIterator)iter.clone());
-        this.mode = mode;
-        this.options = opt;
-        norm2 = mode.getNormalizer2(opt);
-        buffer = new StringBuilder();
-    }
-
-    public NormalizerBase(CharacterIterator iter, Mode mode) {
-       this(iter, mode, 0);
-    }
-
-    /**
-     * Clones this {@code NormalizerBase} object.  All properties of this
-     * object are duplicated in the new object, including the cloning of any
-     * {@link CharacterIterator} that was passed in to the constructor
-     * or to {@link #setText(CharacterIterator) setText}.
-     * However, the text storage underlying
-     * the {@code CharacterIterator} is not duplicated unless the
-     * iterator's {@code clone} method does so.
-     * @stable ICU 2.8
-     */
-    public Object clone() {
-        try {
-            NormalizerBase copy = (NormalizerBase) super.clone();
-            copy.text = (UCharacterIterator) text.clone();
-            copy.mode = mode;
-            copy.options = options;
-            copy.norm2 = norm2;
-            copy.buffer = new StringBuilder(buffer);
-            copy.bufferPos = bufferPos;
-            copy.currentIndex = currentIndex;
-            copy.nextIndex = nextIndex;
-            return copy;
-        }
-        catch (CloneNotSupportedException e) {
-            throw new InternalError(e.toString(), e);
-        }
-    }
-
-    /**
-     * Normalizes a {@code String} using the given normalization operation.
-     * 

-     * The {@code options} parameter specifies which optional
-     * {@code NormalizerBase} features are to be enabled for this operation.
-     * Currently the only available option is {@link #UNICODE_3_2}.
-     * If you want the default behavior corresponding to one of the standard
-     * Unicode Normalization Forms, use 0 for this argument.
-     * 

-     * @param str       the input string to be normalized.
-     * @param mode      the normalization mode
-     * @param options   the optional features to be enabled.
-     * @return String   the normalized string
-     * @stable ICU 2.6
-     */
-    public static String normalize(String str, Mode mode, int options) {
-        return mode.getNormalizer2(options).normalize(str);
-    }
-
-    public static String normalize(String str, Normalizer.Form form) {
-        return NormalizerBase.normalize(str, toMode(form), UNICODE_LATEST);
-    }
-
-    public static String normalize(String str, Normalizer.Form form, int options) {
-        return NormalizerBase.normalize(str, toMode(form), options);
-    }
-
-    /**
-     * Test if a string is in a given normalization form.
-     * This is semantically equivalent to source.equals(normalize(source, mode)).
-     *
-     * Unlike quickCheck(), this function returns a definitive result,
-     * never a "maybe".
-     * For NFD, NFKD, and FCD, both functions work exactly the same.
-     * For NFC and NFKC where quickCheck may return "maybe", this function will
-     * perform further tests to arrive at a true/false result.
-     * @param str       the input string to be checked to see if it is
-     *                   normalized
-     * @param mode      the normalization mode
-     * @param options   Options for use with exclusion set and tailored Normalization
-     *                  The only option that is currently recognized is UNICODE_3_2
-     * @see #isNormalized
-     * @stable ICU 2.6
-     */
-    public static boolean isNormalized(String str, Mode mode, int options) {
-        return mode.getNormalizer2(options).isNormalized(str);
-    }
-
-    public static boolean isNormalized(String str, Normalizer.Form form) {
-        return NormalizerBase.isNormalized(str, toMode(form), UNICODE_LATEST);
-    }
-
-    public static boolean isNormalized(String str, Normalizer.Form form, int options) {
-        return NormalizerBase.isNormalized(str, toMode(form), options);
-    }
-
-    //-------------------------------------------------------------------------
-    // Iteration API
-    //-------------------------------------------------------------------------
-
-    /**
-     * Return the current character in the normalized text.
-     * @return The codepoint as an int
-     * @stable ICU 2.8
-     */
-    public int current() {
-        if(bufferPos0 || previousNormalize()) {
-            int c=buffer.codePointBefore(bufferPos);
-            bufferPos-=Character.charCount(c);
-            return c;
-        } else {
-            return DONE;
-        }
-    }
-
-    /**
-     * Reset the index to the beginning of the text.
-     * This is equivalent to setIndexOnly(startIndex)).
-     * @stable ICU 2.8
-     */
-    public void reset() {
-        text.setIndex(0);
-        currentIndex=nextIndex=0;
-        clearBuffer();
-    }
-
-    /**
-     * Set the iteration position in the input text that is being normalized,
-     * without any immediate normalization.
-     * After setIndexOnly(), getIndex() will return the same index that is
-     * specified here.
-     *
-     * @param index the desired index in the input text.
-     * @stable ICU 2.8
-     */
-    public void setIndexOnly(int index) {
-        text.setIndex(index);  // validates index
-        currentIndex=nextIndex=index;
-        clearBuffer();
-    }
-
-    /**
-     * Set the iteration position in the input text that is being normalized
-     * and return the first normalized character at that position.
-     * 

-     * Note: This method sets the position in the input text,
-     * while {@link #next} and {@link #previous} iterate through characters
-     * in the normalized output.  This means that there is not
-     * necessarily a one-to-one correspondence between characters returned
-     * by {@code next} and {@code previous} and the indices passed to and
-     * returned from {@code setIndex} and {@link #getIndex}.
-     * 

-     * @param index the desired index in the input text.
-     *
-     * @return   the first normalized character that is the result of iterating
-     *            forward starting at the given index.
-     *
-     * @throws IllegalArgumentException if the given index is less than
-     *          {@link #getBeginIndex} or greater than {@link #getEndIndex}.
-     * deprecated ICU 3.2
-     * @obsolete ICU 3.2
-     */
-     public int setIndex(int index) {
-         setIndexOnly(index);
-         return current();
-     }
-
-    /**
-     * Retrieve the index of the start of the input text. This is the begin
-     * index of the {@code CharacterIterator} or the start (i.e. 0) of the
-     * {@code String} over which this {@code NormalizerBase} is iterating
-     * @deprecated ICU 2.2. Use startIndex() instead.
-     * @return The codepoint as an int
-     * @see #startIndex
-     */
-    @Deprecated
-    public int getBeginIndex() {
-        return 0;
-    }
-
-    /**
-     * Retrieve the index of the end of the input text.  This is the end index
-     * of the {@code CharacterIterator} or the length of the {@code String}
-     * over which this {@code NormalizerBase} is iterating
-     * @deprecated ICU 2.2. Use endIndex() instead.
-     * @return The codepoint as an int
-     * @see #endIndex
-     */
-    @Deprecated
-    public int getEndIndex() {
-        return endIndex();
-    }
-
-    /**
-     * Retrieve the current iteration position in the input text that is
-     * being normalized.  This method is useful in applications such as
-     * searching, where you need to be able to determine the position in
-     * the input text that corresponds to a given normalized output character.
-     * 

-     * Note: This method sets the position in the input, while
-     * {@link #next} and {@link #previous} iterate through characters in the
-     * output.  This means that there is not necessarily a one-to-one
-     * correspondence between characters returned by {@code next} and
-     * {@code previous} and the indices passed to and returned from
-     * {@code setIndex} and {@link #getIndex}.
-     * @return The current iteration position
-     * @stable ICU 2.8
-     */
-    public int getIndex() {
-        if(bufferPos
-     * Note:If the normalization mode is changed while iterating
-     * over a string, calls to {@link #next} and {@link #previous} may
-     * return previously buffers characters in the old normalization mode
-     * until the iteration is able to re-sync at the next base character.
-     * It is safest to call {@link #setText setText()}, {@link #first},
-     * {@link #last}, etc. after calling {@code setMode}.
-     * 

-     * @param newMode the new mode for this {@code NormalizerBase}.
-     * The supported modes are:
-     * 
    
-     *  
  • {@link #NFC}    - Unicode canonical decompositiion
-     *                        followed by canonical composition.
-     *  
  • {@link #NFKC}   - Unicode compatibility decompositiion
-     *                        follwed by canonical composition.
-     *  
  • {@link #NFD}    - Unicode canonical decomposition
-     *  
  • {@link #NFKD}   - Unicode compatibility decomposition.
-     *  
  • {@link #NONE}   - Do nothing but return characters
-     *                        from the underlying input text.
-     * 

-     *
-     * @see #getMode
-     * @stable ICU 2.8
-     */
-    public void setMode(Mode newMode) {
-        mode = newMode;
-        norm2 = mode.getNormalizer2(options);
-    }
-
-    /**
-     * Return the basic operation performed by this {@code NormalizerBase}
-     *
-     * @see #setMode
-     * @stable ICU 2.8
-     */
-    public Mode getMode() {
-        return mode;
-    }
-
-    /**
-     * Set the input text over which this {@code NormalizerBase} will iterate.
-     * The iteration position is set to the beginning of the input text.
-     * @param newText   The new string to be normalized.
-     * @stable ICU 2.8
-     */
-    public void setText(String newText) {
-        UCharacterIterator newIter = UCharacterIterator.getInstance(newText);
-        if (newIter == null) {
-            throw new IllegalStateException("Could not create a new UCharacterIterator");
-        }
-        text = newIter;
-        reset();
-    }
-
-    /**
-     * Set the input text over which this {@code NormalizerBase} will iterate.
-     * The iteration position is set to the beginning of the input text.
-     * @param newText   The new string to be normalized.
-     * @stable ICU 2.8
-     */
-    public void setText(CharacterIterator newText) {
-        UCharacterIterator newIter = UCharacterIterator.getInstance(newText);
-        if (newIter == null) {
-            throw new IllegalStateException("Could not create a new UCharacterIterator");
-        }
-        text = newIter;
-        currentIndex=nextIndex=0;
-        clearBuffer();
-    }
-
-    private void clearBuffer() {
-        buffer.setLength(0);
-        bufferPos=0;
-    }
-
-    private boolean nextNormalize() {
-        clearBuffer();
-        currentIndex=nextIndex;
-        text.setIndex(nextIndex);
-        // Skip at least one character so we make progress.
-        int c=text.nextCodePoint();
-        if(c<0) {
-            return false;
-        }
-        StringBuilder segment=new StringBuilder().appendCodePoint(c);
-        while((c=text.nextCodePoint())>=0) {
-            if(norm2.hasBoundaryBefore(c)) {
-                text.moveCodePointIndex(-1);
-                break;
-            }
-            segment.appendCodePoint(c);
-        }
-        nextIndex=text.getIndex();
-        norm2.normalize(segment, buffer);
-        return buffer.length()!=0;
-    }
-
-    private boolean previousNormalize() {
-        clearBuffer();
-        nextIndex=currentIndex;
-        text.setIndex(currentIndex);
-        StringBuilder segment=new StringBuilder();
-        int c;
-        while((c=text.previousCodePoint())>=0) {
-            if(c<=0xffff) {
-                segment.insert(0, (char)c);
-            } else {
-                segment.insert(0, Character.toChars(c));
-            }
-            if(norm2.hasBoundaryBefore(c)) {
-                break;
-            }
-        }
-        currentIndex=text.getIndex();
-        norm2.normalize(segment, buffer);
-        bufferPos=buffer.length();
-        return buffer.length()!=0;
-    }
-
-}
--- /dev/null	2020-01-10 15:58:00.000000000 -0800
+++ new/src/java.base/share/classes/jdk/internal/icu/text/NormalizerBase.java	2020-01-10 15:58:00.000000000 -0800
@@ -0,0 +1,784 @@
+/*
+ * Copyright (c) 2005, 2020, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.  Oracle designates this
+ * particular file as subject to the "Classpath" exception as provided
+ * by Oracle in the LICENSE file that accompanied this code.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+/*
+ *******************************************************************************
+ * Copyright (C) 2000-2014, International Business Machines Corporation and
+ * others. All Rights Reserved.
+ *******************************************************************************
+ */
+package jdk.internal.icu.text;
+
+import jdk.internal.icu.impl.Norm2AllModes;
+
+import java.text.CharacterIterator;
+import java.text.Normalizer;
+
+/**
+ * Unicode Normalization
+ *
+ * 
Unicode normalization API

+ *
+ * normalize transforms Unicode text into an equivalent composed or
+ * decomposed form, allowing for easier sorting and searching of text.
+ * normalize supports the standard normalization forms described in
+ * 
+ * Unicode Standard Annex #15 — Unicode Normalization Forms.
+ *
+ * Characters with accents or other adornments can be encoded in
+ * several different ways in Unicode.  For example, take the character A-acute.
+ * In Unicode, this can be encoded as a single character (the
+ * "composed" form):
+ *
+ * 
+ *      00C1    LATIN CAPITAL LETTER A WITH ACUTE
+ * 

+ *
+ * or as two separate characters (the "decomposed" form):
+ *
+ * 
+ *      0041    LATIN CAPITAL LETTER A
+ *      0301    COMBINING ACUTE ACCENT
+ * 

+ *
+ * To a user of your program, however, both of these sequences should be
+ * treated as the same "user-level" character "A with acute accent".  When you
+ * are searching or comparing text, you must ensure that these two sequences are
+ * treated equivalently.  In addition, you must handle characters with more than
+ * one accent.  Sometimes the order of a character's combining accents is
+ * significant, while in other cases accent sequences in different orders are
+ * really equivalent.
+ *
+ * Similarly, the string "ffi" can be encoded as three separate letters:
+ *
+ * 
+ *      0066    LATIN SMALL LETTER F
+ *      0066    LATIN SMALL LETTER F
+ *      0069    LATIN SMALL LETTER I
+ * 

+ *
+ * or as the single character
+ *
+ * 
+ *      FB03    LATIN SMALL LIGATURE FFI
+ * 

+ *
+ * The ffi ligature is not a distinct semantic character, and strictly speaking
+ * it shouldn't be in Unicode at all, but it was included for compatibility
+ * with existing character sets that already provided it.  The Unicode standard
+ * identifies such characters by giving them "compatibility" decompositions
+ * into the corresponding semantic characters.  When sorting and searching, you
+ * will often want to use these mappings.
+ *
+ * normalize helps solve these problems by transforming text into
+ * the canonical composed and decomposed forms as shown in the first example
+ * above. In addition, you can have it perform compatibility decompositions so
+ * that you can treat compatibility characters the same as their equivalents.
+ * Finally, normalize rearranges accents into the proper canonical
+ * order, so that you do not have to worry about accent rearrangement on your
+ * own.
+ *
+ * Form FCD, "Fast C or D", is also designed for collation.
+ * It allows to work on strings that are not necessarily normalized
+ * with an algorithm (like in collation) that works under "canonical closure",
+ * i.e., it treats precomposed characters and their decomposed equivalents the
+ * same.
+ *
+ * It is not a normalization form because it does not provide for uniqueness of
+ * representation. Multiple strings may be canonically equivalent (their NFDs
+ * are identical) and may all conform to FCD without being identical themselves.
+ *
+ * The form is defined such that the "raw decomposition", the recursive
+ * canonical decomposition of each character, results in a string that is
+ * canonically ordered. This means that precomposed characters are allowed for
+ * as long as their decompositions do not need canonical reordering.
+ *
+ * Its advantage for a process like collation is that all NFD and most NFC texts
+ * - and many unnormalized texts - already conform to FCD and do not need to be
+ * normalized (NFD) for such a process. The FCD quick check will return YES for
+ * most strings in practice.
+ *
+ * normalize(FCD) may be implemented with NFD.
+ *
+ * For more details on FCD see Unicode Technical Note #5 (Canonical Equivalence in Applications):
+ * http://www.unicode.org/notes/tn5/#FCD
+ *
+ * ICU collation performs either NFD or FCD normalization automatically if
+ * normalization is turned on for the collator object. Beyond collation and
+ * string search, normalized strings may be useful for string equivalence
+ * comparisons, transliteration/transcription, unique representations, etc.
+ *
+ * The W3C generally recommends to exchange texts in NFC.
+ * Note also that most legacy character encodings use only precomposed forms and
+ * often do not encode any combining marks by themselves. For conversion to such
+ * character encodings the Unicode text needs to be normalized to NFC.
+ * For more usage examples, see the Unicode Standard Annex.
+ *
+ * Note: The Normalizer class also provides API for iterative normalization.
+ * While the setIndex() and getIndex() refer to indices in the
+ * underlying Unicode input text, the next() and previous() methods
+ * iterate through characters in the normalized output.
+ * This means that there is not necessarily a one-to-one correspondence
+ * between characters returned by next() and previous() and the indices
+ * passed to and returned from setIndex() and getIndex().
+ * It is for this reason that Normalizer does not implement the CharacterIterator interface.
+ *
+ * @stable ICU 2.8
+ */
+// Original filename in ICU4J: Normalizer.java
+public final class NormalizerBase implements Cloneable {
+
+    // The input text and our position in it
+    private UCharacterIterator  text;
+    private Normalizer2         norm2;
+    private Mode                mode;
+    private int                 options;
+
+    // The normalization buffer is the result of normalization
+    // of the source in [currentIndex..nextIndex] .
+    private int                 currentIndex;
+    private int                 nextIndex;
+
+    // A buffer for holding intermediate results
+    private StringBuilder       buffer;
+    private int                 bufferPos;
+
+    // Helper classes to defer loading of normalization data.
+    private static final class ModeImpl {
+        private ModeImpl(Normalizer2 n2) {
+            normalizer2 = n2;
+        }
+        private final Normalizer2 normalizer2;
+    }
+
+    private static final class NFDModeImpl {
+        private static final ModeImpl INSTANCE = new ModeImpl(Normalizer2.getNFDInstance());
+    }
+
+    private static final class NFKDModeImpl {
+        private static final ModeImpl INSTANCE = new ModeImpl(Normalizer2.getNFKDInstance());
+    }
+
+    private static final class NFCModeImpl {
+        private static final ModeImpl INSTANCE = new ModeImpl(Normalizer2.getNFCInstance());
+    }
+
+    private static final class NFKCModeImpl {
+        private static final ModeImpl INSTANCE = new ModeImpl(Normalizer2.getNFKCInstance());
+    }
+
+    private static final class Unicode32 {
+        private static final UnicodeSet INSTANCE = new UnicodeSet("[:age=3.2:]").freeze();
+    }
+
+    private static final class NFD32ModeImpl {
+        private static final ModeImpl INSTANCE =
+            new ModeImpl(new FilteredNormalizer2(Normalizer2.getNFDInstance(),
+                                                 Unicode32.INSTANCE));
+    }
+
+    private static final class NFKD32ModeImpl {
+        private static final ModeImpl INSTANCE =
+            new ModeImpl(new FilteredNormalizer2(Normalizer2.getNFKDInstance(),
+                                                 Unicode32.INSTANCE));
+    }
+
+    private static final class NFC32ModeImpl {
+        private static final ModeImpl INSTANCE =
+            new ModeImpl(new FilteredNormalizer2(Normalizer2.getNFCInstance(),
+                                                 Unicode32.INSTANCE));
+    }
+
+    private static final class NFKC32ModeImpl {
+        private static final ModeImpl INSTANCE =
+            new ModeImpl(new FilteredNormalizer2(Normalizer2.getNFKCInstance(),
+                                                 Unicode32.INSTANCE));
+    }
+
+    /**
+     * Options bit set value to select Unicode 3.2 normalization
+     * (except NormalizationCorrections).
+     * At most one Unicode version can be selected at a time.
+     * @stable ICU 2.6
+     */
+    public static final int UNICODE_3_2=0x20;
+
+    public static final int UNICODE_3_2_0_ORIGINAL=UNICODE_3_2;
+
+    /*
+     * Default option for the latest Unicode normalization. This option is
+     * provided mainly for testing.
+     * The value zero means that normalization is done with the fixes for
+     *   - Corrigendum 4 (Five CJK Canonical Mapping Errors)
+     *   - Corrigendum 5 (Normalization Idempotency)
+     */
+    public static final int UNICODE_LATEST = 0x00;
+
+    /**
+     * Constant indicating that the end of the iteration has been reached.
+     * This is guaranteed to have the same value as {@link UCharacterIterator#DONE}.
+     * @stable ICU 2.8
+     */
+    public static final int DONE = UCharacterIterator.DONE;
+
+    /**
+     * Constants for normalization modes.
+     * 

+     * The Mode class is not intended for public subclassing.
+     * Only the Mode constants provided by the Normalizer class should be used,
+     * and any fields or methods should not be called or overridden by users.
+     * @stable ICU 2.8
+     */
+    public abstract static class Mode {
+
+        /**
+         * Sole constructor
+         * @internal
+         * @deprecated This API is ICU internal only.
+         */
+        @Deprecated
+        protected Mode() {
+        }
+
+        /**
+         * @internal
+         * @deprecated This API is ICU internal only.
+         */
+        @Deprecated
+        protected abstract Normalizer2 getNormalizer2(int options);
+    }
+
+    private static Mode toMode(Normalizer.Form form) {
+        switch (form) {
+        case NFC :
+            return NFC;
+        case NFD :
+            return NFD;
+        case NFKC :
+            return NFKC;
+        case NFKD :
+            return NFKD;
+        }
+
+        throw new IllegalArgumentException("Unexpected normalization form: " +
+                                           form);
+    }
+
+    private static final class NONEMode extends Mode {
+        protected Normalizer2 getNormalizer2(int options) { return Norm2AllModes.NOOP_NORMALIZER2; }
+    }
+
+    private static final class NFDMode extends Mode {
+        protected Normalizer2 getNormalizer2(int options) {
+            return (options&UNICODE_3_2) != 0 ?
+                    NFD32ModeImpl.INSTANCE.normalizer2 :
+                    NFDModeImpl.INSTANCE.normalizer2;
+        }
+    }
+
+    private static final class NFKDMode extends Mode {
+        protected Normalizer2 getNormalizer2(int options) {
+            return (options&UNICODE_3_2) != 0 ?
+                    NFKD32ModeImpl.INSTANCE.normalizer2 :
+                    NFKDModeImpl.INSTANCE.normalizer2;
+        }
+    }
+
+    private static final class NFCMode extends Mode {
+        protected Normalizer2 getNormalizer2(int options) {
+            return (options&UNICODE_3_2) != 0 ?
+                    NFC32ModeImpl.INSTANCE.normalizer2 :
+                    NFCModeImpl.INSTANCE.normalizer2;
+        }
+    }
+
+    private static final class NFKCMode extends Mode {
+        protected Normalizer2 getNormalizer2(int options) {
+            return (options&UNICODE_3_2) != 0 ?
+                    NFKC32ModeImpl.INSTANCE.normalizer2 :
+                    NFKCModeImpl.INSTANCE.normalizer2;
+        }
+    }
+
+    /**
+     * No decomposition/composition.
+     * @stable ICU 2.8
+     */
+    public static final Mode NONE = new NONEMode();
+
+    /**
+     * Canonical decomposition.
+     * @stable ICU 2.8
+     */
+    public static final Mode NFD = new NFDMode();
+
+    /**
+     * Compatibility decomposition.
+     * @stable ICU 2.8
+     */
+    public static final Mode NFKD = new NFKDMode();
+
+    /**
+     * Canonical decomposition followed by canonical composition.
+     * @stable ICU 2.8
+     */
+    public static final Mode NFC = new NFCMode();
+
+    public static final Mode NFKC =new NFKCMode();
+
+    //-------------------------------------------------------------------------
+    // Iterator constructors
+    //-------------------------------------------------------------------------
+
+    /**
+     * Creates a new {@code NormalizerBase} object for iterating over the
+     * normalized form of a given string.
+     * 

+     * The {@code options} parameter specifies which optional
+     * {@code NormalizerBase} features are to be enabled for this object.
+     * 

+     * @param str  The string to be normalized.  The normalization
+     *              will start at the beginning of the string.
+     *
+     * @param mode The normalization mode.
+     *
+     * @param opt Any optional features to be enabled.
+     *            Currently the only available option is {@link #UNICODE_3_2}.
+     *            If you want the default behavior corresponding to one of the
+     *            standard Unicode Normalization Forms, use 0 for this argument.
+     * @stable ICU 2.6
+     */
+    public NormalizerBase(String str, Mode mode, int opt) {
+        this.text = UCharacterIterator.getInstance(str);
+        this.mode = mode;
+        this.options=opt;
+        norm2 = mode.getNormalizer2(opt);
+        buffer = new StringBuilder();
+    }
+
+    public NormalizerBase(String str, Mode mode) {
+       this(str, mode, 0);
+    }
+
+
+    /**
+     * Creates a new {@code NormalizerBase} object for iterating over the
+     * normalized form of the given text.
+     * 

+     * @param iter  The input text to be normalized.  The normalization
+     *              will start at the beginning of the string.
+     *
+     * @param mode  The normalization mode.
+     *
+     * @param opt Any optional features to be enabled.
+     *            Currently the only available option is {@link #UNICODE_3_2}.
+     *            If you want the default behavior corresponding to one of the
+     *            standard Unicode Normalization Forms, use 0 for this argument.
+     * @stable ICU 2.6
+     */
+    public NormalizerBase(CharacterIterator iter, Mode mode, int opt) {
+        this.text = UCharacterIterator.getInstance((CharacterIterator)iter.clone());
+        this.mode = mode;
+        this.options = opt;
+        norm2 = mode.getNormalizer2(opt);
+        buffer = new StringBuilder();
+    }
+
+    public NormalizerBase(CharacterIterator iter, Mode mode) {
+       this(iter, mode, 0);
+    }
+
+    /**
+     * Clones this {@code NormalizerBase} object.  All properties of this
+     * object are duplicated in the new object, including the cloning of any
+     * {@link CharacterIterator} that was passed in to the constructor
+     * or to {@link #setText(CharacterIterator) setText}.
+     * However, the text storage underlying
+     * the {@code CharacterIterator} is not duplicated unless the
+     * iterator's {@code clone} method does so.
+     * @stable ICU 2.8
+     */
+    public Object clone() {
+        try {
+            NormalizerBase copy = (NormalizerBase) super.clone();
+            copy.text = (UCharacterIterator) text.clone();
+            copy.mode = mode;
+            copy.options = options;
+            copy.norm2 = norm2;
+            copy.buffer = new StringBuilder(buffer);
+            copy.bufferPos = bufferPos;
+            copy.currentIndex = currentIndex;
+            copy.nextIndex = nextIndex;
+            return copy;
+        }
+        catch (CloneNotSupportedException e) {
+            throw new InternalError(e.toString(), e);
+        }
+    }
+
+    /**
+     * Normalizes a {@code String} using the given normalization operation.
+     * 

+     * The {@code options} parameter specifies which optional
+     * {@code NormalizerBase} features are to be enabled for this operation.
+     * Currently the only available option is {@link #UNICODE_3_2}.
+     * If you want the default behavior corresponding to one of the standard
+     * Unicode Normalization Forms, use 0 for this argument.
+     * 

+     * @param str       the input string to be normalized.
+     * @param mode      the normalization mode
+     * @param options   the optional features to be enabled.
+     * @return String   the normalized string
+     * @stable ICU 2.6
+     */
+    public static String normalize(String str, Mode mode, int options) {
+        return mode.getNormalizer2(options).normalize(str);
+    }
+
+    public static String normalize(String str, Normalizer.Form form) {
+        return NormalizerBase.normalize(str, toMode(form), UNICODE_LATEST);
+    }
+
+    public static String normalize(String str, Normalizer.Form form, int options) {
+        return NormalizerBase.normalize(str, toMode(form), options);
+    }
+
+    /**
+     * Test if a string is in a given normalization form.
+     * This is semantically equivalent to source.equals(normalize(source, mode)).
+     *
+     * Unlike quickCheck(), this function returns a definitive result,
+     * never a "maybe".
+     * For NFD, NFKD, and FCD, both functions work exactly the same.
+     * For NFC and NFKC where quickCheck may return "maybe", this function will
+     * perform further tests to arrive at a true/false result.
+     * @param str       the input string to be checked to see if it is
+     *                   normalized
+     * @param mode      the normalization mode
+     * @param options   Options for use with exclusion set and tailored Normalization
+     *                  The only option that is currently recognized is UNICODE_3_2
+     * @see #isNormalized
+     * @stable ICU 2.6
+     */
+    public static boolean isNormalized(String str, Mode mode, int options) {
+        return mode.getNormalizer2(options).isNormalized(str);
+    }
+
+    public static boolean isNormalized(String str, Normalizer.Form form) {
+        return NormalizerBase.isNormalized(str, toMode(form), UNICODE_LATEST);
+    }
+
+    public static boolean isNormalized(String str, Normalizer.Form form, int options) {
+        return NormalizerBase.isNormalized(str, toMode(form), options);
+    }
+
+    //-------------------------------------------------------------------------
+    // Iteration API
+    //-------------------------------------------------------------------------
+
+    /**
+     * Return the current character in the normalized text.
+     * @return The codepoint as an int
+     * @stable ICU 2.8
+     */
+    public int current() {
+        if(bufferPos0 || previousNormalize()) {
+            int c=buffer.codePointBefore(bufferPos);
+            bufferPos-=Character.charCount(c);
+            return c;
+        } else {
+            return DONE;
+        }
+    }
+
+    /**
+     * Reset the index to the beginning of the text.
+     * This is equivalent to setIndexOnly(startIndex)).
+     * @stable ICU 2.8
+     */
+    public void reset() {
+        text.setIndex(0);
+        currentIndex=nextIndex=0;
+        clearBuffer();
+    }
+
+    /**
+     * Set the iteration position in the input text that is being normalized,
+     * without any immediate normalization.
+     * After setIndexOnly(), getIndex() will return the same index that is
+     * specified here.
+     *
+     * @param index the desired index in the input text.
+     * @stable ICU 2.8
+     */
+    public void setIndexOnly(int index) {
+        text.setIndex(index);  // validates index
+        currentIndex=nextIndex=index;
+        clearBuffer();
+    }
+
+    /**
+     * Set the iteration position in the input text that is being normalized
+     * and return the first normalized character at that position.
+     * 

+     * Note: This method sets the position in the input text,
+     * while {@link #next} and {@link #previous} iterate through characters
+     * in the normalized output.  This means that there is not
+     * necessarily a one-to-one correspondence between characters returned
+     * by {@code next} and {@code previous} and the indices passed to and
+     * returned from {@code setIndex} and {@link #getIndex}.
+     * 

+     * @param index the desired index in the input text.
+     *
+     * @return   the first normalized character that is the result of iterating
+     *            forward starting at the given index.
+     *
+     * @throws IllegalArgumentException if the given index is less than
+     *          {@link #getBeginIndex} or greater than {@link #getEndIndex}.
+     * deprecated ICU 3.2
+     * @obsolete ICU 3.2
+     */
+     public int setIndex(int index) {
+         setIndexOnly(index);
+         return current();
+     }
+
+    /**
+     * Retrieve the index of the start of the input text. This is the begin
+     * index of the {@code CharacterIterator} or the start (i.e. 0) of the
+     * {@code String} over which this {@code NormalizerBase} is iterating
+     * @deprecated ICU 2.2. Use startIndex() instead.
+     * @return The codepoint as an int
+     * @see #startIndex
+     */
+    @Deprecated
+    public int getBeginIndex() {
+        return 0;
+    }
+
+    /**
+     * Retrieve the index of the end of the input text.  This is the end index
+     * of the {@code CharacterIterator} or the length of the {@code String}
+     * over which this {@code NormalizerBase} is iterating
+     * @deprecated ICU 2.2. Use endIndex() instead.
+     * @return The codepoint as an int
+     * @see #endIndex
+     */
+    @Deprecated
+    public int getEndIndex() {
+        return endIndex();
+    }
+
+    /**
+     * Retrieve the current iteration position in the input text that is
+     * being normalized.  This method is useful in applications such as
+     * searching, where you need to be able to determine the position in
+     * the input text that corresponds to a given normalized output character.
+     * 

+     * Note: This method sets the position in the input, while
+     * {@link #next} and {@link #previous} iterate through characters in the
+     * output.  This means that there is not necessarily a one-to-one
+     * correspondence between characters returned by {@code next} and
+     * {@code previous} and the indices passed to and returned from
+     * {@code setIndex} and {@link #getIndex}.
+     * @return The current iteration position
+     * @stable ICU 2.8
+     */
+    public int getIndex() {
+        if(bufferPos
+     * Note:If the normalization mode is changed while iterating
+     * over a string, calls to {@link #next} and {@link #previous} may
+     * return previously buffers characters in the old normalization mode
+     * until the iteration is able to re-sync at the next base character.
+     * It is safest to call {@link #setText setText()}, {@link #first},
+     * {@link #last}, etc. after calling {@code setMode}.
+     * 

+     * @param newMode the new mode for this {@code NormalizerBase}.
+     * The supported modes are:
+     * 
    
+     *  
  • {@link #NFC}    - Unicode canonical decompositiion
+     *                        followed by canonical composition.
+     *  
  • {@link #NFKC}   - Unicode compatibility decompositiion
+     *                        follwed by canonical composition.
+     *  
  • {@link #NFD}    - Unicode canonical decomposition
+     *  
  • {@link #NFKD}   - Unicode compatibility decomposition.
+     *  
  • {@link #NONE}   - Do nothing but return characters
+     *                        from the underlying input text.
+     * 

+     *
+     * @see #getMode
+     * @stable ICU 2.8
+     */
+    public void setMode(Mode newMode) {
+        mode = newMode;
+        norm2 = mode.getNormalizer2(options);
+    }
+
+    /**
+     * Return the basic operation performed by this {@code NormalizerBase}
+     *
+     * @see #setMode
+     * @stable ICU 2.8
+     */
+    public Mode getMode() {
+        return mode;
+    }
+
+    /**
+     * Set the input text over which this {@code NormalizerBase} will iterate.
+     * The iteration position is set to the beginning of the input text.
+     * @param newText   The new string to be normalized.
+     * @stable ICU 2.8
+     */
+    public void setText(String newText) {
+        UCharacterIterator newIter = UCharacterIterator.getInstance(newText);
+        if (newIter == null) {
+            throw new IllegalStateException("Could not create a new UCharacterIterator");
+        }
+        text = newIter;
+        reset();
+    }
+
+    /**
+     * Set the input text over which this {@code NormalizerBase} will iterate.
+     * The iteration position is set to the beginning of the input text.
+     * @param newText   The new string to be normalized.
+     * @stable ICU 2.8
+     */
+    public void setText(CharacterIterator newText) {
+        UCharacterIterator newIter = UCharacterIterator.getInstance(newText);
+        if (newIter == null) {
+            throw new IllegalStateException("Could not create a new UCharacterIterator");
+        }
+        text = newIter;
+        currentIndex=nextIndex=0;
+        clearBuffer();
+    }
+
+    private void clearBuffer() {
+        buffer.setLength(0);
+        bufferPos=0;
+    }
+
+    private boolean nextNormalize() {
+        clearBuffer();
+        currentIndex=nextIndex;
+        text.setIndex(nextIndex);
+        // Skip at least one character so we make progress.
+        int c=text.nextCodePoint();
+        if(c<0) {
+            return false;
+        }
+        StringBuilder segment=new StringBuilder().appendCodePoint(c);
+        while((c=text.nextCodePoint())>=0) {
+            if(norm2.hasBoundaryBefore(c)) {
+                text.moveCodePointIndex(-1);
+                break;
+            }
+            segment.appendCodePoint(c);
+        }
+        nextIndex=text.getIndex();
+        norm2.normalize(segment, buffer);
+        return buffer.length()!=0;
+    }
+
+    private boolean previousNormalize() {
+        clearBuffer();
+        nextIndex=currentIndex;
+        text.setIndex(currentIndex);
+        StringBuilder segment=new StringBuilder();
+        int c;
+        while((c=text.previousCodePoint())>=0) {
+            if(c<=0xffff) {
+                segment.insert(0, (char)c);
+            } else {
+                segment.insert(0, Character.toChars(c));
+            }
+            if(norm2.hasBoundaryBefore(c)) {
+                break;
+            }
+        }
+        currentIndex=text.getIndex();
+        norm2.normalize(segment, buffer);
+        bufferPos=buffer.length();
+        return buffer.length()!=0;
+    }
+
+}
--- old/src/java.base/share/classes/sun/text/normalizer/Replaceable.java	2020-01-10 15:58:02.000000000 -0800
+++ /dev/null	2020-01-10 15:58:02.000000000 -0800
@@ -1,121 +0,0 @@
-/*
- * Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved.
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.  Oracle designates this
- * particular file as subject to the "Classpath" exception as provided
- * by Oracle in the LICENSE file that accompanied this code.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- */
-
-/*
- *******************************************************************************
- * (C) Copyright IBM Corp. 1996-2005 - All Rights Reserved                     *
- *                                                                             *
- * The original version of this source code and documentation is copyrighted   *
- * and owned by IBM, These materials are provided under terms of a License     *
- * Agreement between IBM and Sun. This technology is protected by multiple     *
- * US and International patents. This notice and attribution to IBM may not    *
- * to removed.                                                                 *
- *******************************************************************************
- */
-
-package sun.text.normalizer;
-
-/**
- * Replaceable is an interface representing a
- * string of characters that supports the replacement of a range of
- * itself with a new string of characters.  It is used by APIs that
- * change a piece of text while retaining metadata.  Metadata is data
- * other than the Unicode characters returned by char32At().  One
- * example of metadata is style attributes; another is an edit
- * history, marking each character with an author and revision number.
- *
- * 
An implicit aspect of the Replaceable API is that
- * during a replace operation, new characters take on the metadata of
- * the old characters.  For example, if the string "the bold
- * font" has range (4, 8) replaced with "strong", then it becomes "the
- * strong font".
- *
- * 
Replaceable specifies ranges using a start
- * offset and a limit offset.  The range of characters thus specified
- * includes the characters at offset start..limit-1.  That is, the
- * start offset is inclusive, and the limit offset is exclusive.
- *
- * 
Replaceable also includes API to access characters
- * in the string: length(), charAt(),
- * char32At(), and extractBetween().
- *
- * 
For a subclass to support metadata, typical behavior of
- * replace() is the following:
- * 
    
- *   
  • Set the metadata of the new text to the metadata of the first
- *   character replaced
    • 
- *   
  • If no characters are replaced, use the metadata of the
- *   previous character
    • 
- *   
  • If there is no previous character (i.e. start == 0), use the
- *   following character
    • 
- *   
  • If there is no following character (i.e. the replaceable was
- *   empty), use default metadata
    • 
- *   
  • If the code point U+FFFF is seen, it should be interpreted as
- *   a special marker having no metadata
    • 
- * 

- * If this is not the behavior, the subclass should document any differences.
- *
- * 
Copyright © IBM Corporation 1999.  All rights reserved.
- *
- * @author Alan Liu
- * @stable ICU 2.0
- */
-interface Replaceable {
-    /**
-     * Returns the number of 16-bit code units in the text.
-     * @return number of 16-bit code units in text
-     * @stable ICU 2.0
-     */
-    int length();
-
-    /**
-     * Returns the 16-bit code unit at the given offset into the text.
-     * @param offset an integer between 0 and length()-1
-     * inclusive
-     * @return 16-bit code unit of text at given offset
-     * @stable ICU 2.0
-     */
-    char charAt(int offset);
-
-    /**
-     * Copies characters from this object into the destination
-     * character array.  The first character to be copied is at index
-     * srcStart; the last character to be copied is at
-     * index srcLimit-1 (thus the total number of
-     * characters to be copied is srcLimit-srcStart). The
-     * characters are copied into the subarray of dst
-     * starting at index dstStart and ending at index
-     * dstStart + (srcLimit-srcStart) - 1.
-     *
-     * @param srcStart the beginning index to copy, inclusive;
-     *        {@code 0 <= start <= limit}.
-     * @param srcLimit the ending index to copy, exclusive;
-     *        {@code start <= limit <= length()}.
-     * @param dst the destination array.
-     * @param dstStart the start offset in the destination array.
-     * @stable ICU 2.0
-     */
-    void getChars(int srcStart, int srcLimit, char dst[], int dstStart);
-}
--- /dev/null	2020-01-10 15:58:02.000000000 -0800
+++ new/src/java.base/share/classes/jdk/internal/icu/text/Replaceable.java	2020-01-10 15:58:01.000000000 -0800
@@ -0,0 +1,121 @@
+/*
+ * Copyright (c) 2005, 2020, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.  Oracle designates this
+ * particular file as subject to the "Classpath" exception as provided
+ * by Oracle in the LICENSE file that accompanied this code.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+/*
+ *******************************************************************************
+ * (C) Copyright IBM Corp. 1996-2005 - All Rights Reserved                     *
+ *                                                                             *
+ * The original version of this source code and documentation is copyrighted   *
+ * and owned by IBM, These materials are provided under terms of a License     *
+ * Agreement between IBM and Sun. This technology is protected by multiple     *
+ * US and International patents. This notice and attribution to IBM may not    *
+ * to removed.                                                                 *
+ *******************************************************************************
+ */
+
+package jdk.internal.icu.text;
+
+/**
+ * Replaceable is an interface representing a
+ * string of characters that supports the replacement of a range of
+ * itself with a new string of characters.  It is used by APIs that
+ * change a piece of text while retaining metadata.  Metadata is data
+ * other than the Unicode characters returned by char32At().  One
+ * example of metadata is style attributes; another is an edit
+ * history, marking each character with an author and revision number.
+ *
+ * 
An implicit aspect of the Replaceable API is that
+ * during a replace operation, new characters take on the metadata of
+ * the old characters.  For example, if the string "the bold
+ * font" has range (4, 8) replaced with "strong", then it becomes "the
+ * strong font".
+ *
+ * 
Replaceable specifies ranges using a start
+ * offset and a limit offset.  The range of characters thus specified
+ * includes the characters at offset start..limit-1.  That is, the
+ * start offset is inclusive, and the limit offset is exclusive.
+ *
+ * 
Replaceable also includes API to access characters
+ * in the string: length(), charAt(),
+ * char32At(), and extractBetween().
+ *
+ * 
For a subclass to support metadata, typical behavior of
+ * replace() is the following:
+ * 
    
+ *   
  • Set the metadata of the new text to the metadata of the first
+ *   character replaced
    • 
+ *   
  • If no characters are replaced, use the metadata of the
+ *   previous character
    • 
+ *   
  • If there is no previous character (i.e. start == 0), use the
+ *   following character
    • 
+ *   
  • If there is no following character (i.e. the replaceable was
+ *   empty), use default metadata
    • 
+ *   
  • If the code point U+FFFF is seen, it should be interpreted as
+ *   a special marker having no metadata
    • 
+ * 

+ * If this is not the behavior, the subclass should document any differences.
+ *
+ * 
Copyright © IBM Corporation 1999.  All rights reserved.
+ *
+ * @author Alan Liu
+ * @stable ICU 2.0
+ */
+public interface Replaceable {
+    /**
+     * Returns the number of 16-bit code units in the text.
+     * @return number of 16-bit code units in text
+     * @stable ICU 2.0
+     */
+    int length();
+
+    /**
+     * Returns the 16-bit code unit at the given offset into the text.
+     * @param offset an integer between 0 and length()-1
+     * inclusive
+     * @return 16-bit code unit of text at given offset
+     * @stable ICU 2.0
+     */
+    char charAt(int offset);
+
+    /**
+     * Copies characters from this object into the destination
+     * character array.  The first character to be copied is at index
+     * srcStart; the last character to be copied is at
+     * index srcLimit-1 (thus the total number of
+     * characters to be copied is srcLimit-srcStart). The
+     * characters are copied into the subarray of dst
+     * starting at index dstStart and ending at index
+     * dstStart + (srcLimit-srcStart) - 1.
+     *
+     * @param srcStart the beginning index to copy, inclusive;
+     *        {@code 0 <= start <= limit}.
+     * @param srcLimit the ending index to copy, exclusive;
+     *        {@code start <= limit <= length()}.
+     * @param dst the destination array.
+     * @param dstStart the start offset in the destination array.
+     * @stable ICU 2.0
+     */
+    void getChars(int srcStart, int srcLimit, char dst[], int dstStart);
+}
--- old/src/java.base/share/classes/sun/text/normalizer/ReplaceableString.java	2020-01-10 15:58:03.000000000 -0800
+++ /dev/null	2020-01-10 15:58:03.000000000 -0800
@@ -1,118 +0,0 @@
-/*
- * Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved.
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.  Oracle designates this
- * particular file as subject to the "Classpath" exception as provided
- * by Oracle in the LICENSE file that accompanied this code.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- */
-
-/*
- *******************************************************************************
- * Copyright (C) 1996-2009, International Business Machines Corporation and    *
- * others. All Rights Reserved.                                                *
- *******************************************************************************
- */
-
-package sun.text.normalizer;
-
-/**
- * ReplaceableString is an adapter class that implements the
- * Replaceable API around an ordinary StringBuffer.
- *
- * 
Note: This class does not support attributes and is not
- * intended for general use.  Most clients will need to implement
- * {@link Replaceable} in their text representation class.
- *
- * 
Copyright © IBM Corporation 1999.  All rights reserved.
- *
- * @see Replaceable
- * @author Alan Liu
- * @stable ICU 2.0
- */
-class ReplaceableString implements Replaceable {
-
-    private StringBuffer buf;
-
-    /**
-     * Construct a new object with the given initial contents.
-     * @param str initial contents
-     * @stable ICU 2.0
-     */
-    public ReplaceableString(String str) {
-        buf = new StringBuffer(str);
-    }
-
-    /**
-     * Construct a new object using buf for internal
-     * storage.  The contents of buf at the time of
-     * construction are used as the initial contents.  Note!
-     * Modifications to buf will modify this object, and
-     * vice versa.
-     * @param buf object to be used as internal storage
-     * @stable ICU 2.0
-     */
-    public ReplaceableString(StringBuffer buf) {
-        this.buf = buf;
-    }
-
-    /**
-     * Return the number of characters contained in this object.
-     * Replaceable API.
-     * @stable ICU 2.0
-     */
-    public int length() {
-        return buf.length();
-    }
-
-    /**
-     * Return the character at the given position in this object.
-     * Replaceable API.
-     * @param offset offset into the contents, from 0 to
-     * length() - 1
-     * @stable ICU 2.0
-     */
-    public char charAt(int offset) {
-        return buf.charAt(offset);
-    }
-
-    /**
-     * Copies characters from this object into the destination
-     * character array.  The first character to be copied is at index
-     * srcStart; the last character to be copied is at
-     * index srcLimit-1 (thus the total number of
-     * characters to be copied is srcLimit-srcStart). The
-     * characters are copied into the subarray of dst
-     * starting at index dstStart and ending at index
-     * dstStart + (srcLimit-srcStart) - 1.
-     *
-     * @param srcStart the beginning index to copy, inclusive;
-     *        {@code 0 <= start <= limit}.
-     * @param srcLimit the ending index to copy, exclusive;
-     *        {@code start <= limit <= length()}.
-     * @param dst the destination array.
-     * @param dstStart the start offset in the destination array.
-     * @stable ICU 2.0
-     */
-    public void getChars(int srcStart, int srcLimit, char dst[], int dstStart) {
-        if (srcStart != srcLimit) {
-            buf.getChars(srcStart, srcLimit, dst, dstStart);
-        }
-    }
-}
--- /dev/null	2020-01-10 15:58:03.000000000 -0800
+++ new/src/java.base/share/classes/jdk/internal/icu/text/ReplaceableString.java	2020-01-10 15:58:03.000000000 -0800
@@ -0,0 +1,118 @@
+/*
+ * Copyright (c) 2005, 2020, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.  Oracle designates this
+ * particular file as subject to the "Classpath" exception as provided
+ * by Oracle in the LICENSE file that accompanied this code.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+/*
+ *******************************************************************************
+ * Copyright (C) 1996-2009, International Business Machines Corporation and    *
+ * others. All Rights Reserved.                                                *
+ *******************************************************************************
+ */
+
+package jdk.internal.icu.text;
+
+/**
+ * ReplaceableString is an adapter class that implements the
+ * Replaceable API around an ordinary StringBuffer.
+ *
+ * 
Note: This class does not support attributes and is not
+ * intended for general use.  Most clients will need to implement
+ * {@link Replaceable} in their text representation class.
+ *
+ * 
Copyright © IBM Corporation 1999.  All rights reserved.
+ *
+ * @see Replaceable
+ * @author Alan Liu
+ * @stable ICU 2.0
+ */
+public class ReplaceableString implements Replaceable {
+
+    private StringBuffer buf;
+
+    /**
+     * Construct a new object with the given initial contents.
+     * @param str initial contents
+     * @stable ICU 2.0
+     */
+    public ReplaceableString(String str) {
+        buf = new StringBuffer(str);
+    }
+
+    /**
+     * Construct a new object using buf for internal
+     * storage.  The contents of buf at the time of
+     * construction are used as the initial contents.  Note!
+     * Modifications to buf will modify this object, and
+     * vice versa.
+     * @param buf object to be used as internal storage
+     * @stable ICU 2.0
+     */
+    public ReplaceableString(StringBuffer buf) {
+        this.buf = buf;
+    }
+
+    /**
+     * Return the number of characters contained in this object.
+     * Replaceable API.
+     * @stable ICU 2.0
+     */
+    public int length() {
+        return buf.length();
+    }
+
+    /**
+     * Return the character at the given position in this object.
+     * Replaceable API.
+     * @param offset offset into the contents, from 0 to
+     * length() - 1
+     * @stable ICU 2.0
+     */
+    public char charAt(int offset) {
+        return buf.charAt(offset);
+    }
+
+    /**
+     * Copies characters from this object into the destination
+     * character array.  The first character to be copied is at index
+     * srcStart; the last character to be copied is at
+     * index srcLimit-1 (thus the total number of
+     * characters to be copied is srcLimit-srcStart). The
+     * characters are copied into the subarray of dst
+     * starting at index dstStart and ending at index
+     * dstStart + (srcLimit-srcStart) - 1.
+     *
+     * @param srcStart the beginning index to copy, inclusive;
+     *        {@code 0 <= start <= limit}.
+     * @param srcLimit the ending index to copy, exclusive;
+     *        {@code start <= limit <= length()}.
+     * @param dst the destination array.
+     * @param dstStart the start offset in the destination array.
+     * @stable ICU 2.0
+     */
+    public void getChars(int srcStart, int srcLimit, char dst[], int dstStart) {
+        if (srcStart != srcLimit) {
+            buf.getChars(srcStart, srcLimit, dst, dstStart);
+        }
+    }
+}
--- old/src/java.base/share/classes/sun/net/idn/StringPrep.java	2020-01-10 15:58:04.000000000 -0800
+++ /dev/null	2020-01-10 15:58:04.000000000 -0800
@@ -1,486 +0,0 @@
-/*
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.  Oracle designates this
- * particular file as subject to the "Classpath" exception as provided
- * by Oracle in the LICENSE file that accompanied this code.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- */
-/*
-/*
- *******************************************************************************
- * Copyright (C) 2003-2004, International Business Machines Corporation and         *
- * others. All Rights Reserved.                                                *
- *******************************************************************************
- */
-//
-// CHANGELOG
-//      2005-05-19 Edward Wang
-//          - copy this file from icu4jsrc_3_2/src/com/ibm/icu/text/StringPrep.java
-//          - move from package com.ibm.icu.text to package sun.net.idn
-//          - use ParseException instead of StringPrepParseException
-//          - change 'Normalizer.getUnicodeVersion()' to 'NormalizerImpl.getUnicodeVersion()'
-//          - remove all @deprecated tag to make compiler happy
-//      2007-08-14 Martin Buchholz
-//          - remove redundant casts
-//
-package sun.net.idn;
-
-import java.io.BufferedInputStream;
-import java.io.ByteArrayInputStream;
-import java.io.IOException;
-import java.io.InputStream;
-import java.text.ParseException;
-
-import sun.text.Normalizer;
-import sun.text.normalizer.CharTrie;
-import sun.text.normalizer.Trie;
-import sun.text.normalizer.VersionInfo;
-import sun.text.normalizer.UCharacter;
-import sun.text.normalizer.UCharacterIterator;
-import sun.text.normalizer.UTF16;
-import sun.net.idn.UCharacterDirection;
-import sun.net.idn.StringPrepDataReader;
-
-/**
- * StringPrep API implements the StingPrep framework as described by
- * RFC 3454.
- * StringPrep prepares Unicode strings for use in network protocols.
- * Profiles of StingPrep are set of rules and data according to which the
- * Unicode Strings are prepared. Each profiles contains tables which describe
- * how a code point should be treated. The tables are broadly classied into
- * 
    
- *     
  •  Unassigned Table: Contains code points that are unassigned
- *          in the Unicode Version supported by StringPrep. Currently
- *          RFC 3454 supports Unicode 3.2. 
    • 
- *     
  •  Prohibited Table: Contains code points that are prohibted from
- *          the output of the StringPrep processing function. 
    • 
- *     
  •  Mapping Table: Contains code ponts that are deleted from the output or case mapped. 
    • 
- * 

- *
- * The procedure for preparing Unicode strings:
- * 
    
- *      
  1.  Map: For each character in the input, check if it has a mapping
- *           and, if so, replace it with its mapping. 
    2. 
- *      
  2.  Normalize: Possibly normalize the result of step 1 using Unicode
- *           normalization. 
    3. 
- *      
  3.  Prohibit: Check for any characters that are not allowed in the
- *           output.  If any are found, return an error.
    4. 
- *      
  4.  Check bidi: Possibly check for right-to-left characters, and if
- *           any are found, make sure that the whole string satisfies the
- *           requirements for bidirectional strings.  If the string does not
- *           satisfy the requirements for bidirectional strings, return an
- *           error.  
    5. 
- * 

- * @author Ram Viswanadha
- * @draft ICU 2.8
- */
-public final class StringPrep {
-    /**
-     * Option to prohibit processing of unassigned code points in the input
-     *
-     * @see   #prepare
-     * @draft ICU 2.8
-     */
-    public static final int DEFAULT = 0x0000;
-
-    /**
-     * Option to allow processing of unassigned code points in the input
-     *
-     * @see   #prepare
-     * @draft ICU 2.8
-     */
-    public static final int ALLOW_UNASSIGNED = 0x0001;
-
-    private static final int UNASSIGNED        = 0x0000;
-    private static final int MAP               = 0x0001;
-    private static final int PROHIBITED        = 0x0002;
-    private static final int DELETE            = 0x0003;
-    private static final int TYPE_LIMIT        = 0x0004;
-
-    private static final int NORMALIZATION_ON  = 0x0001;
-    private static final int CHECK_BIDI_ON     = 0x0002;
-
-    private static final int TYPE_THRESHOLD       = 0xFFF0;
-    private static final int MAX_INDEX_VALUE      = 0x3FBF;   /*16139*/
-    private static final int MAX_INDEX_TOP_LENGTH = 0x0003;
-
-    /* indexes[] value names */
-    private static final int INDEX_TRIE_SIZE                  =  0; /* number of bytes in normalization trie */
-    private static final int INDEX_MAPPING_DATA_SIZE          =  1; /* The array that contains the mapping   */
-    private static final int NORM_CORRECTNS_LAST_UNI_VERSION  =  2; /* The index of Unicode version of last entry in NormalizationCorrections.txt */
-    private static final int ONE_UCHAR_MAPPING_INDEX_START    =  3; /* The starting index of 1 UChar mapping index in the mapping data array */
-    private static final int TWO_UCHARS_MAPPING_INDEX_START   =  4; /* The starting index of 2 UChars mapping index in the mapping data array */
-    private static final int THREE_UCHARS_MAPPING_INDEX_START =  5;
-    private static final int FOUR_UCHARS_MAPPING_INDEX_START  =  6;
-    private static final int OPTIONS                          =  7; /* Bit set of options to turn on in the profile */
-    private static final int INDEX_TOP                        = 16;                          /* changing this requires a new formatVersion */
-
-
-    /**
-     * Default buffer size of datafile
-     */
-    private static final int DATA_BUFFER_SIZE = 25000;
-
-    /* Wrappers for Trie implementations */
-    private static final class StringPrepTrieImpl implements Trie.DataManipulate{
-        private CharTrie sprepTrie = null;
-       /**
-        * Called by com.ibm.icu.util.Trie to extract from a lead surrogate's
-        * data the index array offset of the indexes for that lead surrogate.
-        * @param property data value for a surrogate from the trie, including
-        *        the folding offset
-        * @return data offset or 0 if there is no data for the lead surrogate
-        */
-         public int getFoldingOffset(int value){
-            return value;
-        }
-    }
-
-    // CharTrie implementation for reading the trie data
-    private StringPrepTrieImpl sprepTrieImpl;
-    // Indexes read from the data file
-    private int[] indexes;
-    // mapping data read from the data file
-    private char[] mappingData;
-    // format version of the data file
-    private byte[] formatVersion;
-    // the version of Unicode supported by the data file
-    private VersionInfo sprepUniVer;
-    // the Unicode version of last entry in the
-    // NormalizationCorrections.txt file if normalization
-    // is turned on
-    private VersionInfo normCorrVer;
-    // Option to turn on Normalization
-    private boolean doNFKC;
-    // Option to turn on checking for BiDi rules
-    private boolean checkBiDi;
-
-
-    private char getCodePointValue(int ch){
-        return sprepTrieImpl.sprepTrie.getCodePointValue(ch);
-    }
-
-    private static VersionInfo getVersionInfo(int comp){
-        int micro = comp & 0xFF;
-        int milli =(comp >> 8)  & 0xFF;
-        int minor =(comp >> 16) & 0xFF;
-        int major =(comp >> 24) & 0xFF;
-        return VersionInfo.getInstance(major,minor,milli,micro);
-    }
-    private static VersionInfo getVersionInfo(byte[] version){
-        if(version.length != 4){
-            return null;
-        }
-        return VersionInfo.getInstance((int)version[0],(int) version[1],(int) version[2],(int) version[3]);
-    }
-    /**
-     * Creates an StringPrep object after reading the input stream.
-     * The object does not hold a reference to the input steam, so the stream can be
-     * closed after the method returns.
-     *
-     * @param inputStream The stream for reading the StringPrep profile binarySun
-     * @throws IOException
-     * @draft ICU 2.8
-     */
-    public StringPrep(InputStream inputStream) throws IOException{
-
-        BufferedInputStream b = new BufferedInputStream(inputStream,DATA_BUFFER_SIZE);
-
-        StringPrepDataReader reader = new StringPrepDataReader(b);
-
-        // read the indexes
-        indexes = reader.readIndexes(INDEX_TOP);
-
-        byte[] sprepBytes = new byte[indexes[INDEX_TRIE_SIZE]];
-
-
-        //indexes[INDEX_MAPPING_DATA_SIZE] store the size of mappingData in bytes
-        mappingData = new char[indexes[INDEX_MAPPING_DATA_SIZE]/2];
-        // load the rest of the data and initialize the data members
-        reader.read(sprepBytes,mappingData);
-
-        sprepTrieImpl           = new StringPrepTrieImpl();
-        sprepTrieImpl.sprepTrie = new CharTrie( new ByteArrayInputStream(sprepBytes),sprepTrieImpl  );
-
-        // get the data format version
-        formatVersion = reader.getDataFormatVersion();
-
-        // get the options
-        doNFKC            = ((indexes[OPTIONS] & NORMALIZATION_ON) > 0);
-        checkBiDi         = ((indexes[OPTIONS] & CHECK_BIDI_ON) > 0);
-        sprepUniVer   = getVersionInfo(reader.getUnicodeVersion());
-        normCorrVer   = getVersionInfo(indexes[NORM_CORRECTNS_LAST_UNI_VERSION]);
-        VersionInfo normUniVer = UCharacter.getUnicodeVersion();
-        if(normUniVer.compareTo(sprepUniVer) < 0 && /* the Unicode version of SPREP file must be less than the Unicode Vesion of the normalization data */
-           normUniVer.compareTo(normCorrVer) < 0 && /* the Unicode version of the NormalizationCorrections.txt file should be less than the Unicode Vesion of the normalization data */
-           ((indexes[OPTIONS] & NORMALIZATION_ON) > 0) /* normalization turned on*/
-           ){
-            throw new IOException("Normalization Correction version not supported");
-        }
-        b.close();
-    }
-
-    private static final class Values{
-        boolean isIndex;
-        int value;
-        int type;
-        public void reset(){
-            isIndex = false;
-            value = 0;
-            type = -1;
-        }
-    }
-
-    private static final void getValues(char trieWord,Values values){
-        values.reset();
-        if(trieWord == 0){
-            /*
-             * Initial value stored in the mapping table
-             * just return TYPE_LIMIT .. so that
-             * the source codepoint is copied to the destination
-             */
-            values.type = TYPE_LIMIT;
-        }else if(trieWord >= TYPE_THRESHOLD){
-            values.type = (trieWord - TYPE_THRESHOLD);
-        }else{
-            /* get the type */
-            values.type = MAP;
-            /* ascertain if the value is index or delta */
-            if((trieWord & 0x02)>0){
-                values.isIndex = true;
-                values.value = trieWord  >> 2; //mask off the lower 2 bits and shift
-
-            }else{
-                values.isIndex = false;
-                values.value = (trieWord<<16)>>16;
-                values.value =  (values.value >> 2);
-
-            }
-
-            if((trieWord>>2) == MAX_INDEX_VALUE){
-                values.type = DELETE;
-                values.isIndex = false;
-                values.value = 0;
-            }
-        }
-    }
-
-
-
-    private StringBuffer map( UCharacterIterator iter, int options)
-                            throws ParseException {
-
-        Values val = new Values();
-        char result = 0;
-        int ch  = UCharacterIterator.DONE;
-        StringBuffer dest = new StringBuffer();
-        boolean allowUnassigned = ((options & ALLOW_UNASSIGNED)>0);
-
-        while((ch=iter.nextCodePoint())!= UCharacterIterator.DONE){
-
-            result = getCodePointValue(ch);
-            getValues(result,val);
-
-            // check if the source codepoint is unassigned
-            if(val.type == UNASSIGNED && allowUnassigned == false){
-                 throw new ParseException("An unassigned code point was found in the input " +
-                                          iter.getText(), iter.getIndex());
-            }else if((val.type == MAP)){
-                int index, length;
-
-                if(val.isIndex){
-                    index = val.value;
-                    if(index >= indexes[ONE_UCHAR_MAPPING_INDEX_START] &&
-                             index < indexes[TWO_UCHARS_MAPPING_INDEX_START]){
-                        length = 1;
-                    }else if(index >= indexes[TWO_UCHARS_MAPPING_INDEX_START] &&
-                             index < indexes[THREE_UCHARS_MAPPING_INDEX_START]){
-                        length = 2;
-                    }else if(index >= indexes[THREE_UCHARS_MAPPING_INDEX_START] &&
-                             index < indexes[FOUR_UCHARS_MAPPING_INDEX_START]){
-                        length = 3;
-                    }else{
-                        length = mappingData[index++];
-                    }
-                    /* copy mapping to destination */
-                    dest.append(mappingData,index,length);
-                    continue;
-
-                }else{
-                    ch -= val.value;
-                }
-            }else if(val.type == DELETE){
-                // just consume the codepoint and contine
-                continue;
-            }
-            //copy the source into destination
-            UTF16.append(dest,ch);
-        }
-
-        return dest;
-    }
-
-
-    private StringBuffer normalize(StringBuffer src){
-        /*
-         * Option UNORM_BEFORE_PRI_29:
-         *
-         * IDNA as interpreted by IETF members (see unicode mailing list 2004H1)
-         * requires strict adherence to Unicode 3.2 normalization,
-         * including buggy composition from before fixing Public Review Issue #29.
-         * Note that this results in some valid but nonsensical text to be
-         * either corrupted or rejected, depending on the text.
-         * See http://www.unicode.org/review/resolved-pri.html#pri29
-         * See unorm.cpp and cnormtst.c
-         */
-        return new StringBuffer(
-            Normalizer.normalize(
-                src.toString(),
-                java.text.Normalizer.Form.NFKC,
-                Normalizer.UNICODE_3_2));
-    }
-    /*
-    boolean isLabelSeparator(int ch){
-        int result = getCodePointValue(ch);
-        if( (result & 0x07)  == LABEL_SEPARATOR){
-            return true;
-        }
-        return false;
-    }
-    */
-     /*
-       1) Map -- For each character in the input, check if it has a mapping
-          and, if so, replace it with its mapping.
-
-       2) Normalize -- Possibly normalize the result of step 1 using Unicode
-          normalization.
-
-       3) Prohibit -- Check for any characters that are not allowed in the
-          output.  If any are found, return an error.
-
-       4) Check bidi -- Possibly check for right-to-left characters, and if
-          any are found, make sure that the whole string satisfies the
-          requirements for bidirectional strings.  If the string does not
-          satisfy the requirements for bidirectional strings, return an
-          error.
-          [Unicode3.2] defines several bidirectional categories; each character
-           has one bidirectional category assigned to it.  For the purposes of
-           the requirements below, an "RandALCat character" is a character that
-           has Unicode bidirectional categories "R" or "AL"; an "LCat character"
-           is a character that has Unicode bidirectional category "L".  Note
-
-
-           that there are many characters which fall in neither of the above
-           definitions; Latin digits ( through ) are examples of
-           this because they have bidirectional category "EN".
-
-           In any profile that specifies bidirectional character handling, all
-           three of the following requirements MUST be met:
-
-           1) The characters in section 5.8 MUST be prohibited.
-
-           2) If a string contains any RandALCat character, the string MUST NOT
-              contain any LCat character.
-
-           3) If a string contains any RandALCat character, a RandALCat
-              character MUST be the first character of the string, and a
-              RandALCat character MUST be the last character of the string.
-    */
-    /**
-     * Prepare the input buffer for use in applications with the given profile. This operation maps, normalizes(NFKC),
-     * checks for prohited and BiDi characters in the order defined by RFC 3454
-     * depending on the options specified in the profile.
-     *
-     * @param src           A UCharacterIterator object containing the source string
-     * @param options       A bit set of options:
-     *
-     *  - StringPrep.NONE               Prohibit processing of unassigned code points in the input
-     *
-     *  - StringPrep.ALLOW_UNASSIGNED   Treat the unassigned code points are in the input
-     *                                  as normal Unicode code points.
-     *
-     * @return StringBuffer A StringBuffer containing the output
-     * @throws ParseException
-     * @draft ICU 2.8
-     */
-    public StringBuffer prepare(UCharacterIterator src, int options)
-                        throws ParseException{
-
-        // map
-        StringBuffer mapOut = map(src,options);
-        StringBuffer normOut = mapOut;// initialize
-
-        if(doNFKC){
-            // normalize
-            normOut = normalize(mapOut);
-        }
-
-        int ch;
-        char result;
-        UCharacterIterator iter = UCharacterIterator.getInstance(normOut);
-        Values val = new Values();
-        int direction=UCharacterDirection.CHAR_DIRECTION_COUNT,
-            firstCharDir=UCharacterDirection.CHAR_DIRECTION_COUNT;
-        int rtlPos=-1, ltrPos=-1;
-        boolean rightToLeft=false, leftToRight=false;
-
-        while((ch=iter.nextCodePoint())!= UCharacterIterator.DONE){
-            result = getCodePointValue(ch);
-            getValues(result,val);
-
-            if(val.type == PROHIBITED ){
-                throw new ParseException("A prohibited code point was found in the input" +
-                                         iter.getText(), val.value);
-            }
-
-            direction = UCharacter.getDirection(ch);
-            if(firstCharDir == UCharacterDirection.CHAR_DIRECTION_COUNT){
-                firstCharDir = direction;
-            }
-            if(direction == UCharacterDirection.LEFT_TO_RIGHT){
-                leftToRight = true;
-                ltrPos = iter.getIndex()-1;
-            }
-            if(direction == UCharacterDirection.RIGHT_TO_LEFT || direction == UCharacterDirection.RIGHT_TO_LEFT_ARABIC){
-                rightToLeft = true;
-                rtlPos = iter.getIndex()-1;
-            }
-        }
-        if(checkBiDi == true){
-            // satisfy 2
-            if( leftToRight == true && rightToLeft == true){
-                throw new ParseException("The input does not conform to the rules for BiDi code points." +
-                                         iter.getText(),
-                                         (rtlPos>ltrPos) ? rtlPos : ltrPos);
-             }
-
-            //satisfy 3
-            if( rightToLeft == true &&
-                !((firstCharDir == UCharacterDirection.RIGHT_TO_LEFT || firstCharDir == UCharacterDirection.RIGHT_TO_LEFT_ARABIC) &&
-                (direction == UCharacterDirection.RIGHT_TO_LEFT || direction == UCharacterDirection.RIGHT_TO_LEFT_ARABIC))
-              ){
-                throw new ParseException("The input does not conform to the rules for BiDi code points." +
-                                         iter.getText(),
-                                         (rtlPos>ltrPos) ? rtlPos : ltrPos);
-            }
-        }
-        return normOut;
-
-      }
-}
--- /dev/null	2020-01-10 15:58:04.000000000 -0800
+++ new/src/java.base/share/classes/jdk/internal/icu/text/StringPrep.java	2020-01-10 15:58:04.000000000 -0800
@@ -0,0 +1,485 @@
+/*
+ * Copyright (c) 2005, 2020, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.  Oracle designates this
+ * particular file as subject to the "Classpath" exception as provided
+ * by Oracle in the LICENSE file that accompanied this code.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+/*
+/*
+ *******************************************************************************
+ * Copyright (C) 2003-2004, International Business Machines Corporation and         *
+ * others. All Rights Reserved.                                                *
+ *******************************************************************************
+ */
+//
+// CHANGELOG
+//      2005-05-19 Edward Wang
+//          - copy this file from icu4jsrc_3_2/src/com/ibm/icu/text/StringPrep.java
+//          - move from package com.ibm.icu.text to package sun.net.idn
+//          - use ParseException instead of StringPrepParseException
+//          - change 'Normalizer.getUnicodeVersion()' to 'NormalizerImpl.getUnicodeVersion()'
+//          - remove all @deprecated tag to make compiler happy
+//      2007-08-14 Martin Buchholz
+//          - remove redundant casts
+//
+package jdk.internal.icu.text;
+
+import java.io.BufferedInputStream;
+import java.io.ByteArrayInputStream;
+import java.io.IOException;
+import java.io.InputStream;
+import java.text.ParseException;
+
+import sun.text.Normalizer;
+import jdk.internal.icu.impl.CharTrie;
+import jdk.internal.icu.impl.StringPrepDataReader;
+import jdk.internal.icu.impl.Trie;
+import jdk.internal.icu.lang.UCharacter;
+import jdk.internal.icu.lang.UCharacterDirection;
+import jdk.internal.icu.util.VersionInfo;
+
+/**
+ * StringPrep API implements the StingPrep framework as described by
+ * RFC 3454.
+ * StringPrep prepares Unicode strings for use in network protocols.
+ * Profiles of StingPrep are set of rules and data according to which the
+ * Unicode Strings are prepared. Each profiles contains tables which describe
+ * how a code point should be treated. The tables are broadly classied into
+ * 
    
+ *     
  •  Unassigned Table: Contains code points that are unassigned
+ *          in the Unicode Version supported by StringPrep. Currently
+ *          RFC 3454 supports Unicode 3.2. 
    • 
+ *     
  •  Prohibited Table: Contains code points that are prohibted from
+ *          the output of the StringPrep processing function. 
    • 
+ *     
  •  Mapping Table: Contains code ponts that are deleted from the output or case mapped. 
    • 
+ * 

+ *
+ * The procedure for preparing Unicode strings:
+ * 
    
+ *      
  1.  Map: For each character in the input, check if it has a mapping
+ *           and, if so, replace it with its mapping. 
    2. 
+ *      
  2.  Normalize: Possibly normalize the result of step 1 using Unicode
+ *           normalization. 
    3. 
+ *      
  3.  Prohibit: Check for any characters that are not allowed in the
+ *           output.  If any are found, return an error.
    4. 
+ *      
  4.  Check bidi: Possibly check for right-to-left characters, and if
+ *           any are found, make sure that the whole string satisfies the
+ *           requirements for bidirectional strings.  If the string does not
+ *           satisfy the requirements for bidirectional strings, return an
+ *           error.  
    5. 
+ * 

+ * @author Ram Viswanadha
+ * @draft ICU 2.8
+ */
+public final class StringPrep {
+    /**
+     * Option to prohibit processing of unassigned code points in the input
+     *
+     * @see   #prepare
+     * @draft ICU 2.8
+     */
+    public static final int DEFAULT = 0x0000;
+
+    /**
+     * Option to allow processing of unassigned code points in the input
+     *
+     * @see   #prepare
+     * @draft ICU 2.8
+     */
+    public static final int ALLOW_UNASSIGNED = 0x0001;
+
+    private static final int UNASSIGNED        = 0x0000;
+    private static final int MAP               = 0x0001;
+    private static final int PROHIBITED        = 0x0002;
+    private static final int DELETE            = 0x0003;
+    private static final int TYPE_LIMIT        = 0x0004;
+
+    private static final int NORMALIZATION_ON  = 0x0001;
+    private static final int CHECK_BIDI_ON     = 0x0002;
+
+    private static final int TYPE_THRESHOLD       = 0xFFF0;
+    private static final int MAX_INDEX_VALUE      = 0x3FBF;   /*16139*/
+    private static final int MAX_INDEX_TOP_LENGTH = 0x0003;
+
+    /* indexes[] value names */
+    private static final int INDEX_TRIE_SIZE                  =  0; /* number of bytes in normalization trie */
+    private static final int INDEX_MAPPING_DATA_SIZE          =  1; /* The array that contains the mapping   */
+    private static final int NORM_CORRECTNS_LAST_UNI_VERSION  =  2; /* The index of Unicode version of last entry in NormalizationCorrections.txt */
+    private static final int ONE_UCHAR_MAPPING_INDEX_START    =  3; /* The starting index of 1 UChar mapping index in the mapping data array */
+    private static final int TWO_UCHARS_MAPPING_INDEX_START   =  4; /* The starting index of 2 UChars mapping index in the mapping data array */
+    private static final int THREE_UCHARS_MAPPING_INDEX_START =  5;
+    private static final int FOUR_UCHARS_MAPPING_INDEX_START  =  6;
+    private static final int OPTIONS                          =  7; /* Bit set of options to turn on in the profile */
+    private static final int INDEX_TOP                        = 16;                          /* changing this requires a new formatVersion */
+
+
+    /**
+     * Default buffer size of datafile
+     */
+    private static final int DATA_BUFFER_SIZE = 25000;
+
+    /* Wrappers for Trie implementations */
+    private static final class StringPrepTrieImpl implements Trie.DataManipulate{
+        private CharTrie sprepTrie = null;
+       /**
+        * Called by com.ibm.icu.util.Trie to extract from a lead surrogate's
+        * data the index array offset of the indexes for that lead surrogate.
+        * @param property data value for a surrogate from the trie, including
+        *        the folding offset
+        * @return data offset or 0 if there is no data for the lead surrogate
+        */
+         public int getFoldingOffset(int value){
+            return value;
+        }
+    }
+
+    // CharTrie implementation for reading the trie data
+    private StringPrepTrieImpl sprepTrieImpl;
+    // Indexes read from the data file
+    private int[] indexes;
+    // mapping data read from the data file
+    private char[] mappingData;
+    // format version of the data file
+    private byte[] formatVersion;
+    // the version of Unicode supported by the data file
+    private VersionInfo sprepUniVer;
+    // the Unicode version of last entry in the
+    // NormalizationCorrections.txt file if normalization
+    // is turned on
+    private VersionInfo normCorrVer;
+    // Option to turn on Normalization
+    private boolean doNFKC;
+    // Option to turn on checking for BiDi rules
+    private boolean checkBiDi;
+
+
+    private char getCodePointValue(int ch){
+        return sprepTrieImpl.sprepTrie.getCodePointValue(ch);
+    }
+
+    private static VersionInfo getVersionInfo(int comp){
+        int micro = comp & 0xFF;
+        int milli =(comp >> 8)  & 0xFF;
+        int minor =(comp >> 16) & 0xFF;
+        int major =(comp >> 24) & 0xFF;
+        return VersionInfo.getInstance(major,minor,milli,micro);
+    }
+    private static VersionInfo getVersionInfo(byte[] version){
+        if(version.length != 4){
+            return null;
+        }
+        return VersionInfo.getInstance((int)version[0],(int) version[1],(int) version[2],(int) version[3]);
+    }
+    /**
+     * Creates an StringPrep object after reading the input stream.
+     * The object does not hold a reference to the input steam, so the stream can be
+     * closed after the method returns.
+     *
+     * @param inputStream The stream for reading the StringPrep profile binarySun
+     * @throws IOException
+     * @draft ICU 2.8
+     */
+    public StringPrep(InputStream inputStream) throws IOException{
+
+        BufferedInputStream b = new BufferedInputStream(inputStream,DATA_BUFFER_SIZE);
+
+        StringPrepDataReader reader = new StringPrepDataReader(b);
+
+        // read the indexes
+        indexes = reader.readIndexes(INDEX_TOP);
+
+        byte[] sprepBytes = new byte[indexes[INDEX_TRIE_SIZE]];
+
+
+        //indexes[INDEX_MAPPING_DATA_SIZE] store the size of mappingData in bytes
+        mappingData = new char[indexes[INDEX_MAPPING_DATA_SIZE]/2];
+        // load the rest of the data and initialize the data members
+        reader.read(sprepBytes,mappingData);
+
+        sprepTrieImpl           = new StringPrepTrieImpl();
+        sprepTrieImpl.sprepTrie = new CharTrie( new ByteArrayInputStream(sprepBytes),sprepTrieImpl  );
+
+        // get the data format version
+        formatVersion = reader.getDataFormatVersion();
+
+        // get the options
+        doNFKC            = ((indexes[OPTIONS] & NORMALIZATION_ON) > 0);
+        checkBiDi         = ((indexes[OPTIONS] & CHECK_BIDI_ON) > 0);
+        sprepUniVer   = getVersionInfo(reader.getUnicodeVersion());
+        normCorrVer   = getVersionInfo(indexes[NORM_CORRECTNS_LAST_UNI_VERSION]);
+        VersionInfo normUniVer = UCharacter.getUnicodeVersion();
+        if(normUniVer.compareTo(sprepUniVer) < 0 && /* the Unicode version of SPREP file must be less than the Unicode Vesion of the normalization data */
+           normUniVer.compareTo(normCorrVer) < 0 && /* the Unicode version of the NormalizationCorrections.txt file should be less than the Unicode Vesion of the normalization data */
+           ((indexes[OPTIONS] & NORMALIZATION_ON) > 0) /* normalization turned on*/
+           ){
+            throw new IOException("Normalization Correction version not supported");
+        }
+        b.close();
+    }
+
+    private static final class Values{
+        boolean isIndex;
+        int value;
+        int type;
+        public void reset(){
+            isIndex = false;
+            value = 0;
+            type = -1;
+        }
+    }
+
+    private static final void getValues(char trieWord,Values values){
+        values.reset();
+        if(trieWord == 0){
+            /*
+             * Initial value stored in the mapping table
+             * just return TYPE_LIMIT .. so that
+             * the source codepoint is copied to the destination
+             */
+            values.type = TYPE_LIMIT;
+        }else if(trieWord >= TYPE_THRESHOLD){
+            values.type = (trieWord - TYPE_THRESHOLD);
+        }else{
+            /* get the type */
+            values.type = MAP;
+            /* ascertain if the value is index or delta */
+            if((trieWord & 0x02)>0){
+                values.isIndex = true;
+                values.value = trieWord  >> 2; //mask off the lower 2 bits and shift
+
+            }else{
+                values.isIndex = false;
+                values.value = (trieWord<<16)>>16;
+                values.value =  (values.value >> 2);
+
+            }
+
+            if((trieWord>>2) == MAX_INDEX_VALUE){
+                values.type = DELETE;
+                values.isIndex = false;
+                values.value = 0;
+            }
+        }
+    }
+
+
+
+    private StringBuffer map( UCharacterIterator iter, int options)
+                            throws ParseException {
+
+        Values val = new Values();
+        char result = 0;
+        int ch  = UCharacterIterator.DONE;
+        StringBuffer dest = new StringBuffer();
+        boolean allowUnassigned = ((options & ALLOW_UNASSIGNED)>0);
+
+        while((ch=iter.nextCodePoint())!= UCharacterIterator.DONE){
+
+            result = getCodePointValue(ch);
+            getValues(result,val);
+
+            // check if the source codepoint is unassigned
+            if(val.type == UNASSIGNED && allowUnassigned == false){
+                 throw new ParseException("An unassigned code point was found in the input " +
+                                          iter.getText(), iter.getIndex());
+            }else if((val.type == MAP)){
+                int index, length;
+
+                if(val.isIndex){
+                    index = val.value;
+                    if(index >= indexes[ONE_UCHAR_MAPPING_INDEX_START] &&
+                             index < indexes[TWO_UCHARS_MAPPING_INDEX_START]){
+                        length = 1;
+                    }else if(index >= indexes[TWO_UCHARS_MAPPING_INDEX_START] &&
+                             index < indexes[THREE_UCHARS_MAPPING_INDEX_START]){
+                        length = 2;
+                    }else if(index >= indexes[THREE_UCHARS_MAPPING_INDEX_START] &&
+                             index < indexes[FOUR_UCHARS_MAPPING_INDEX_START]){
+                        length = 3;
+                    }else{
+                        length = mappingData[index++];
+                    }
+                    /* copy mapping to destination */
+                    dest.append(mappingData,index,length);
+                    continue;
+
+                }else{
+                    ch -= val.value;
+                }
+            }else if(val.type == DELETE){
+                // just consume the codepoint and contine
+                continue;
+            }
+            //copy the source into destination
+            UTF16.append(dest,ch);
+        }
+
+        return dest;
+    }
+
+
+    private StringBuffer normalize(StringBuffer src){
+        /*
+         * Option UNORM_BEFORE_PRI_29:
+         *
+         * IDNA as interpreted by IETF members (see unicode mailing list 2004H1)
+         * requires strict adherence to Unicode 3.2 normalization,
+         * including buggy composition from before fixing Public Review Issue #29.
+         * Note that this results in some valid but nonsensical text to be
+         * either corrupted or rejected, depending on the text.
+         * See http://www.unicode.org/review/resolved-pri.html#pri29
+         * See unorm.cpp and cnormtst.c
+         */
+        return new StringBuffer(
+            Normalizer.normalize(
+                src.toString(),
+                java.text.Normalizer.Form.NFKC,
+                Normalizer.UNICODE_3_2));
+    }
+    /*
+    boolean isLabelSeparator(int ch){
+        int result = getCodePointValue(ch);
+        if( (result & 0x07)  == LABEL_SEPARATOR){
+            return true;
+        }
+        return false;
+    }
+    */
+     /*
+       1) Map -- For each character in the input, check if it has a mapping
+          and, if so, replace it with its mapping.
+
+       2) Normalize -- Possibly normalize the result of step 1 using Unicode
+          normalization.
+
+       3) Prohibit -- Check for any characters that are not allowed in the
+          output.  If any are found, return an error.
+
+       4) Check bidi -- Possibly check for right-to-left characters, and if
+          any are found, make sure that the whole string satisfies the
+          requirements for bidirectional strings.  If the string does not
+          satisfy the requirements for bidirectional strings, return an
+          error.
+          [Unicode3.2] defines several bidirectional categories; each character
+           has one bidirectional category assigned to it.  For the purposes of
+           the requirements below, an "RandALCat character" is a character that
+           has Unicode bidirectional categories "R" or "AL"; an "LCat character"
+           is a character that has Unicode bidirectional category "L".  Note
+
+
+           that there are many characters which fall in neither of the above
+           definitions; Latin digits ( through ) are examples of
+           this because they have bidirectional category "EN".
+
+           In any profile that specifies bidirectional character handling, all
+           three of the following requirements MUST be met:
+
+           1) The characters in section 5.8 MUST be prohibited.
+
+           2) If a string contains any RandALCat character, the string MUST NOT
+              contain any LCat character.
+
+           3) If a string contains any RandALCat character, a RandALCat
+              character MUST be the first character of the string, and a
+              RandALCat character MUST be the last character of the string.
+    */
+    /**
+     * Prepare the input buffer for use in applications with the given profile. This operation maps, normalizes(NFKC),
+     * checks for prohited and BiDi characters in the order defined by RFC 3454
+     * depending on the options specified in the profile.
+     *
+     * @param src           A UCharacterIterator object containing the source string
+     * @param options       A bit set of options:
+     *
+     *  - StringPrep.NONE               Prohibit processing of unassigned code points in the input
+     *
+     *  - StringPrep.ALLOW_UNASSIGNED   Treat the unassigned code points are in the input
+     *                                  as normal Unicode code points.
+     *
+     * @return StringBuffer A StringBuffer containing the output
+     * @throws ParseException
+     * @draft ICU 2.8
+     */
+    public StringBuffer prepare(UCharacterIterator src, int options)
+                        throws ParseException{
+
+        // map
+        StringBuffer mapOut = map(src,options);
+        StringBuffer normOut = mapOut;// initialize
+
+        if(doNFKC){
+            // normalize
+            normOut = normalize(mapOut);
+        }
+
+        int ch;
+        char result;
+        UCharacterIterator iter = UCharacterIterator.getInstance(normOut);
+        Values val = new Values();
+        int direction=UCharacterDirection.CHAR_DIRECTION_COUNT,
+            firstCharDir=UCharacterDirection.CHAR_DIRECTION_COUNT;
+        int rtlPos=-1, ltrPos=-1;
+        boolean rightToLeft=false, leftToRight=false;
+
+        while((ch=iter.nextCodePoint())!= UCharacterIterator.DONE){
+            result = getCodePointValue(ch);
+            getValues(result,val);
+
+            if(val.type == PROHIBITED ){
+                throw new ParseException("A prohibited code point was found in the input" +
+                                         iter.getText(), val.value);
+            }
+
+            direction = UCharacter.getDirection(ch);
+            if(firstCharDir == UCharacterDirection.CHAR_DIRECTION_COUNT){
+                firstCharDir = direction;
+            }
+            if(direction == UCharacterDirection.LEFT_TO_RIGHT){
+                leftToRight = true;
+                ltrPos = iter.getIndex()-1;
+            }
+            if(direction == UCharacterDirection.RIGHT_TO_LEFT || direction == UCharacterDirection.RIGHT_TO_LEFT_ARABIC){
+                rightToLeft = true;
+                rtlPos = iter.getIndex()-1;
+            }
+        }
+        if(checkBiDi == true){
+            // satisfy 2
+            if( leftToRight == true && rightToLeft == true){
+                throw new ParseException("The input does not conform to the rules for BiDi code points." +
+                                         iter.getText(),
+                                         (rtlPos>ltrPos) ? rtlPos : ltrPos);
+             }
+
+            //satisfy 3
+            if( rightToLeft == true &&
+                !((firstCharDir == UCharacterDirection.RIGHT_TO_LEFT || firstCharDir == UCharacterDirection.RIGHT_TO_LEFT_ARABIC) &&
+                (direction == UCharacterDirection.RIGHT_TO_LEFT || direction == UCharacterDirection.RIGHT_TO_LEFT_ARABIC))
+              ){
+                throw new ParseException("The input does not conform to the rules for BiDi code points." +
+                                         iter.getText(),
+                                         (rtlPos>ltrPos) ? rtlPos : ltrPos);
+            }
+        }
+        return normOut;
+
+      }
+}
--- old/src/java.base/share/classes/sun/text/normalizer/UCharacterIterator.java	2020-01-10 15:58:06.000000000 -0800
+++ /dev/null	2020-01-10 15:58:06.000000000 -0800
@@ -1,313 +0,0 @@
-/*
- * Copyright (c) 2003, 2015, Oracle and/or its affiliates. All rights reserved.
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.  Oracle designates this
- * particular file as subject to the "Classpath" exception as provided
- * by Oracle in the LICENSE file that accompanied this code.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- */
-
-/*
- *******************************************************************************
- * Copyright (C) 1996-2014, International Business Machines Corporation and    *
- * others. All Rights Reserved.                                                *
- *******************************************************************************
- */
-
-package sun.text.normalizer;
-
-import java.text.CharacterIterator;
-
-/**
- * Abstract class that defines an API for iteration on text objects.This is an
- * interface for forward and backward iteration and random access into a text
- * object. Forward iteration is done with post-increment and backward iteration
- * is done with pre-decrement semantics, while the
- * java.text.CharacterIterator interface methods provided forward
- * iteration with "pre-increment" and backward iteration with pre-decrement
- * semantics. This API is more efficient for forward iteration over code points.
- * The other major difference is that this API can do both code unit and code point
- * iteration, java.text.CharacterIterator can only iterate over
- * code units and is limited to BMP (0 - 0xFFFF)
- * @author Ram
- * @stable ICU 2.4
- */
-public abstract class UCharacterIterator
-                      implements Cloneable {
-
-    /**
-     * Protected default constructor for the subclasses
-     * @stable ICU 2.4
-     */
-    protected UCharacterIterator(){
-    }
-
-    /**
-     * Indicator that we have reached the ends of the UTF16 text.
-     * Moved from UForwardCharacterIterator.java
-     * @stable ICU 2.4
-     */
-    public static final int DONE = -1;
-
-    // static final methods ----------------------------------------------------
-
-    /**
-     * Returns a UCharacterIterator object given a
-     * source string.
-     * @param source a string
-     * @return UCharacterIterator object
-     * @exception IllegalArgumentException if the argument is null
-     * @stable ICU 2.4
-     */
-    public static final UCharacterIterator getInstance(String source){
-        return new ReplaceableUCharacterIterator(source);
-    }
-
-    /**
-     * Returns a UCharacterIterator object given a
-     * source StringBuffer.
-     * @param source an string buffer of UTF-16 code units
-     * @return UCharacterIterator object
-     * @exception IllegalArgumentException if the argument is null
-     * @stable ICU 2.4
-     */
-    public static final UCharacterIterator getInstance(StringBuffer source){
-        return new ReplaceableUCharacterIterator(source);
-    }
-
-   /**
-     * Returns a UCharacterIterator object given a
-     * CharacterIterator.
-     * @param source a valid CharacterIterator object.
-     * @return UCharacterIterator object
-     * @exception IllegalArgumentException if the argument is null
-     * @stable ICU 2.4
-     */
-    public static final UCharacterIterator getInstance(CharacterIterator source){
-        return new CharacterIteratorWrapper(source);
-    }
-
-    // public methods ----------------------------------------------------------
-
-    /**
-     * Returns the length of the text
-     * @return length of the text
-     * @stable ICU 2.4
-     */
-    public abstract int getLength();
-
-    /**
-     * Gets the current index in text.
-     * @return current index in text.
-     * @stable ICU 2.4
-     */
-    public abstract int getIndex();
-
-    /**
-     * Returns the UTF16 code unit at index, and increments to the next
-     * code unit (post-increment semantics).  If index is out of
-     * range, DONE is returned, and the iterator is reset to the limit
-     * of the text.
-     * @return the next UTF16 code unit, or DONE if the index is at the limit
-     *         of the text.
-     * @stable ICU 2.4
-     */
-    public abstract int next();
-
-    /**
-     * Returns the code point at index, and increments to the next code
-     * point (post-increment semantics).  If index does not point to a
-     * valid surrogate pair, the behavior is the same as
-     * next().  Otherwise the iterator is incremented past
-     * the surrogate pair, and the code point represented by the pair
-     * is returned.
-     * @return the next codepoint in text, or DONE if the index is at
-     *         the limit of the text.
-     * @stable ICU 2.4
-     */
-    public int nextCodePoint(){
-        int ch1 = next();
-        if(UTF16.isLeadSurrogate((char)ch1)){
-            int ch2 = next();
-            if(UTF16.isTrailSurrogate((char)ch2)){
-                return UCharacterProperty.getRawSupplementary((char)ch1,
-                                                              (char)ch2);
-            }else if (ch2 != DONE) {
-                // unmatched surrogate so back out
-                previous();
-            }
-        }
-        return ch1;
-    }
-
-    /**
-     * Decrement to the position of the previous code unit in the
-     * text, and return it (pre-decrement semantics).  If the
-     * resulting index is less than 0, the index is reset to 0 and
-     * DONE is returned.
-     * @return the previous code unit in the text, or DONE if the new
-     *         index is before the start of the text.
-     * @stable ICU 2.4
-     */
-    public abstract int previous();
-
-
-    /**
-     * Retreat to the start of the previous code point in the text,
-     * and return it (pre-decrement semantics).  If the index is not
-     * preceeded by a valid surrogate pair, the behavior is the same
-     * as previous().  Otherwise the iterator is
-     * decremented to the start of the surrogate pair, and the code
-     * point represented by the pair is returned.
-     * @return the previous code point in the text, or DONE if the new
-     *         index is before the start of the text.
-     * @stable ICU 2.4
-     */
-    public int previousCodePoint(){
-        int ch1 = previous();
-        if(UTF16.isTrailSurrogate((char)ch1)){
-            int ch2 = previous();
-            if(UTF16.isLeadSurrogate((char)ch2)){
-                return UCharacterProperty.getRawSupplementary((char)ch2,
-                                                              (char)ch1);
-            }else if (ch2 != DONE) {
-                //unmatched trail surrogate so back out
-                next();
-            }
-        }
-        return ch1;
-    }
-
-    /**
-     * Sets the index to the specified index in the text.
-     * @param index the index within the text.
-     * @exception IndexOutOfBoundsException is thrown if an invalid index is
-     *            supplied
-     * @stable ICU 2.4
-     */
-    public abstract void setIndex(int index);
-
-    /**
-     * Sets the current index to the start.
-     * @stable ICU 2.4
-     */
-    public void setToStart() {
-        setIndex(0);
-    }
-
-    /**
-     * Fills the buffer with the underlying text storage of the iterator
-     * If the buffer capacity is not enough a exception is thrown. The capacity
-     * of the fill in buffer should at least be equal to length of text in the
-     * iterator obtained by calling getLength().
-     * Usage:
-     *
-     * 
{@code
-     *         UChacterIterator iter = new UCharacterIterator.getInstance(text);
-     *         char[] buf = new char[iter.getLength()];
-     *         iter.getText(buf);
-     *
-     *         OR
-     *         char[] buf= new char[1];
-     *         int len = 0;
-     *         for(;;){
-     *             try{
-     *                 len = iter.getText(buf);
-     *                 break;
-     *             }catch(IndexOutOfBoundsException e){
-     *                 buf = new char[iter.getLength()];
-     *             }
-     *         }
-     * }

-     *
-     * @param fillIn an array of chars to fill with the underlying UTF-16 code
-     *         units.
-     * @param offset the position within the array to start putting the data.
-     * @return the number of code units added to fillIn, as a convenience
-     * @exception IndexOutOfBoundsException exception if there is not enough
-     *            room after offset in the array, or if offset < 0.
-     * @stable ICU 2.4
-     */
-    public abstract int getText(char[] fillIn, int offset);
-
-    /**
-     * Convenience override for getText(char[], int) that provides
-     * an offset of 0.
-     * @param fillIn an array of chars to fill with the underlying UTF-16 code
-     *         units.
-     * @return the number of code units added to fillIn, as a convenience
-     * @exception IndexOutOfBoundsException exception if there is not enough
-     *            room in the array.
-     * @stable ICU 2.4
-     */
-    public final int getText(char[] fillIn) {
-        return getText(fillIn, 0);
-    }
-
-    /**
-     * Convenience method for returning the underlying text storage as a string
-     * @return the underlying text storage in the iterator as a string
-     * @stable ICU 2.4
-     */
-    public String getText() {
-        char[] text = new char[getLength()];
-        getText(text);
-        return new String(text);
-    }
-
-    /**
-     * Moves the current position by the number of code points
-     * specified, either forward or backward depending on the sign of
-     * delta (positive or negative respectively). If the current index
-     * is at a trail surrogate then the first adjustment is by code
-     * unit, and the remaining adjustments are by code points.  If the
-     * resulting index would be less than zero, the index is set to
-     * zero, and if the resulting index would be greater than limit,
-     * the index is set to limit.
-     * @param delta the number of code units to move the current index.
-     * @return the new index
-     * @exception IndexOutOfBoundsException is thrown if an invalid delta is
-     *            supplied
-     * @stable ICU 2.4
-     *
-     */
-    public int moveCodePointIndex(int delta){
-        if(delta>0){
-            while(delta>0 && nextCodePoint() != DONE){delta--;}
-        }else{
-            while(delta<0 && previousCodePoint() != DONE){delta++;}
-        }
-        if(delta!=0){
-            throw new IndexOutOfBoundsException();
-        }
-
-        return getIndex();
-    }
-
-    /**
-     * Creates a copy of this iterator, independent from other iterators.
-     * If it is not possible to clone the iterator, returns null.
-     * @return copy of this iterator
-     * @stable ICU 2.4
-     */
-    public Object clone() throws CloneNotSupportedException{
-        return super.clone();
-    }
-
-}
--- /dev/null	2020-01-10 15:58:06.000000000 -0800
+++ new/src/java.base/share/classes/jdk/internal/icu/text/UCharacterIterator.java	2020-01-10 15:58:05.000000000 -0800
@@ -0,0 +1,317 @@
+/*
+ * Copyright (c) 2003, 2020, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.  Oracle designates this
+ * particular file as subject to the "Classpath" exception as provided
+ * by Oracle in the LICENSE file that accompanied this code.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+/*
+ *******************************************************************************
+ * Copyright (C) 1996-2014, International Business Machines Corporation and    *
+ * others. All Rights Reserved.                                                *
+ *******************************************************************************
+ */
+
+package jdk.internal.icu.text;
+
+import jdk.internal.icu.impl.CharacterIteratorWrapper;
+import jdk.internal.icu.impl.ReplaceableUCharacterIterator;
+import jdk.internal.icu.impl.UCharacterProperty;
+
+import java.text.CharacterIterator;
+
+/**
+ * Abstract class that defines an API for iteration on text objects.This is an
+ * interface for forward and backward iteration and random access into a text
+ * object. Forward iteration is done with post-increment and backward iteration
+ * is done with pre-decrement semantics, while the
+ * java.text.CharacterIterator interface methods provided forward
+ * iteration with "pre-increment" and backward iteration with pre-decrement
+ * semantics. This API is more efficient for forward iteration over code points.
+ * The other major difference is that this API can do both code unit and code point
+ * iteration, java.text.CharacterIterator can only iterate over
+ * code units and is limited to BMP (0 - 0xFFFF)
+ * @author Ram
+ * @stable ICU 2.4
+ */
+public abstract class UCharacterIterator
+                      implements Cloneable {
+
+    /**
+     * Protected default constructor for the subclasses
+     * @stable ICU 2.4
+     */
+    protected UCharacterIterator(){
+    }
+
+    /**
+     * Indicator that we have reached the ends of the UTF16 text.
+     * Moved from UForwardCharacterIterator.java
+     * @stable ICU 2.4
+     */
+    public static final int DONE = -1;
+
+    // static final methods ----------------------------------------------------
+
+    /**
+     * Returns a UCharacterIterator object given a
+     * source string.
+     * @param source a string
+     * @return UCharacterIterator object
+     * @exception IllegalArgumentException if the argument is null
+     * @stable ICU 2.4
+     */
+    public static final UCharacterIterator getInstance(String source){
+        return new ReplaceableUCharacterIterator(source);
+    }
+
+    /**
+     * Returns a UCharacterIterator object given a
+     * source StringBuffer.
+     * @param source an string buffer of UTF-16 code units
+     * @return UCharacterIterator object
+     * @exception IllegalArgumentException if the argument is null
+     * @stable ICU 2.4
+     */
+    public static final UCharacterIterator getInstance(StringBuffer source){
+        return new ReplaceableUCharacterIterator(source);
+    }
+
+   /**
+     * Returns a UCharacterIterator object given a
+     * CharacterIterator.
+     * @param source a valid CharacterIterator object.
+     * @return UCharacterIterator object
+     * @exception IllegalArgumentException if the argument is null
+     * @stable ICU 2.4
+     */
+    public static final UCharacterIterator getInstance(CharacterIterator source){
+        return new CharacterIteratorWrapper(source);
+    }
+
+    // public methods ----------------------------------------------------------
+
+    /**
+     * Returns the length of the text
+     * @return length of the text
+     * @stable ICU 2.4
+     */
+    public abstract int getLength();
+
+    /**
+     * Gets the current index in text.
+     * @return current index in text.
+     * @stable ICU 2.4
+     */
+    public abstract int getIndex();
+
+    /**
+     * Returns the UTF16 code unit at index, and increments to the next
+     * code unit (post-increment semantics).  If index is out of
+     * range, DONE is returned, and the iterator is reset to the limit
+     * of the text.
+     * @return the next UTF16 code unit, or DONE if the index is at the limit
+     *         of the text.
+     * @stable ICU 2.4
+     */
+    public abstract int next();
+
+    /**
+     * Returns the code point at index, and increments to the next code
+     * point (post-increment semantics).  If index does not point to a
+     * valid surrogate pair, the behavior is the same as
+     * next().  Otherwise the iterator is incremented past
+     * the surrogate pair, and the code point represented by the pair
+     * is returned.
+     * @return the next codepoint in text, or DONE if the index is at
+     *         the limit of the text.
+     * @stable ICU 2.4
+     */
+    public int nextCodePoint(){
+        int ch1 = next();
+        if(UTF16.isLeadSurrogate((char)ch1)){
+            int ch2 = next();
+            if(UTF16.isTrailSurrogate((char)ch2)){
+                return UCharacterProperty.getRawSupplementary((char)ch1,
+                                                              (char)ch2);
+            }else if (ch2 != DONE) {
+                // unmatched surrogate so back out
+                previous();
+            }
+        }
+        return ch1;
+    }
+
+    /**
+     * Decrement to the position of the previous code unit in the
+     * text, and return it (pre-decrement semantics).  If the
+     * resulting index is less than 0, the index is reset to 0 and
+     * DONE is returned.
+     * @return the previous code unit in the text, or DONE if the new
+     *         index is before the start of the text.
+     * @stable ICU 2.4
+     */
+    public abstract int previous();
+
+
+    /**
+     * Retreat to the start of the previous code point in the text,
+     * and return it (pre-decrement semantics).  If the index is not
+     * preceeded by a valid surrogate pair, the behavior is the same
+     * as previous().  Otherwise the iterator is
+     * decremented to the start of the surrogate pair, and the code
+     * point represented by the pair is returned.
+     * @return the previous code point in the text, or DONE if the new
+     *         index is before the start of the text.
+     * @stable ICU 2.4
+     */
+    public int previousCodePoint(){
+        int ch1 = previous();
+        if(UTF16.isTrailSurrogate((char)ch1)){
+            int ch2 = previous();
+            if(UTF16.isLeadSurrogate((char)ch2)){
+                return UCharacterProperty.getRawSupplementary((char)ch2,
+                                                              (char)ch1);
+            }else if (ch2 != DONE) {
+                //unmatched trail surrogate so back out
+                next();
+            }
+        }
+        return ch1;
+    }
+
+    /**
+     * Sets the index to the specified index in the text.
+     * @param index the index within the text.
+     * @exception IndexOutOfBoundsException is thrown if an invalid index is
+     *            supplied
+     * @stable ICU 2.4
+     */
+    public abstract void setIndex(int index);
+
+    /**
+     * Sets the current index to the start.
+     * @stable ICU 2.4
+     */
+    public void setToStart() {
+        setIndex(0);
+    }
+
+    /**
+     * Fills the buffer with the underlying text storage of the iterator
+     * If the buffer capacity is not enough a exception is thrown. The capacity
+     * of the fill in buffer should at least be equal to length of text in the
+     * iterator obtained by calling getLength().
+     * Usage:
+     *
+     * 
{@code
+     *         UChacterIterator iter = new UCharacterIterator.getInstance(text);
+     *         char[] buf = new char[iter.getLength()];
+     *         iter.getText(buf);
+     *
+     *         OR
+     *         char[] buf= new char[1];
+     *         int len = 0;
+     *         for(;;){
+     *             try{
+     *                 len = iter.getText(buf);
+     *                 break;
+     *             }catch(IndexOutOfBoundsException e){
+     *                 buf = new char[iter.getLength()];
+     *             }
+     *         }
+     * }

+     *
+     * @param fillIn an array of chars to fill with the underlying UTF-16 code
+     *         units.
+     * @param offset the position within the array to start putting the data.
+     * @return the number of code units added to fillIn, as a convenience
+     * @exception IndexOutOfBoundsException exception if there is not enough
+     *            room after offset in the array, or if offset < 0.
+     * @stable ICU 2.4
+     */
+    public abstract int getText(char[] fillIn, int offset);
+
+    /**
+     * Convenience override for getText(char[], int) that provides
+     * an offset of 0.
+     * @param fillIn an array of chars to fill with the underlying UTF-16 code
+     *         units.
+     * @return the number of code units added to fillIn, as a convenience
+     * @exception IndexOutOfBoundsException exception if there is not enough
+     *            room in the array.
+     * @stable ICU 2.4
+     */
+    public final int getText(char[] fillIn) {
+        return getText(fillIn, 0);
+    }
+
+    /**
+     * Convenience method for returning the underlying text storage as a string
+     * @return the underlying text storage in the iterator as a string
+     * @stable ICU 2.4
+     */
+    public String getText() {
+        char[] text = new char[getLength()];
+        getText(text);
+        return new String(text);
+    }
+
+    /**
+     * Moves the current position by the number of code points
+     * specified, either forward or backward depending on the sign of
+     * delta (positive or negative respectively). If the current index
+     * is at a trail surrogate then the first adjustment is by code
+     * unit, and the remaining adjustments are by code points.  If the
+     * resulting index would be less than zero, the index is set to
+     * zero, and if the resulting index would be greater than limit,
+     * the index is set to limit.
+     * @param delta the number of code units to move the current index.
+     * @return the new index
+     * @exception IndexOutOfBoundsException is thrown if an invalid delta is
+     *            supplied
+     * @stable ICU 2.4
+     *
+     */
+    public int moveCodePointIndex(int delta){
+        if(delta>0){
+            while(delta>0 && nextCodePoint() != DONE){delta--;}
+        }else{
+            while(delta<0 && previousCodePoint() != DONE){delta++;}
+        }
+        if(delta!=0){
+            throw new IndexOutOfBoundsException();
+        }
+
+        return getIndex();
+    }
+
+    /**
+     * Creates a copy of this iterator, independent from other iterators.
+     * If it is not possible to clone the iterator, returns null.
+     * @return copy of this iterator
+     * @stable ICU 2.4
+     */
+    public Object clone() throws CloneNotSupportedException{
+        return super.clone();
+    }
+
+}
--- old/src/java.base/share/classes/sun/text/normalizer/UTF16.java	2020-01-10 15:58:07.000000000 -0800
+++ /dev/null	2020-01-10 15:58:07.000000000 -0800
@@ -1,616 +0,0 @@
-/*
- * Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved.
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.  Oracle designates this
- * particular file as subject to the "Classpath" exception as provided
- * by Oracle in the LICENSE file that accompanied this code.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- */
-/**
- *******************************************************************************
- * Copyright (C) 1996-2014, International Business Machines Corporation and
- * others. All Rights Reserved.
- *******************************************************************************
- */
-
-package sun.text.normalizer;
-
-/**
- * 
Standalone utility class providing UTF16 character conversions and
- * indexing conversions.
- * 
Code that uses strings alone rarely need modification.
- * By design, UTF-16 does not allow overlap, so searching for strings is a safe
- * operation. Similarly, concatenation is always safe. Substringing is safe if
- * the start and end are both on UTF-32 boundaries. In normal code, the values
- * for start and end are on those boundaries, since they arose from operations
- * like searching. If not, the nearest UTF-32 boundaries can be determined
- * using bounds().
- * Examples:
- * 
The following examples illustrate use of some of these methods.
- * 
{@code
- * // iteration forwards: Original
- * for (int i = 0; i < s.length(); ++i) {
- *     char ch = s.charAt(i);
- *     doSomethingWith(ch);
- * }
- *
- * // iteration forwards: Changes for UTF-32
- * int ch;
- * for (int i = 0; i < s.length(); i += UTF16.getCharCount(ch)) {
- *     ch = UTF16.charAt(s, i);
- *     doSomethingWith(ch);
- * }
- *
- * // iteration backwards: Original
- * for (int i = s.length() - 1; i >= 0; --i) {
- *     char ch = s.charAt(i);
- *     doSomethingWith(ch);
- * }
- *
- * // iteration backwards: Changes for UTF-32
- * int ch;
- * for (int i = s.length() - 1; i > 0; i -= UTF16.getCharCount(ch)) {
- *     ch = UTF16.charAt(s, i);
- *     doSomethingWith(ch);
- * }
- * }

- * Notes:
- * 
    
- *   
  • 
- *   Naming: For clarity, High and Low surrogates are called
- *   Lead and Trail in the API, which gives a better
- *   sense of their ordering in a string. offset16 and
- *   offset32 are used to distinguish offsets to UTF-16
- *   boundaries vs offsets to UTF-32 boundaries. int char32 is
- *   used to contain UTF-32 characters, as opposed to char16,
- *   which is a UTF-16 code unit.
- *   
    • 
- *   
  • 
- *   Roundtripping Offsets: You can always roundtrip from a
- *   UTF-32 offset to a UTF-16 offset and back. Because of the difference in
- *   structure, you can roundtrip from a UTF-16 offset to a UTF-32 offset and
- *   back if and only if bounds(string, offset16) != TRAIL.
- *   
    • 
- *   
  • 
- *   Exceptions: The error checking will throw an exception
- *   if indices are out of bounds. Other than that, all methods will
- *   behave reasonably, even if unmatched surrogates or out-of-bounds UTF-32
- *   values are present. UCharacter.isLegal() can be used to check
- *   for validity if desired.
- *   
    • 
- *   
  • 
- *   Unmatched Surrogates: If the string contains unmatched
- *   surrogates, then these are counted as one UTF-32 value. This matches
- *   their iteration behavior, which is vital. It also matches common display
- *   practice as missing glyphs (see the Unicode Standard Section 5.4, 5.5).
- *   
    • 
- *   
  • 
- *   Optimization: The method implementations may need
- *   optimization if the compiler doesn't fold static final methods. Since
- *   surrogate pairs will form an exceeding small percentage of all the text
- *   in the world, the singleton case should always be optimized for.
- *   
    • 
- * 

- * @author Mark Davis, with help from Markus Scherer
- * @stable ICU 2.1
- */
-
-public final class UTF16
-{
-    // public variables ---------------------------------------------------
-
-    /**
-     * The lowest Unicode code point value.
-     * @stable ICU 2.1
-     */
-    public static final int CODEPOINT_MIN_VALUE = 0;
-    /**
-     * The highest Unicode code point value (scalar value) according to the
-     * Unicode Standard.
-     * @stable ICU 2.1
-     */
-    public static final int CODEPOINT_MAX_VALUE = 0x10ffff;
-    /**
-     * The minimum value for Supplementary code points
-     * @stable ICU 2.1
-     */
-    public static final int SUPPLEMENTARY_MIN_VALUE = 0x10000;
-    /**
-     * Lead surrogate minimum value
-     * @stable ICU 2.1
-     */
-    public static final int LEAD_SURROGATE_MIN_VALUE = 0xD800;
-    /**
-     * Trail surrogate minimum value
-     * @stable ICU 2.1
-     */
-    public static final int TRAIL_SURROGATE_MIN_VALUE = 0xDC00;
-    /**
-     * Lead surrogate maximum value
-     * @stable ICU 2.1
-     */
-    public static final int LEAD_SURROGATE_MAX_VALUE = 0xDBFF;
-    /**
-     * Trail surrogate maximum value
-     * @stable ICU 2.1
-     */
-    public static final int TRAIL_SURROGATE_MAX_VALUE = 0xDFFF;
-    /**
-     * Surrogate minimum value
-     * @stable ICU 2.1
-     */
-    public static final int SURROGATE_MIN_VALUE = LEAD_SURROGATE_MIN_VALUE;
-    /**
-     * Lead surrogate bitmask
-     */
-    private static final int LEAD_SURROGATE_BITMASK = 0xFFFFFC00;
-    /**
-     * Trail surrogate bitmask
-     */
-    private static final int TRAIL_SURROGATE_BITMASK = 0xFFFFFC00;
-    /**
-     * Surrogate bitmask
-     */
-    private static final int SURROGATE_BITMASK = 0xFFFFF800;
-    /**
-     * Lead surrogate bits
-     */
-    private static final int LEAD_SURROGATE_BITS = 0xD800;
-    /**
-     * Trail surrogate bits
-     */
-    private static final int TRAIL_SURROGATE_BITS = 0xDC00;
-    /**
-     * Surrogate bits
-     */
-    private static final int SURROGATE_BITS = 0xD800;
-
-    // constructor --------------------------------------------------------
-
-    // /CLOVER:OFF
-    /**
-     * Prevent instance from being created.
-     */
-    private UTF16() {
-    }
-
-    // /CLOVER:ON
-    // public method ------------------------------------------------------
-
-    /**
-     * Extract a single UTF-32 value from a string.
-     * Used when iterating forwards or backwards (with
-     * UTF16.getCharCount(), as well as random access. If a
-     * validity check is required, use
-     * 
-     * UCharacter.isLegal() on the return value.
-     * If the char retrieved is part of a surrogate pair, its supplementary
-     * character will be returned. If a complete supplementary character is
-     * not found the incomplete character will be returned
-     * @param source array of UTF-16 chars
-     * @param offset16 UTF-16 offset to the start of the character.
-     * @return UTF-32 value for the UTF-32 value that contains the char at
-     *         offset16. The boundaries of that codepoint are the same as in
-     *         bounds32().
-     * @exception IndexOutOfBoundsException thrown if offset16 is out of
-     *            bounds.
-     * @stable ICU 2.1
-     */
-    public static int charAt(String source, int offset16) {
-        char single = source.charAt(offset16);
-        if (single < LEAD_SURROGATE_MIN_VALUE) {
-            return single;
-        }
-        return _charAt(source, offset16, single);
-    }
-
-    private static int _charAt(String source, int offset16, char single) {
-        if (single > TRAIL_SURROGATE_MAX_VALUE) {
-            return single;
-        }
-
-        // Convert the UTF-16 surrogate pair if necessary.
-        // For simplicity in usage, and because the frequency of pairs is
-        // low, look both directions.
-
-        if (single <= LEAD_SURROGATE_MAX_VALUE) {
-            ++offset16;
-            if (source.length() != offset16) {
-                char trail = source.charAt(offset16);
-                if (trail >= TRAIL_SURROGATE_MIN_VALUE && trail <= TRAIL_SURROGATE_MAX_VALUE) {
-                    return UCharacterProperty.getRawSupplementary(single, trail);
-                }
-            }
-        } else {
-            --offset16;
-            if (offset16 >= 0) {
-                // single is a trail surrogate so
-                char lead = source.charAt(offset16);
-                if (lead >= LEAD_SURROGATE_MIN_VALUE && lead <= LEAD_SURROGATE_MAX_VALUE) {
-                    return UCharacterProperty.getRawSupplementary(lead, single);
-                }
-            }
-        }
-        return single; // return unmatched surrogate
-    }
-
-    /**
-     * Extract a single UTF-32 value from a string.
-     * Used when iterating forwards or backwards (with
-     * UTF16.getCharCount(), as well as random access. If a
-     * validity check is required, use
-     * UCharacter.isLegal()
-     *  on the return value.
-     * If the char retrieved is part of a surrogate pair, its supplementary
-     * character will be returned. If a complete supplementary character is
-     * not found the incomplete character will be returned
-     * @param source array of UTF-16 chars
-     * @param offset16 UTF-16 offset to the start of the character.
-     * @return UTF-32 value for the UTF-32 value that contains the char at
-     *         offset16. The boundaries of that codepoint are the same as in
-     *         bounds32().
-     * @exception IndexOutOfBoundsException thrown if offset16 is out of bounds.
-     * @stable ICU 2.1
-     */
-    public static int charAt(CharSequence source, int offset16) {
-        char single = source.charAt(offset16);
-        if (single < UTF16.LEAD_SURROGATE_MIN_VALUE) {
-            return single;
-        }
-        return _charAt(source, offset16, single);
-    }
-
-    private static int _charAt(CharSequence source, int offset16, char single) {
-        if (single > UTF16.TRAIL_SURROGATE_MAX_VALUE) {
-            return single;
-        }
-
-        // Convert the UTF-16 surrogate pair if necessary.
-        // For simplicity in usage, and because the frequency of pairs is
-        // low, look both directions.
-
-        if (single <= UTF16.LEAD_SURROGATE_MAX_VALUE) {
-            ++offset16;
-            if (source.length() != offset16) {
-                char trail = source.charAt(offset16);
-                if (trail >= UTF16.TRAIL_SURROGATE_MIN_VALUE
-                        && trail <= UTF16.TRAIL_SURROGATE_MAX_VALUE) {
-                    return UCharacterProperty.getRawSupplementary(single, trail);
-                }
-            }
-        } else {
-            --offset16;
-            if (offset16 >= 0) {
-                // single is a trail surrogate so
-                char lead = source.charAt(offset16);
-                if (lead >= UTF16.LEAD_SURROGATE_MIN_VALUE
-                        && lead <= UTF16.LEAD_SURROGATE_MAX_VALUE) {
-                    return UCharacterProperty.getRawSupplementary(lead, single);
-                }
-            }
-        }
-        return single; // return unmatched surrogate
-    }
-
-    /**
-     * Extract a single UTF-32 value from a substring. Used when iterating forwards or backwards
-     * (with UTF16.getCharCount(), as well as random access. If a validity check is
-     * required, use UCharacter.isLegal()
-     * 
-     * on the return value. If the char retrieved is part of a surrogate pair, its supplementary
-     * character will be returned. If a complete supplementary character is not found the incomplete
-     * character will be returned
-     *
-     * @param source Array of UTF-16 chars
-     * @param start Offset to substring in the source array for analyzing
-     * @param limit Offset to substring in the source array for analyzing
-     * @param offset16 UTF-16 offset relative to start
-     * @return UTF-32 value for the UTF-32 value that contains the char at offset16. The boundaries
-     *         of that codepoint are the same as in bounds32().
-     * @exception IndexOutOfBoundsException Thrown if offset16 is not within the range of start and limit.
-     * @stable ICU 2.1
-     */
-    public static int charAt(char source[], int start, int limit, int offset16) {
-        offset16 += start;
-        if (offset16 < start || offset16 >= limit) {
-            throw new ArrayIndexOutOfBoundsException(offset16);
-        }
-
-        char single = source[offset16];
-        if (!isSurrogate(single)) {
-            return single;
-        }
-
-        // Convert the UTF-16 surrogate pair if necessary.
-        // For simplicity in usage, and because the frequency of pairs is
-        // low, look both directions.
-        if (single <= LEAD_SURROGATE_MAX_VALUE) {
-            offset16++;
-            if (offset16 >= limit) {
-                return single;
-            }
-            char trail = source[offset16];
-            if (isTrailSurrogate(trail)) {
-                return UCharacterProperty.getRawSupplementary(single, trail);
-            }
-        }
-        else { // isTrailSurrogate(single), so
-            if (offset16 == start) {
-                return single;
-            }
-            offset16--;
-            char lead = source[offset16];
-            if (isLeadSurrogate(lead))
-                return UCharacterProperty.getRawSupplementary(lead, single);
-        }
-        return single; // return unmatched surrogate
-    }
-
-    /**
-     * Determines how many chars this char32 requires.
-     * If a validity check is required, use 
-     * isLegal() on
-     * char32 before calling.
-     * @param char32 the input codepoint.
-     * @return 2 if is in supplementary space, otherwise 1.
-     * @stable ICU 2.1
-     */
-    public static int getCharCount(int char32)
-    {
-        if (char32 < SUPPLEMENTARY_MIN_VALUE) {
-            return 1;
-        }
-        return 2;
-    }
-
-    /**
-     * Determines whether the code value is a surrogate.
-     * @param char16 the input character.
-     * @return true if the input character is a surrogate.
-     * @stable ICU 2.1
-     */
-    public static boolean isSurrogate(char char16)
-    {
-        return (char16 & SURROGATE_BITMASK) == SURROGATE_BITS;
-    }
-
-    /**
-     * Determines whether the character is a trail surrogate.
-     * @param char16 the input character.
-     * @return true if the input character is a trail surrogate.
-     * @stable ICU 2.1
-     */
-    public static boolean isTrailSurrogate(char char16)
-    {
-        return (char16 & TRAIL_SURROGATE_BITMASK) == TRAIL_SURROGATE_BITS;
-    }
-
-    /**
-     * Determines whether the character is a lead surrogate.
-     * @param char16 the input character.
-     * @return true if the input character is a lead surrogate
-     * @stable ICU 2.1
-     */
-    public static boolean isLeadSurrogate(char char16)
-    {
-        return (char16 & LEAD_SURROGATE_BITMASK) == LEAD_SURROGATE_BITS;
-    }
-
-    /**
-     * Returns the lead surrogate.
-     * If a validity check is required, use
-     * isLegal()
-     * on char32 before calling.
-     * @param char32 the input character.
-     * @return lead surrogate if the getCharCount(ch) is 2; 

-     *         and 0 otherwise (note: 0 is not a valid lead surrogate).
-     * @stable ICU 2.1
-     */
-    public static char getLeadSurrogate(int char32)
-    {
-        if (char32 >= SUPPLEMENTARY_MIN_VALUE) {
-            return (char)(LEAD_SURROGATE_OFFSET_ +
-                          (char32 >> LEAD_SURROGATE_SHIFT_));
-        }
-
-        return 0;
-    }
-
-    /**
-     * Returns the trail surrogate.
-     * If a validity check is required, use
-     * isLegal()
-     * on char32 before calling.
-     * @param char32 the input character.
-     * @return the trail surrogate if the getCharCount(ch) is 2; 
 otherwise
-     *         the character itself
-     * @stable ICU 2.1
-     */
-    public static char getTrailSurrogate(int char32)
-    {
-        if (char32 >= SUPPLEMENTARY_MIN_VALUE) {
-            return (char)(TRAIL_SURROGATE_MIN_VALUE +
-                          (char32 & TRAIL_SURROGATE_MASK_));
-        }
-
-        return (char) char32;
-    }
-
-    /**
-     * Convenience method corresponding to String.valueOf(char). Returns a one
-     * or two char string containing the UTF-32 value in UTF16 format. If a
-     * validity check is required, use
-     * isLegal()
-     * on char32 before calling.
-     * @param char32 the input character.
-     * @return string value of char32 in UTF16 format
-     * @exception IllegalArgumentException thrown if char32 is a invalid
-     *            codepoint.
-     * @stable ICU 2.1
-     */
-    public static String valueOf(int char32)
-    {
-        if (char32 < CODEPOINT_MIN_VALUE || char32 > CODEPOINT_MAX_VALUE) {
-            throw new IllegalArgumentException("Illegal codepoint");
-        }
-        return toString(char32);
-    }
-
-    /**
-     * Append a single UTF-32 value to the end of a StringBuffer.
-     * If a validity check is required, use
-     * isLegal()
-     * on char32 before calling.
-     * @param target the buffer to append to
-     * @param char32 value to append.
-     * @return the updated StringBuffer
-     * @exception IllegalArgumentException thrown when char32 does not lie
-     *            within the range of the Unicode codepoints
-     * @stable ICU 2.1
-     */
-    public static StringBuffer append(StringBuffer target, int char32)
-    {
-        // Check for irregular values
-        if (char32 < CODEPOINT_MIN_VALUE || char32 > CODEPOINT_MAX_VALUE) {
-            throw new IllegalArgumentException("Illegal codepoint: " + Integer.toHexString(char32));
-        }
-
-        // Write the UTF-16 values
-        if (char32 >= SUPPLEMENTARY_MIN_VALUE)
-            {
-            target.append(getLeadSurrogate(char32));
-            target.append(getTrailSurrogate(char32));
-        }
-        else {
-            target.append((char) char32);
-        }
-        return target;
-    }
-
-    /**
-     * Shifts offset16 by the argument number of codepoints within a subarray.
-     * @param source char array
-     * @param start position of the subarray to be performed on
-     * @param limit position of the subarray to be performed on
-     * @param offset16 UTF16 position to shift relative to start
-     * @param shift32 number of codepoints to shift
-     * @return new shifted offset16 relative to start
-     * @exception IndexOutOfBoundsException if the new offset16 is out of
-     *            bounds with respect to the subarray or the subarray bounds
-     *            are out of range.
-     * @stable ICU 2.1
-     */
-    public static int moveCodePointOffset(char source[], int start, int limit,
-                                          int offset16, int shift32)
-    {
-        int size = source.length;
-        int count;
-        char ch;
-        int result = offset16 + start;
-        if (start < 0 || limit < start) {
-            throw new StringIndexOutOfBoundsException(start);
-        }
-        if (limit > size) {
-            throw new StringIndexOutOfBoundsException(limit);
-        }
-        if (offset16 < 0 || result > limit) {
-            throw new StringIndexOutOfBoundsException(offset16);
-        }
-        if (shift32 > 0) {
-            if (shift32 + result > size) {
-                throw new StringIndexOutOfBoundsException(result);
-            }
-            count = shift32;
-            while (result < limit && count > 0)
-            {
-                ch = source[result];
-                if (isLeadSurrogate(ch) && (result + 1 < limit) &&
-                    isTrailSurrogate(source[result + 1])) {
-                    result++;
-                }
-                count--;
-                result++;
-            }
-        } else {
-            if (result + shift32 < start) {
-                throw new StringIndexOutOfBoundsException(result);
-            }
-            for (count = -shift32; count > 0; count--) {
-                result--;
-                if (result < start) {
-                    break;
-                }
-                ch = source[result];
-                if (isTrailSurrogate(ch) && result > start && isLeadSurrogate(source[result - 1])) {
-                    result--;
-                }
-            }
-        }
-        if (count != 0) {
-            throw new StringIndexOutOfBoundsException(shift32);
-        }
-        result -= start;
-        return result;
-    }
-
-    // private data members -------------------------------------------------
-
-    /**
-     * Shift value for lead surrogate to form a supplementary character.
-     */
-    private static final int LEAD_SURROGATE_SHIFT_ = 10;
-
-    /**
-     * Mask to retrieve the significant value from a trail surrogate.
-     */
-    private static final int TRAIL_SURROGATE_MASK_ = 0x3FF;
-
-    /**
-     * Value that all lead surrogate starts with
-     */
-    private static final int LEAD_SURROGATE_OFFSET_ =
-        LEAD_SURROGATE_MIN_VALUE -
-        (SUPPLEMENTARY_MIN_VALUE
-        >> LEAD_SURROGATE_SHIFT_);
-
-    // private methods ------------------------------------------------------
-
-    /**
-     * 
Converts argument code point and returns a String object representing
-     * the code point's value in UTF16 format.
-     * 
This method does not check for the validity of the codepoint, the
-     * results are not guaranteed if a invalid codepoint is passed as
-     * argument.
-     * 
The result is a string whose length is 1 for non-supplementary code
-     * points, 2 otherwise.
-     * @param ch code point
-     * @return string representation of the code point
-     */
-    private static String toString(int ch)
-    {
-        if (ch < SUPPLEMENTARY_MIN_VALUE) {
-            return String.valueOf((char) ch);
-        }
-
-        StringBuilder result = new StringBuilder();
-        result.append(getLeadSurrogate(ch));
-        result.append(getTrailSurrogate(ch));
-        return result.toString();
-    }
-}
--- /dev/null	2020-01-10 15:58:07.000000000 -0800
+++ new/src/java.base/share/classes/jdk/internal/icu/text/UTF16.java	2020-01-10 15:58:07.000000000 -0800
@@ -0,0 +1,618 @@
+/*
+ * Copyright (c) 2005, 2020, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.  Oracle designates this
+ * particular file as subject to the "Classpath" exception as provided
+ * by Oracle in the LICENSE file that accompanied this code.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+/**
+ *******************************************************************************
+ * Copyright (C) 1996-2014, International Business Machines Corporation and
+ * others. All Rights Reserved.
+ *******************************************************************************
+ */
+
+package jdk.internal.icu.text;
+
+import jdk.internal.icu.impl.UCharacterProperty;
+
+/**
+ * 
Standalone utility class providing UTF16 character conversions and
+ * indexing conversions.
+ * 
Code that uses strings alone rarely need modification.
+ * By design, UTF-16 does not allow overlap, so searching for strings is a safe
+ * operation. Similarly, concatenation is always safe. Substringing is safe if
+ * the start and end are both on UTF-32 boundaries. In normal code, the values
+ * for start and end are on those boundaries, since they arose from operations
+ * like searching. If not, the nearest UTF-32 boundaries can be determined
+ * using bounds().
+ * Examples:
+ * 
The following examples illustrate use of some of these methods.
+ * 
{@code
+ * // iteration forwards: Original
+ * for (int i = 0; i < s.length(); ++i) {
+ *     char ch = s.charAt(i);
+ *     doSomethingWith(ch);
+ * }
+ *
+ * // iteration forwards: Changes for UTF-32
+ * int ch;
+ * for (int i = 0; i < s.length(); i += UTF16.getCharCount(ch)) {
+ *     ch = UTF16.charAt(s, i);
+ *     doSomethingWith(ch);
+ * }
+ *
+ * // iteration backwards: Original
+ * for (int i = s.length() - 1; i >= 0; --i) {
+ *     char ch = s.charAt(i);
+ *     doSomethingWith(ch);
+ * }
+ *
+ * // iteration backwards: Changes for UTF-32
+ * int ch;
+ * for (int i = s.length() - 1; i > 0; i -= UTF16.getCharCount(ch)) {
+ *     ch = UTF16.charAt(s, i);
+ *     doSomethingWith(ch);
+ * }
+ * }

+ * Notes:
+ * 
    
+ *   
  • 
+ *   Naming: For clarity, High and Low surrogates are called
+ *   Lead and Trail in the API, which gives a better
+ *   sense of their ordering in a string. offset16 and
+ *   offset32 are used to distinguish offsets to UTF-16
+ *   boundaries vs offsets to UTF-32 boundaries. int char32 is
+ *   used to contain UTF-32 characters, as opposed to char16,
+ *   which is a UTF-16 code unit.
+ *   
    • 
+ *   
  • 
+ *   Roundtripping Offsets: You can always roundtrip from a
+ *   UTF-32 offset to a UTF-16 offset and back. Because of the difference in
+ *   structure, you can roundtrip from a UTF-16 offset to a UTF-32 offset and
+ *   back if and only if bounds(string, offset16) != TRAIL.
+ *   
    • 
+ *   
  • 
+ *   Exceptions: The error checking will throw an exception
+ *   if indices are out of bounds. Other than that, all methods will
+ *   behave reasonably, even if unmatched surrogates or out-of-bounds UTF-32
+ *   values are present. UCharacter.isLegal() can be used to check
+ *   for validity if desired.
+ *   
    • 
+ *   
  • 
+ *   Unmatched Surrogates: If the string contains unmatched
+ *   surrogates, then these are counted as one UTF-32 value. This matches
+ *   their iteration behavior, which is vital. It also matches common display
+ *   practice as missing glyphs (see the Unicode Standard Section 5.4, 5.5).
+ *   
    • 
+ *   
  • 
+ *   Optimization: The method implementations may need
+ *   optimization if the compiler doesn't fold static final methods. Since
+ *   surrogate pairs will form an exceeding small percentage of all the text
+ *   in the world, the singleton case should always be optimized for.
+ *   
    • 
+ * 

+ * @author Mark Davis, with help from Markus Scherer
+ * @stable ICU 2.1
+ */
+
+public final class UTF16
+{
+    // public variables ---------------------------------------------------
+
+    /**
+     * The lowest Unicode code point value.
+     * @stable ICU 2.1
+     */
+    public static final int CODEPOINT_MIN_VALUE = 0;
+    /**
+     * The highest Unicode code point value (scalar value) according to the
+     * Unicode Standard.
+     * @stable ICU 2.1
+     */
+    public static final int CODEPOINT_MAX_VALUE = 0x10ffff;
+    /**
+     * The minimum value for Supplementary code points
+     * @stable ICU 2.1
+     */
+    public static final int SUPPLEMENTARY_MIN_VALUE = 0x10000;
+    /**
+     * Lead surrogate minimum value
+     * @stable ICU 2.1
+     */
+    public static final int LEAD_SURROGATE_MIN_VALUE = 0xD800;
+    /**
+     * Trail surrogate minimum value
+     * @stable ICU 2.1
+     */
+    public static final int TRAIL_SURROGATE_MIN_VALUE = 0xDC00;
+    /**
+     * Lead surrogate maximum value
+     * @stable ICU 2.1
+     */
+    public static final int LEAD_SURROGATE_MAX_VALUE = 0xDBFF;
+    /**
+     * Trail surrogate maximum value
+     * @stable ICU 2.1
+     */
+    public static final int TRAIL_SURROGATE_MAX_VALUE = 0xDFFF;
+    /**
+     * Surrogate minimum value
+     * @stable ICU 2.1
+     */
+    public static final int SURROGATE_MIN_VALUE = LEAD_SURROGATE_MIN_VALUE;
+    /**
+     * Lead surrogate bitmask
+     */
+    private static final int LEAD_SURROGATE_BITMASK = 0xFFFFFC00;
+    /**
+     * Trail surrogate bitmask
+     */
+    private static final int TRAIL_SURROGATE_BITMASK = 0xFFFFFC00;
+    /**
+     * Surrogate bitmask
+     */
+    private static final int SURROGATE_BITMASK = 0xFFFFF800;
+    /**
+     * Lead surrogate bits
+     */
+    private static final int LEAD_SURROGATE_BITS = 0xD800;
+    /**
+     * Trail surrogate bits
+     */
+    private static final int TRAIL_SURROGATE_BITS = 0xDC00;
+    /**
+     * Surrogate bits
+     */
+    private static final int SURROGATE_BITS = 0xD800;
+
+    // constructor --------------------------------------------------------
+
+    // /CLOVER:OFF
+    /**
+     * Prevent instance from being created.
+     */
+    private UTF16() {
+    }
+
+    // /CLOVER:ON
+    // public method ------------------------------------------------------
+
+    /**
+     * Extract a single UTF-32 value from a string.
+     * Used when iterating forwards or backwards (with
+     * UTF16.getCharCount(), as well as random access. If a
+     * validity check is required, use
+     * 
+     * UCharacter.isLegal() on the return value.
+     * If the char retrieved is part of a surrogate pair, its supplementary
+     * character will be returned. If a complete supplementary character is
+     * not found the incomplete character will be returned
+     * @param source array of UTF-16 chars
+     * @param offset16 UTF-16 offset to the start of the character.
+     * @return UTF-32 value for the UTF-32 value that contains the char at
+     *         offset16. The boundaries of that codepoint are the same as in
+     *         bounds32().
+     * @exception IndexOutOfBoundsException thrown if offset16 is out of
+     *            bounds.
+     * @stable ICU 2.1
+     */
+    public static int charAt(String source, int offset16) {
+        char single = source.charAt(offset16);
+        if (single < LEAD_SURROGATE_MIN_VALUE) {
+            return single;
+        }
+        return _charAt(source, offset16, single);
+    }
+
+    private static int _charAt(String source, int offset16, char single) {
+        if (single > TRAIL_SURROGATE_MAX_VALUE) {
+            return single;
+        }
+
+        // Convert the UTF-16 surrogate pair if necessary.
+        // For simplicity in usage, and because the frequency of pairs is
+        // low, look both directions.
+
+        if (single <= LEAD_SURROGATE_MAX_VALUE) {
+            ++offset16;
+            if (source.length() != offset16) {
+                char trail = source.charAt(offset16);
+                if (trail >= TRAIL_SURROGATE_MIN_VALUE && trail <= TRAIL_SURROGATE_MAX_VALUE) {
+                    return UCharacterProperty.getRawSupplementary(single, trail);
+                }
+            }
+        } else {
+            --offset16;
+            if (offset16 >= 0) {
+                // single is a trail surrogate so
+                char lead = source.charAt(offset16);
+                if (lead >= LEAD_SURROGATE_MIN_VALUE && lead <= LEAD_SURROGATE_MAX_VALUE) {
+                    return UCharacterProperty.getRawSupplementary(lead, single);
+                }
+            }
+        }
+        return single; // return unmatched surrogate
+    }
+
+    /**
+     * Extract a single UTF-32 value from a string.
+     * Used when iterating forwards or backwards (with
+     * UTF16.getCharCount(), as well as random access. If a
+     * validity check is required, use
+     * UCharacter.isLegal()
+     *  on the return value.
+     * If the char retrieved is part of a surrogate pair, its supplementary
+     * character will be returned. If a complete supplementary character is
+     * not found the incomplete character will be returned
+     * @param source array of UTF-16 chars
+     * @param offset16 UTF-16 offset to the start of the character.
+     * @return UTF-32 value for the UTF-32 value that contains the char at
+     *         offset16. The boundaries of that codepoint are the same as in
+     *         bounds32().
+     * @exception IndexOutOfBoundsException thrown if offset16 is out of bounds.
+     * @stable ICU 2.1
+     */
+    public static int charAt(CharSequence source, int offset16) {
+        char single = source.charAt(offset16);
+        if (single < UTF16.LEAD_SURROGATE_MIN_VALUE) {
+            return single;
+        }
+        return _charAt(source, offset16, single);
+    }
+
+    private static int _charAt(CharSequence source, int offset16, char single) {
+        if (single > UTF16.TRAIL_SURROGATE_MAX_VALUE) {
+            return single;
+        }
+
+        // Convert the UTF-16 surrogate pair if necessary.
+        // For simplicity in usage, and because the frequency of pairs is
+        // low, look both directions.
+
+        if (single <= UTF16.LEAD_SURROGATE_MAX_VALUE) {
+            ++offset16;
+            if (source.length() != offset16) {
+                char trail = source.charAt(offset16);
+                if (trail >= UTF16.TRAIL_SURROGATE_MIN_VALUE
+                        && trail <= UTF16.TRAIL_SURROGATE_MAX_VALUE) {
+                    return UCharacterProperty.getRawSupplementary(single, trail);
+                }
+            }
+        } else {
+            --offset16;
+            if (offset16 >= 0) {
+                // single is a trail surrogate so
+                char lead = source.charAt(offset16);
+                if (lead >= UTF16.LEAD_SURROGATE_MIN_VALUE
+                        && lead <= UTF16.LEAD_SURROGATE_MAX_VALUE) {
+                    return UCharacterProperty.getRawSupplementary(lead, single);
+                }
+            }
+        }
+        return single; // return unmatched surrogate
+    }
+
+    /**
+     * Extract a single UTF-32 value from a substring. Used when iterating forwards or backwards
+     * (with UTF16.getCharCount(), as well as random access. If a validity check is
+     * required, use UCharacter.isLegal()
+     * 
+     * on the return value. If the char retrieved is part of a surrogate pair, its supplementary
+     * character will be returned. If a complete supplementary character is not found the incomplete
+     * character will be returned
+     *
+     * @param source Array of UTF-16 chars
+     * @param start Offset to substring in the source array for analyzing
+     * @param limit Offset to substring in the source array for analyzing
+     * @param offset16 UTF-16 offset relative to start
+     * @return UTF-32 value for the UTF-32 value that contains the char at offset16. The boundaries
+     *         of that codepoint are the same as in bounds32().
+     * @exception IndexOutOfBoundsException Thrown if offset16 is not within the range of start and limit.
+     * @stable ICU 2.1
+     */
+    public static int charAt(char source[], int start, int limit, int offset16) {
+        offset16 += start;
+        if (offset16 < start || offset16 >= limit) {
+            throw new ArrayIndexOutOfBoundsException(offset16);
+        }
+
+        char single = source[offset16];
+        if (!isSurrogate(single)) {
+            return single;
+        }
+
+        // Convert the UTF-16 surrogate pair if necessary.
+        // For simplicity in usage, and because the frequency of pairs is
+        // low, look both directions.
+        if (single <= LEAD_SURROGATE_MAX_VALUE) {
+            offset16++;
+            if (offset16 >= limit) {
+                return single;
+            }
+            char trail = source[offset16];
+            if (isTrailSurrogate(trail)) {
+                return UCharacterProperty.getRawSupplementary(single, trail);
+            }
+        }
+        else { // isTrailSurrogate(single), so
+            if (offset16 == start) {
+                return single;
+            }
+            offset16--;
+            char lead = source[offset16];
+            if (isLeadSurrogate(lead))
+                return UCharacterProperty.getRawSupplementary(lead, single);
+        }
+        return single; // return unmatched surrogate
+    }
+
+    /**
+     * Determines how many chars this char32 requires.
+     * If a validity check is required, use 
+     * isLegal() on
+     * char32 before calling.
+     * @param char32 the input codepoint.
+     * @return 2 if is in supplementary space, otherwise 1.
+     * @stable ICU 2.1
+     */
+    public static int getCharCount(int char32)
+    {
+        if (char32 < SUPPLEMENTARY_MIN_VALUE) {
+            return 1;
+        }
+        return 2;
+    }
+
+    /**
+     * Determines whether the code value is a surrogate.
+     * @param char16 the input character.
+     * @return true if the input character is a surrogate.
+     * @stable ICU 2.1
+     */
+    public static boolean isSurrogate(char char16)
+    {
+        return (char16 & SURROGATE_BITMASK) == SURROGATE_BITS;
+    }
+
+    /**
+     * Determines whether the character is a trail surrogate.
+     * @param char16 the input character.
+     * @return true if the input character is a trail surrogate.
+     * @stable ICU 2.1
+     */
+    public static boolean isTrailSurrogate(char char16)
+    {
+        return (char16 & TRAIL_SURROGATE_BITMASK) == TRAIL_SURROGATE_BITS;
+    }
+
+    /**
+     * Determines whether the character is a lead surrogate.
+     * @param char16 the input character.
+     * @return true if the input character is a lead surrogate
+     * @stable ICU 2.1
+     */
+    public static boolean isLeadSurrogate(char char16)
+    {
+        return (char16 & LEAD_SURROGATE_BITMASK) == LEAD_SURROGATE_BITS;
+    }
+
+    /**
+     * Returns the lead surrogate.
+     * If a validity check is required, use
+     * isLegal()
+     * on char32 before calling.
+     * @param char32 the input character.
+     * @return lead surrogate if the getCharCount(ch) is 2; 

+     *         and 0 otherwise (note: 0 is not a valid lead surrogate).
+     * @stable ICU 2.1
+     */
+    public static char getLeadSurrogate(int char32)
+    {
+        if (char32 >= SUPPLEMENTARY_MIN_VALUE) {
+            return (char)(LEAD_SURROGATE_OFFSET_ +
+                          (char32 >> LEAD_SURROGATE_SHIFT_));
+        }
+
+        return 0;
+    }
+
+    /**
+     * Returns the trail surrogate.
+     * If a validity check is required, use
+     * isLegal()
+     * on char32 before calling.
+     * @param char32 the input character.
+     * @return the trail surrogate if the getCharCount(ch) is 2; 
 otherwise
+     *         the character itself
+     * @stable ICU 2.1
+     */
+    public static char getTrailSurrogate(int char32)
+    {
+        if (char32 >= SUPPLEMENTARY_MIN_VALUE) {
+            return (char)(TRAIL_SURROGATE_MIN_VALUE +
+                          (char32 & TRAIL_SURROGATE_MASK_));
+        }
+
+        return (char) char32;
+    }
+
+    /**
+     * Convenience method corresponding to String.valueOf(char). Returns a one
+     * or two char string containing the UTF-32 value in UTF16 format. If a
+     * validity check is required, use
+     * isLegal()
+     * on char32 before calling.
+     * @param char32 the input character.
+     * @return string value of char32 in UTF16 format
+     * @exception IllegalArgumentException thrown if char32 is a invalid
+     *            codepoint.
+     * @stable ICU 2.1
+     */
+    public static String valueOf(int char32)
+    {
+        if (char32 < CODEPOINT_MIN_VALUE || char32 > CODEPOINT_MAX_VALUE) {
+            throw new IllegalArgumentException("Illegal codepoint");
+        }
+        return toString(char32);
+    }
+
+    /**
+     * Append a single UTF-32 value to the end of a StringBuffer.
+     * If a validity check is required, use
+     * isLegal()
+     * on char32 before calling.
+     * @param target the buffer to append to
+     * @param char32 value to append.
+     * @return the updated StringBuffer
+     * @exception IllegalArgumentException thrown when char32 does not lie
+     *            within the range of the Unicode codepoints
+     * @stable ICU 2.1
+     */
+    public static StringBuffer append(StringBuffer target, int char32)
+    {
+        // Check for irregular values
+        if (char32 < CODEPOINT_MIN_VALUE || char32 > CODEPOINT_MAX_VALUE) {
+            throw new IllegalArgumentException("Illegal codepoint: " + Integer.toHexString(char32));
+        }
+
+        // Write the UTF-16 values
+        if (char32 >= SUPPLEMENTARY_MIN_VALUE)
+            {
+            target.append(getLeadSurrogate(char32));
+            target.append(getTrailSurrogate(char32));
+        }
+        else {
+            target.append((char) char32);
+        }
+        return target;
+    }
+
+    /**
+     * Shifts offset16 by the argument number of codepoints within a subarray.
+     * @param source char array
+     * @param start position of the subarray to be performed on
+     * @param limit position of the subarray to be performed on
+     * @param offset16 UTF16 position to shift relative to start
+     * @param shift32 number of codepoints to shift
+     * @return new shifted offset16 relative to start
+     * @exception IndexOutOfBoundsException if the new offset16 is out of
+     *            bounds with respect to the subarray or the subarray bounds
+     *            are out of range.
+     * @stable ICU 2.1
+     */
+    public static int moveCodePointOffset(char source[], int start, int limit,
+                                          int offset16, int shift32)
+    {
+        int size = source.length;
+        int count;
+        char ch;
+        int result = offset16 + start;
+        if (start < 0 || limit < start) {
+            throw new StringIndexOutOfBoundsException(start);
+        }
+        if (limit > size) {
+            throw new StringIndexOutOfBoundsException(limit);
+        }
+        if (offset16 < 0 || result > limit) {
+            throw new StringIndexOutOfBoundsException(offset16);
+        }
+        if (shift32 > 0) {
+            if (shift32 + result > size) {
+                throw new StringIndexOutOfBoundsException(result);
+            }
+            count = shift32;
+            while (result < limit && count > 0)
+            {
+                ch = source[result];
+                if (isLeadSurrogate(ch) && (result + 1 < limit) &&
+                    isTrailSurrogate(source[result + 1])) {
+                    result++;
+                }
+                count--;
+                result++;
+            }
+        } else {
+            if (result + shift32 < start) {
+                throw new StringIndexOutOfBoundsException(result);
+            }
+            for (count = -shift32; count > 0; count--) {
+                result--;
+                if (result < start) {
+                    break;
+                }
+                ch = source[result];
+                if (isTrailSurrogate(ch) && result > start && isLeadSurrogate(source[result - 1])) {
+                    result--;
+                }
+            }
+        }
+        if (count != 0) {
+            throw new StringIndexOutOfBoundsException(shift32);
+        }
+        result -= start;
+        return result;
+    }
+
+    // private data members -------------------------------------------------
+
+    /**
+     * Shift value for lead surrogate to form a supplementary character.
+     */
+    private static final int LEAD_SURROGATE_SHIFT_ = 10;
+
+    /**
+     * Mask to retrieve the significant value from a trail surrogate.
+     */
+    private static final int TRAIL_SURROGATE_MASK_ = 0x3FF;
+
+    /**
+     * Value that all lead surrogate starts with
+     */
+    private static final int LEAD_SURROGATE_OFFSET_ =
+        LEAD_SURROGATE_MIN_VALUE -
+        (SUPPLEMENTARY_MIN_VALUE
+        >> LEAD_SURROGATE_SHIFT_);
+
+    // private methods ------------------------------------------------------
+
+    /**
+     * 
Converts argument code point and returns a String object representing
+     * the code point's value in UTF16 format.
+     * 
This method does not check for the validity of the codepoint, the
+     * results are not guaranteed if a invalid codepoint is passed as
+     * argument.
+     * 
The result is a string whose length is 1 for non-supplementary code
+     * points, 2 otherwise.
+     * @param ch code point
+     * @return string representation of the code point
+     */
+    private static String toString(int ch)
+    {
+        if (ch < SUPPLEMENTARY_MIN_VALUE) {
+            return String.valueOf((char) ch);
+        }
+
+        StringBuilder result = new StringBuilder();
+        result.append(getLeadSurrogate(ch));
+        result.append(getTrailSurrogate(ch));
+        return result.toString();
+    }
+}
--- old/src/java.base/share/classes/sun/text/normalizer/UnicodeSet.java	2020-01-10 15:58:09.000000000 -0800
+++ /dev/null	2020-01-10 15:58:09.000000000 -0800
@@ -1,1407 +0,0 @@
-/*
- * Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved.
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.  Oracle designates this
- * particular file as subject to the "Classpath" exception as provided
- * by Oracle in the LICENSE file that accompanied this code.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- */
-
-/*
- *******************************************************************************
- * Copyright (C) 1996-2015, International Business Machines Corporation and
- * others. All Rights Reserved.
- *******************************************************************************
- */
-package sun.text.normalizer;
-
-import java.io.IOException;
-import java.text.ParsePosition;
-import java.util.ArrayList;
-import java.util.TreeSet;
-
-/**
- * A mutable set of Unicode characters and multicharacter strings.
- * Objects of this class represent character classes used
- * in regular expressions. A character specifies a subset of Unicode
- * code points.  Legal code points are U+0000 to U+10FFFF, inclusive.
- *
- * Note: method freeze() will not only make the set immutable, but
- * also makes important methods much higher performance:
- * contains(c), containsNone(...), span(...), spanBack(...) etc.
- * After the object is frozen, any subsequent call that wants to change
- * the object will throw UnsupportedOperationException.
- *
- * 
The UnicodeSet class is not designed to be subclassed.
- *
- * 
UnicodeSet supports two APIs. The first is the
- * operand API that allows the caller to modify the value of
- * a UnicodeSet object. It conforms to Java 2's
- * java.util.Set interface, although
- * UnicodeSet does not actually implement that
- * interface. All methods of Set are supported, with the
- * modification that they take a character range or single character
- * instead of an Object, and they take a
- * UnicodeSet instead of a Collection.  The
- * operand API may be thought of in terms of boolean logic: a boolean
- * OR is implemented by add, a boolean AND is implemented
- * by retain, a boolean XOR is implemented by
- * complement taking an argument, and a boolean NOT is
- * implemented by complement with no argument.  In terms
- * of traditional set theory function names, add is a
- * union, retain is an intersection, remove
- * is an asymmetric difference, and complement with no
- * argument is a set complement with respect to the superset range
- * MIN_VALUE-MAX_VALUE
- *
- * 
The second API is the
- * applyPattern()/toPattern() API from the
- * java.text.Format-derived classes.  Unlike the
- * methods that add characters, add categories, and control the logic
- * of the set, the method applyPattern() sets all
- * attributes of a UnicodeSet at once, based on a
- * string pattern.
- *
- * 
Pattern syntax

- *
- * Patterns are accepted by the constructors and the
- * applyPattern() methods and returned by the
- * toPattern() method.  These patterns follow a syntax
- * similar to that employed by version 8 regular expression character
- * classes.  Here are some simple examples:
- *
- * 

- *   
- *     
- *       
- *       
- *     
- *       
- *       
- *     
- *       
- *       
- *     
- *     
- *       
- *       
- *     
- *     
- *       
- *       
- *     
- *     
- *       
- *       
- *     
- *     
- *       
- *       
- *     
- *   
[]No characters
[a]The character 'a'
[ae]The characters 'a' and 'e'
[a-e]The characters 'a' through 'e' inclusive, in Unicode code
- *       point order
[\\u4E01]The character U+4E01
[a{ab}{ac}]The character 'a' and the multicharacter strings "ab" and
- *       "ac"
[\p{Lu}]All characters in the general category Uppercase Letter

- * 

- *
- * Any character may be preceded by a backslash in order to remove any special
- * meaning.  White space characters, as defined by the Unicode Pattern_White_Space property, are
- * ignored, unless they are escaped.
- *
- * 
Property patterns specify a set of characters having a certain
- * property as defined by the Unicode standard.  Both the POSIX-like
- * "[:Lu:]" and the Perl-like syntax "\p{Lu}" are recognized.  For a
- * complete list of supported property patterns, see the User's Guide
- * for UnicodeSet at
- * 
- * http://www.icu-project.org/userguide/unicodeSet.html.
- * Actual determination of property data is defined by the underlying
- * Unicode database as implemented by UCharacter.
- *
- * 
Patterns specify individual characters, ranges of characters, and
- * Unicode property sets.  When elements are concatenated, they
- * specify their union.  To complement a set, place a '^' immediately
- * after the opening '['.  Property patterns are inverted by modifying
- * their delimiters; "[:^foo]" and "\P{foo}".  In any other location,
- * '^' has no special meaning.
- *
- * 
Ranges are indicated by placing two a '-' between two
- * characters, as in "a-z".  This specifies the range of all
- * characters from the left to the right, in Unicode order.  If the
- * left character is greater than or equal to the
- * right character it is a syntax error.  If a '-' occurs as the first
- * character after the opening '[' or '[^', or if it occurs as the
- * last character before the closing ']', then it is taken as a
- * literal.  Thus "[a\\-b]", "[-ab]", and "[ab-]" all indicate the same
- * set of three characters, 'a', 'b', and '-'.
- *
- * 
Sets may be intersected using the {@literal '&'} operator or the asymmetric
- * set difference may be taken using the '-' operator, for example,
- * "{@code [[:L:]&[\\u0000-\\u0FFF]]}" indicates the set of all Unicode letters
- * with values less than 4096.  Operators ({@literal '&'} and '|') have equal
- * precedence and bind left-to-right.  Thus
- * "[[:L:]-[a-z]-[\\u0100-\\u01FF]]" is equivalent to
- * "[[[:L:]-[a-z]]-[\\u0100-\\u01FF]]".  This only really matters for
- * difference; intersection is commutative.
- *
- * 
- * 
[a]The set containing 'a'
- * 
[a-z]The set containing 'a'
- * through 'z' and all letters in between, in Unicode order
- * 
[^a-z]The set containing
- * all characters but 'a' through 'z',
- * that is, U+0000 through 'a'-1 and 'z'+1 through U+10FFFF
- * 
[[pat1][pat2]]
- * The union of sets specified by pat1 and pat2
- * 
[[pat1]&[pat2]]
- * The intersection of sets specified by pat1 and pat2
- * 
[[pat1]-[pat2]]
- * The asymmetric difference of sets specified by pat1 and
- * pat2
- * 
[:Lu:] or \p{Lu}
- * The set of characters having the specified
- * Unicode property; in
- * this case, Unicode uppercase letters
- * 
[:^Lu:] or \P{Lu}
- * The set of characters not having the given
- * Unicode property
- * 

- *
- * 
Warning: you cannot add an empty string ("") to a UnicodeSet.

- *
- * 
Formal syntax

- *
- * 

- *   
- *     
- *       
- *       
- *     
- *     
- *       
- *       
- *     
- *     
- *       
- *       
- *     
- *     
- *       
- *       
- *     
- *     
- *       
- *       
- *     
- *     
- *       
- *       
- *     
- *     
- *       
- *       
- *     
- *     
- *       
- *       
- *     
- *   
pattern :=  ('[' '^'? item* ']') |
- *       property
item :=  char | (char '-' char) | pattern-expr

- *       
pattern-expr :=  pattern | pattern-expr pattern |
- *       pattern-expr op pattern

- *       
op :=  '&' | '-'

- *       
special :=  '[' | ']' | '-'

- *       
char :=  any character that is not special

- *       | ('\\' any character)

- *       | ('\u' hex hex hex hex)

- *       
hex :=  any character for which
- *       Character.digit(c, 16)
- *       returns a non-negative result
property :=  a Unicode property set pattern

- *   

- *   
- *     
- *       
- *     
- *   
Legend: 
- *         
- *           
- *           
- *           
- *         
- *         
- *           
- *           
- *           
- *         
- *         
- *           
- *           
- *           
- *         
- *         
- *           
- *           
- *           
- *         
- *         
- *           
- *           
- *           
- *         
- *       
a := b  a may be replaced by b 
a?zero or one instance of a

- *           
a*one or more instances of a

- *           
a | beither a or b

- *           
'a'the literal string between the quotes 

- *       

- * 

- * 
To iterate over contents of UnicodeSet, the following are available:
- * 
  • {@link #ranges()} to iterate through the ranges
    • 
- * 
  • {@link #strings()} to iterate through the strings
    • 
- * 
  • {@link #iterator()} to iterate through the entire contents in a single loop.
- * That method is, however, not particularly efficient, since it "boxes" each code point into a String.
- * 

- * All of the above can be used in for loops.
- * The {@link com.ibm.icu.text.UnicodeSetIterator UnicodeSetIterator} can also be used, but not in for loops.
- * 
To replace, count elements, or delete spans, see {@link com.ibm.icu.text.UnicodeSetSpanner UnicodeSetSpanner}.
- *
- * @author Alan Liu
- * @stable ICU 2.0
- */
-class UnicodeSet {
-
-    private static final int LOW = 0x000000; // LOW <= all valid values. ZERO for codepoints
-    private static final int HIGH = 0x110000; // HIGH > all valid values. 10000 for code units.
-    // 110000 for codepoints
-
-    /**
-     * Minimum value that can be stored in a UnicodeSet.
-     * @stable ICU 2.0
-     */
-    public static final int MIN_VALUE = LOW;
-
-    /**
-     * Maximum value that can be stored in a UnicodeSet.
-     * @stable ICU 2.0
-     */
-    public static final int MAX_VALUE = HIGH - 1;
-
-    private int len;      // length used; list may be longer to minimize reallocs
-    private int[] list;   // MUST be terminated with HIGH
-    private int[] rangeList; // internal buffer
-    private int[] buffer; // internal buffer
-
-    // NOTE: normally the field should be of type SortedSet; but that is missing a public clone!!
-    // is not private so that UnicodeSetIterator can get access
-    TreeSet strings = new TreeSet();
-
-    /**
-     * The pattern representation of this set.  This may not be the
-     * most economical pattern.  It is the pattern supplied to
-     * applyPattern(), with variables substituted and whitespace
-     * removed.  For sets constructed without applyPattern(), or
-     * modified using the non-pattern API, this string will be null,
-     * indicating that toPattern() must generate a pattern
-     * representation from the inversion list.
-     */
-
-    private static final int START_EXTRA = 16;         // initial storage. Must be >= 0
-    private static final int GROW_EXTRA = START_EXTRA; // extra amount for growth. Must be >= 0
-
-    private static UnicodeSet INCLUSION = null;
-
-    private volatile BMPSet bmpSet; // The set is frozen if bmpSet or stringSpan is not null.
-    private volatile UnicodeSetStringSpan stringSpan;
-
-    //----------------------------------------------------------------
-    // Public API
-    //----------------------------------------------------------------
-
-    /**
-     * Constructs an empty set.
-     * @stable ICU 2.0
-     */
-    private UnicodeSet() {
-        list = new int[1 + START_EXTRA];
-        list[len++] = HIGH;
-    }
-
-    /**
-     * Constructs a copy of an existing set.
-     * @stable ICU 2.0
-     */
-    private UnicodeSet(UnicodeSet other) {
-        set(other);
-    }
-
-    /**
-     * Constructs a set containing the given range. If end >
-     * start then an empty set is created.
-     *
-     * @param start first character, inclusive, of range
-     * @param end last character, inclusive, of range
-     * @stable ICU 2.0
-     */
-    public UnicodeSet(int start, int end) {
-        this();
-        complement(start, end);
-    }
-
-    /**
-     * Constructs a set from the given pattern.  See the class description
-     * for the syntax of the pattern language.  Whitespace is ignored.
-     * @param pattern a string specifying what characters are in the set
-     * @exception java.lang.IllegalArgumentException if the pattern contains
-     * a syntax error.
-     * @stable ICU 2.0
-     */
-    public UnicodeSet(String pattern) {
-        this();
-        applyPattern(pattern, null);
-    }
-
-    /**
-     * Make this object represent the same set as other.
-     * @param other a UnicodeSet whose value will be
-     * copied to this object
-     * @stable ICU 2.0
-     */
-    public UnicodeSet set(UnicodeSet other) {
-        checkFrozen();
-        list = other.list.clone();
-        len = other.len;
-        strings = new TreeSet(other.strings);
-        return this;
-    }
-
-    /**
-     * Returns the number of elements in this set (its cardinality)
-     * Note than the elements of a set may include both individual
-     * codepoints and strings.
-     *
-     * @return the number of elements in this set (its cardinality).
-     * @stable ICU 2.0
-     */
-    public int size() {
-        int n = 0;
-        int count = getRangeCount();
-        for (int i = 0; i < count; ++i) {
-            n += getRangeEnd(i) - getRangeStart(i) + 1;
-        }
-        return n + strings.size();
-    }
-
-    // for internal use, after checkFrozen has been called
-    private UnicodeSet add_unchecked(int start, int end) {
-        if (start < MIN_VALUE || start > MAX_VALUE) {
-            throw new IllegalArgumentException("Invalid code point U+" + Utility.hex(start, 6));
-        }
-        if (end < MIN_VALUE || end > MAX_VALUE) {
-            throw new IllegalArgumentException("Invalid code point U+" + Utility.hex(end, 6));
-        }
-        if (start < end) {
-            add(range(start, end), 2, 0);
-        } else if (start == end) {
-            add(start);
-        }
-        return this;
-    }
-
-    /**
-     * Adds the specified character to this set if it is not already
-     * present.  If this set already contains the specified character,
-     * the call leaves this set unchanged.
-     * @stable ICU 2.0
-     */
-    public final UnicodeSet add(int c) {
-        checkFrozen();
-        return add_unchecked(c);
-    }
-
-    // for internal use only, after checkFrozen has been called
-    private final UnicodeSet add_unchecked(int c) {
-        if (c < MIN_VALUE || c > MAX_VALUE) {
-            throw new IllegalArgumentException("Invalid code point U+" + Utility.hex(c, 6));
-        }
-
-        // find smallest i such that c < list[i]
-        // if odd, then it is IN the set
-        // if even, then it is OUT of the set
-        int i = findCodePoint(c);
-
-        // already in set?
-        if ((i & 1) != 0) return this;
-
-        // HIGH is 0x110000
-        // assert(list[len-1] == HIGH);
-
-        // empty = [HIGH]
-        // [start_0, limit_0, start_1, limit_1, HIGH]
-
-        // [..., start_k-1, limit_k-1, start_k, limit_k, ..., HIGH]
-        //                             ^
-        //                             list[i]
-
-        // i == 0 means c is before the first range
-
-        if (c == list[i]-1) {
-            // c is before start of next range
-            list[i] = c;
-            // if we touched the HIGH mark, then add a new one
-            if (c == MAX_VALUE) {
-                ensureCapacity(len+1);
-                list[len++] = HIGH;
-            }
-            if (i > 0 && c == list[i-1]) {
-                // collapse adjacent ranges
-
-                // [..., start_k-1, c, c, limit_k, ..., HIGH]
-                //                     ^
-                //                     list[i]
-                System.arraycopy(list, i+1, list, i-1, len-i-1);
-                len -= 2;
-            }
-        }
-
-        else if (i > 0 && c == list[i-1]) {
-            // c is after end of prior range
-            list[i-1]++;
-            // no need to chcek for collapse here
-        }
-
-        else {
-            // At this point we know the new char is not adjacent to
-            // any existing ranges, and it is not 10FFFF.
-
-
-            // [..., start_k-1, limit_k-1, start_k, limit_k, ..., HIGH]
-            //                             ^
-            //                             list[i]
-
-            // [..., start_k-1, limit_k-1, c, c+1, start_k, limit_k, ..., HIGH]
-            //                             ^
-            //                             list[i]
-
-            // Don't use ensureCapacity() to save on copying.
-            // NOTE: This has no measurable impact on performance,
-            // but it might help in some usage patterns.
-            if (len+2 > list.length) {
-                int[] temp = new int[len + 2 + GROW_EXTRA];
-                if (i != 0) System.arraycopy(list, 0, temp, 0, i);
-                System.arraycopy(list, i, temp, i+2, len-i);
-                list = temp;
-            } else {
-                System.arraycopy(list, i, list, i+2, len-i);
-            }
-
-            list[i] = c;
-            list[i+1] = c+1;
-            len += 2;
-        }
-
-        return this;
-    }
-
-    /**
-     * Adds the specified multicharacter to this set if it is not already
-     * present.  If this set already contains the multicharacter,
-     * the call leaves this set unchanged.
-     * Thus {@code "ch" => {"ch"}}
-     * 
Warning: you cannot add an empty string ("") to a UnicodeSet.
-     * @param s the source string
-     * @return this object, for chaining
-     * @stable ICU 2.0
-     */
-    public final UnicodeSet add(CharSequence s) {
-        checkFrozen();
-        int cp = getSingleCP(s);
-        if (cp < 0) {
-            strings.add(s.toString());
-        } else {
-            add_unchecked(cp, cp);
-        }
-        return this;
-    }
-
-    /**
-     * Utility for getting code point from single code point CharSequence.
-     * See the public UTF16.getSingleCodePoint()
-     * @return a code point IF the string consists of a single one.
-     * otherwise returns -1.
-     * @param s to test
-     */
-    private static int getSingleCP(CharSequence s) {
-        if (s.length() < 1) {
-            throw new IllegalArgumentException("Can't use zero-length strings in UnicodeSet");
-        }
-        if (s.length() > 2) return -1;
-        if (s.length() == 1) return s.charAt(0);
-
-        // at this point, len = 2
-        int cp = UTF16.charAt(s, 0);
-        if (cp > 0xFFFF) { // is surrogate pair
-            return cp;
-        }
-        return -1;
-    }
-
-    /**
-     * Complements the specified range in this set.  Any character in
-     * the range will be removed if it is in this set, or will be
-     * added if it is not in this set.  If {@code end > start}
-     * then an empty range is complemented, leaving the set unchanged.
-     *
-     * @param start first character, inclusive, of range to be removed
-     * from this set.
-     * @param end last character, inclusive, of range to be removed
-     * from this set.
-     * @stable ICU 2.0
-     */
-    public UnicodeSet complement(int start, int end) {
-        checkFrozen();
-        if (start < MIN_VALUE || start > MAX_VALUE) {
-            throw new IllegalArgumentException("Invalid code point U+" + Utility.hex(start, 6));
-        }
-        if (end < MIN_VALUE || end > MAX_VALUE) {
-            throw new IllegalArgumentException("Invalid code point U+" + Utility.hex(end, 6));
-        }
-        if (start <= end) {
-            xor(range(start, end), 2, 0);
-        }
-        return this;
-    }
-
-    /**
-     * Returns true if this set contains the given character.
-     * @param c character to be checked for containment
-     * @return true if the test condition is met
-     * @stable ICU 2.0
-     */
-    public boolean contains(int c) {
-        if (c < MIN_VALUE || c > MAX_VALUE) {
-            throw new IllegalArgumentException("Invalid code point U+" + Utility.hex(c, 6));
-        }
-        if (bmpSet != null) {
-            return bmpSet.contains(c);
-        }
-        if (stringSpan != null) {
-            return stringSpan.contains(c);
-        }
-
-        /*
-        // Set i to the index of the start item greater than ch
-        // We know we will terminate without length test!
-        int i = -1;
-        while (true) {
-            if (c < list[++i]) break;
-        }
-         */
-
-        int i = findCodePoint(c);
-
-        return ((i & 1) != 0); // return true if odd
-    }
-
-    /**
-     * Returns the smallest value i such that c < list[i].  Caller
-     * must ensure that c is a legal value or this method will enter
-     * an infinite loop.  This method performs a binary search.
-     * @param c a character in the range MIN_VALUE..MAX_VALUE
-     * inclusive
-     * @return the smallest integer i in the range 0..len-1,
-     * inclusive, such that c < list[i]
-     */
-    private final int findCodePoint(int c) {
-        /* Examples:
-                                           findCodePoint(c)
-           set              list[]         c=0 1 3 4 7 8
-           ===              ==============   ===========
-           []               [110000]         0 0 0 0 0 0
-           [\u0000-\u0003]  [0, 4, 110000]   1 1 1 2 2 2
-           [\u0004-\u0007]  [4, 8, 110000]   0 0 0 1 1 2
-           [:all:]          [0, 110000]      1 1 1 1 1 1
-         */
-
-        // Return the smallest i such that c < list[i].  Assume
-        // list[len - 1] == HIGH and that c is legal (0..HIGH-1).
-        if (c < list[0]) return 0;
-        // High runner test.  c is often after the last range, so an
-        // initial check for this condition pays off.
-        if (len >= 2 && c >= list[len-2]) return len-1;
-        int lo = 0;
-        int hi = len - 1;
-        // invariant: c >= list[lo]
-        // invariant: c < list[hi]
-        for (;;) {
-            int i = (lo + hi) >>> 1;
-        if (i == lo) return hi;
-            if (c < list[i]) {
-                hi = i;
-            } else {
-                lo = i;
-            }
-        }
-    }
-
-    /**
-     * Retains only the elements in this set that are contained in the
-     * specified set.  In other words, removes from this set all of
-     * its elements that are not contained in the specified set.  This
-     * operation effectively modifies this set so that its value is
-     * the intersection of the two sets.
-     *
-     * @param c set that defines which elements this set will retain.
-     * @stable ICU 2.0
-     */
-    public UnicodeSet retainAll(UnicodeSet c) {
-        checkFrozen();
-        retain(c.list, c.len, 0);
-        strings.retainAll(c.strings);
-        return this;
-    }
-
-    /**
-     * Removes all of the elements from this set.  This set will be
-     * empty after this call returns.
-     * @stable ICU 2.0
-     */
-    public UnicodeSet clear() {
-        checkFrozen();
-        list[0] = HIGH;
-        len = 1;
-        strings.clear();
-        return this;
-    }
-
-    /**
-     * Iteration method that returns the number of ranges contained in
-     * this set.
-     * @see #getRangeStart
-     * @see #getRangeEnd
-     * @stable ICU 2.0
-     */
-    public int getRangeCount() {
-        return len/2;
-    }
-
-    /**
-     * Iteration method that returns the first character in the
-     * specified range of this set.
-     * @exception ArrayIndexOutOfBoundsException if index is outside
-     * the range 0..getRangeCount()-1
-     * @see #getRangeCount
-     * @see #getRangeEnd
-     * @stable ICU 2.0
-     */
-    public int getRangeStart(int index) {
-        return list[index*2];
-    }
-
-    /**
-     * Iteration method that returns the last character in the
-     * specified range of this set.
-     * @exception ArrayIndexOutOfBoundsException if index is outside
-     * the range 0..getRangeCount()-1
-     * @see #getRangeStart
-     * @see #getRangeEnd
-     * @stable ICU 2.0
-     */
-    public int getRangeEnd(int index) {
-        return (list[index*2 + 1] - 1);
-    }
-
-    //----------------------------------------------------------------
-    // Implementation: Pattern parsing
-    //----------------------------------------------------------------
-
-    /**
-     * Parses the given pattern, starting at the given position.  The character
-     * at pattern.charAt(pos.getIndex()) must be '[', or the parse fails.
-     * Parsing continues until the corresponding closing ']'.  If a syntax error
-     * is encountered between the opening and closing brace, the parse fails.
-     * Upon return from a successful parse, the ParsePosition is updated to
-     * point to the character following the closing ']', and an inversion
-     * list for the parsed pattern is returned.  This method
-     * calls itself recursively to parse embedded subpatterns.
-     *
-     * @param pattern the string containing the pattern to be parsed.  The
-     * portion of the string from pos.getIndex(), which must be a '[', to the
-     * corresponding closing ']', is parsed.
-     * @param pos upon entry, the position at which to being parsing.  The
-     * character at pattern.charAt(pos.getIndex()) must be a '['.  Upon return
-     * from a successful parse, pos.getIndex() is either the character after the
-     * closing ']' of the parsed pattern, or pattern.length() if the closing ']'
-     * is the last character of the pattern string.
-     * @return an inversion list for the parsed substring
-     * of pattern
-     * @exception java.lang.IllegalArgumentException if the parse fails.
-     */
-    private UnicodeSet applyPattern(String pattern,
-            ParsePosition pos) {
-        if ("[:age=3.2:]".equals(pattern)) {
-            checkFrozen();
-            VersionInfo version = VersionInfo.getInstance("3.2");
-            applyFilter(new VersionFilter(version), UCharacterProperty.SRC_PROPSVEC);
-        } else {
-            throw new IllegalStateException("UnicodeSet.applyPattern(unexpected pattern "
-                          + pattern + ")");
-        }
-
-        return this;
-    }
-
-    //----------------------------------------------------------------
-    // Implementation: Utility methods
-    //----------------------------------------------------------------
-
-    private void ensureCapacity(int newLen) {
-        if (newLen <= list.length) return;
-        int[] temp = new int[newLen + GROW_EXTRA];
-        System.arraycopy(list, 0, temp, 0, len);
-        list = temp;
-    }
-
-    private void ensureBufferCapacity(int newLen) {
-        if (buffer != null && newLen <= buffer.length) return;
-        buffer = new int[newLen + GROW_EXTRA];
-    }
-
-    /**
-     * Assumes start <= end.
-     */
-    private int[] range(int start, int end) {
-        if (rangeList == null) {
-            rangeList = new int[] { start, end+1, HIGH };
-        } else {
-            rangeList[0] = start;
-            rangeList[1] = end+1;
-        }
-        return rangeList;
-    }
-
-    //----------------------------------------------------------------
-    // Implementation: Fundamental operations
-    //----------------------------------------------------------------
-
-    // polarity = 0, 3 is normal: x xor y
-    // polarity = 1, 2: x xor ~y == x === y
-
-    private UnicodeSet xor(int[] other, int otherLen, int polarity) {
-        ensureBufferCapacity(len + otherLen);
-        int i = 0, j = 0, k = 0;
-        int a = list[i++];
-        int b;
-        if (polarity == 1 || polarity == 2) {
-            b = LOW;
-            if (other[j] == LOW) { // skip base if already LOW
-                ++j;
-                b = other[j];
-            }
-        } else {
-            b = other[j++];
-        }
-        // simplest of all the routines
-        // sort the values, discarding identicals!
-        while (true) {
-            if (a < b) {
-                buffer[k++] = a;
-                a = list[i++];
-            } else if (b < a) {
-                buffer[k++] = b;
-                b = other[j++];
-            } else if (a != HIGH) { // at this point, a == b
-                // discard both values!
-                a = list[i++];
-                b = other[j++];
-            } else { // DONE!
-                buffer[k++] = HIGH;
-                len = k;
-                break;
-            }
-        }
-        // swap list and buffer
-        int[] temp = list;
-        list = buffer;
-        buffer = temp;
-        return this;
-    }
-
-    // polarity = 0 is normal: x union y
-    // polarity = 2: x union ~y
-    // polarity = 1: ~x union y
-    // polarity = 3: ~x union ~y
-
-    private UnicodeSet add(int[] other, int otherLen, int polarity) {
-        ensureBufferCapacity(len + otherLen);
-        int i = 0, j = 0, k = 0;
-        int a = list[i++];
-        int b = other[j++];
-        // change from xor is that we have to check overlapping pairs
-        // polarity bit 1 means a is second, bit 2 means b is.
-        main:
-            while (true) {
-                switch (polarity) {
-                case 0: // both first; take lower if unequal
-                    if (a < b) { // take a
-                        // Back up over overlapping ranges in buffer[]
-                        if (k > 0 && a <= buffer[k-1]) {
-                            // Pick latter end value in buffer[] vs. list[]
-                            a = max(list[i], buffer[--k]);
-                        } else {
-                            // No overlap
-                            buffer[k++] = a;
-                            a = list[i];
-                        }
-                        i++; // Common if/else code factored out
-                        polarity ^= 1;
-                    } else if (b < a) { // take b
-                        if (k > 0 && b <= buffer[k-1]) {
-                            b = max(other[j], buffer[--k]);
-                        } else {
-                            buffer[k++] = b;
-                            b = other[j];
-                        }
-                        j++;
-                        polarity ^= 2;
-                    } else { // a == b, take a, drop b
-                        if (a == HIGH) break main;
-                        // This is symmetrical; it doesn't matter if
-                        // we backtrack with a or b. - liu
-                        if (k > 0 && a <= buffer[k-1]) {
-                            a = max(list[i], buffer[--k]);
-                        } else {
-                            // No overlap
-                            buffer[k++] = a;
-                            a = list[i];
-                        }
-                        i++;
-                        polarity ^= 1;
-                        b = other[j++]; polarity ^= 2;
-                    }
-                    break;
-                case 3: // both second; take higher if unequal, and drop other
-                    if (b <= a) { // take a
-                        if (a == HIGH) break main;
-                        buffer[k++] = a;
-                    } else { // take b
-                        if (b == HIGH) break main;
-                        buffer[k++] = b;
-                    }
-                    a = list[i++]; polarity ^= 1;   // factored common code
-                    b = other[j++]; polarity ^= 2;
-                    break;
-                case 1: // a second, b first; if b < a, overlap
-                    if (a < b) { // no overlap, take a
-                        buffer[k++] = a; a = list[i++]; polarity ^= 1;
-                    } else if (b < a) { // OVERLAP, drop b
-                        b = other[j++]; polarity ^= 2;
-                    } else { // a == b, drop both!
-                        if (a == HIGH) break main;
-                        a = list[i++]; polarity ^= 1;
-                        b = other[j++]; polarity ^= 2;
-                    }
-                    break;
-                case 2: // a first, b second; if a < b, overlap
-                    if (b < a) { // no overlap, take b
-                        buffer[k++] = b; b = other[j++]; polarity ^= 2;
-                    } else  if (a < b) { // OVERLAP, drop a
-                        a = list[i++]; polarity ^= 1;
-                    } else { // a == b, drop both!
-                        if (a == HIGH) break main;
-                        a = list[i++]; polarity ^= 1;
-                        b = other[j++]; polarity ^= 2;
-                    }
-                    break;
-                }
-            }
-        buffer[k++] = HIGH;    // terminate
-        len = k;
-        // swap list and buffer
-        int[] temp = list;
-        list = buffer;
-        buffer = temp;
-        return this;
-    }
-
-    // polarity = 0 is normal: x intersect y
-    // polarity = 2: x intersect ~y == set-minus
-    // polarity = 1: ~x intersect y
-    // polarity = 3: ~x intersect ~y
-
-    private UnicodeSet retain(int[] other, int otherLen, int polarity) {
-        ensureBufferCapacity(len + otherLen);
-        int i = 0, j = 0, k = 0;
-        int a = list[i++];
-        int b = other[j++];
-        // change from xor is that we have to check overlapping pairs
-        // polarity bit 1 means a is second, bit 2 means b is.
-        main:
-            while (true) {
-                switch (polarity) {
-                case 0: // both first; drop the smaller
-                    if (a < b) { // drop a
-                        a = list[i++]; polarity ^= 1;
-                    } else if (b < a) { // drop b
-                        b = other[j++]; polarity ^= 2;
-                    } else { // a == b, take one, drop other
-                        if (a == HIGH) break main;
-                        buffer[k++] = a; a = list[i++]; polarity ^= 1;
-                        b = other[j++]; polarity ^= 2;
-                    }
-                    break;
-                case 3: // both second; take lower if unequal
-                    if (a < b) { // take a
-                        buffer[k++] = a; a = list[i++]; polarity ^= 1;
-                    } else if (b < a) { // take b
-                        buffer[k++] = b; b = other[j++]; polarity ^= 2;
-                    } else { // a == b, take one, drop other
-                        if (a == HIGH) break main;
-                        buffer[k++] = a; a = list[i++]; polarity ^= 1;
-                        b = other[j++]; polarity ^= 2;
-                    }
-                    break;
-                case 1: // a second, b first;
-                    if (a < b) { // NO OVERLAP, drop a
-                        a = list[i++]; polarity ^= 1;
-                    } else if (b < a) { // OVERLAP, take b
-                        buffer[k++] = b; b = other[j++]; polarity ^= 2;
-                    } else { // a == b, drop both!
-                        if (a == HIGH) break main;
-                        a = list[i++]; polarity ^= 1;
-                        b = other[j++]; polarity ^= 2;
-                    }
-                    break;
-                case 2: // a first, b second; if a < b, overlap
-                    if (b < a) { // no overlap, drop b
-                        b = other[j++]; polarity ^= 2;
-                    } else  if (a < b) { // OVERLAP, take a
-                        buffer[k++] = a; a = list[i++]; polarity ^= 1;
-                    } else { // a == b, drop both!
-                        if (a == HIGH) break main;
-                        a = list[i++]; polarity ^= 1;
-                        b = other[j++]; polarity ^= 2;
-                    }
-                    break;
-                }
-            }
-        buffer[k++] = HIGH;    // terminate
-        len = k;
-        // swap list and buffer
-        int[] temp = list;
-        list = buffer;
-        buffer = temp;
-        return this;
-    }
-
-    private static final int max(int a, int b) {
-        return (a > b) ? a : b;
-    }
-
-    //----------------------------------------------------------------
-    // Generic filter-based scanning code
-    //----------------------------------------------------------------
-
-    private static interface Filter {
-        boolean contains(int codePoint);
-    }
-
-    private static final VersionInfo NO_VERSION = VersionInfo.getInstance(0, 0, 0, 0);
-
-    private static class VersionFilter implements Filter {
-        VersionInfo version;
-        VersionFilter(VersionInfo version) { this.version = version; }
-        public boolean contains(int ch) {
-            VersionInfo v = UCharacter.getAge(ch);
-            // Reference comparison ok; VersionInfo caches and reuses
-            // unique objects.
-            return v != NO_VERSION &&
-                    v.compareTo(version) <= 0;
-        }
-    }
-
-    private static synchronized UnicodeSet getInclusions(int src) {
-        if (src != UCharacterProperty.SRC_PROPSVEC) {
-            throw new IllegalStateException("UnicodeSet.getInclusions(unknown src "+src+")");
-        }
-
-        if (INCLUSION == null) {
-            UnicodeSet incl = new UnicodeSet();
-            UCharacterProperty.INSTANCE.upropsvec_addPropertyStarts(incl);
-            INCLUSION = incl;
-        }
-        return INCLUSION;
-    }
-
-    /**
-     * Generic filter-based scanning code for UCD property UnicodeSets.
-     */
-    private UnicodeSet applyFilter(Filter filter, int src) {
-        // Logically, walk through all Unicode characters, noting the start
-        // and end of each range for which filter.contain(c) is
-        // true.  Add each range to a set.
-        //
-        // To improve performance, use an inclusions set which
-        // encodes information about character ranges that are known
-        // to have identical properties.
-        // getInclusions(src) contains exactly the first characters of
-        // same-value ranges for the given properties "source".
-
-        clear();
-
-        int startHasProperty = -1;
-        UnicodeSet inclusions = getInclusions(src);
-        int limitRange = inclusions.getRangeCount();
-
-        for (int j=0; j= 0) {
-                    add_unchecked(startHasProperty, ch-1);
-                    startHasProperty = -1;
-                }
-            }
-        }
-        if (startHasProperty >= 0) {
-            add_unchecked(startHasProperty, 0x10FFFF);
-        }
-
-        return this;
-    }
-
-    /**
-     * Is this frozen, according to the Freezable interface?
-     *
-     * @return value
-     * @stable ICU 3.8
-     */
-    public boolean isFrozen() {
-        return (bmpSet != null || stringSpan != null);
-    }
-
-    /**
-     * Freeze this class, according to the Freezable interface.
-     *
-     * @return this
-     * @stable ICU 4.4
-     */
-    public UnicodeSet freeze() {
-        if (!isFrozen()) {
-            // Do most of what compact() does before freezing because
-            // compact() will not work when the set is frozen.
-            // Small modification: Don't shrink if the savings would be tiny (<=GROW_EXTRA).
-
-            // Delete buffer first to defragment memory less.
-            buffer = null;
-            if (list.length > (len + GROW_EXTRA)) {
-                // Make the capacity equal to len or 1.
-                // We don't want to realloc of 0 size.
-                int capacity = (len == 0) ? 1 : len;
-                int[] oldList = list;
-                list = new int[capacity];
-                for (int i = capacity; i-- > 0;) {
-                    list[i] = oldList[i];
-                }
-            }
-
-            // Optimize contains() and span() and similar functions.
-            if (!strings.isEmpty()) {
-                stringSpan = new UnicodeSetStringSpan(this, new ArrayList(strings), UnicodeSetStringSpan.ALL);
-            }
-            if (stringSpan == null || !stringSpan.needsStringSpanUTF16()) {
-                // Optimize for code point spans.
-                // There are no strings, or
-                // all strings are irrelevant for span() etc. because
-                // all of each string's code points are contained in this set.
-                // However, fully contained strings are relevant for spanAndCount(),
-                // so we create both objects.
-                bmpSet = new BMPSet(list, len);
-            }
-        }
-        return this;
-    }
-
-    /**
-     * Span a string using this UnicodeSet.
-     * 
To replace, count elements, or delete spans, see {@link com.ibm.icu.text.UnicodeSetSpanner UnicodeSetSpanner}.
-     * @param s The string to be spanned
-     * @param spanCondition The span condition
-     * @return the length of the span
-     * @stable ICU 4.4
-     */
-    public int span(CharSequence s, SpanCondition spanCondition) {
-        return span(s, 0, spanCondition);
-    }
-
-    /**
-     * Span a string using this UnicodeSet.
-     *   If the start index is less than 0, span will start from 0.
-     *   If the start index is greater than the string length, span returns the string length.
-     * 
To replace, count elements, or delete spans, see {@link com.ibm.icu.text.UnicodeSetSpanner UnicodeSetSpanner}.
-     * @param s The string to be spanned
-     * @param start The start index that the span begins
-     * @param spanCondition The span condition
-     * @return the string index which ends the span (i.e. exclusive)
-     * @stable ICU 4.4
-     */
-    public int span(CharSequence s, int start, SpanCondition spanCondition) {
-        int end = s.length();
-        if (start < 0) {
-            start = 0;
-        } else if (start >= end) {
-            return end;
-        }
-        if (bmpSet != null) {
-            // Frozen set without strings, or no string is relevant for span().
-            return bmpSet.span(s, start, spanCondition, null);
-        }
-        if (stringSpan != null) {
-            return stringSpan.span(s, start, spanCondition);
-        } else if (!strings.isEmpty()) {
-            int which = spanCondition == SpanCondition.NOT_CONTAINED ? UnicodeSetStringSpan.FWD_UTF16_NOT_CONTAINED
-                    : UnicodeSetStringSpan.FWD_UTF16_CONTAINED;
-            UnicodeSetStringSpan strSpan = new UnicodeSetStringSpan(this, new ArrayList(strings), which);
-            if (strSpan.needsStringSpanUTF16()) {
-                return strSpan.span(s, start, spanCondition);
-            }
-        }
-
-        return spanCodePointsAndCount(s, start, spanCondition, null);
-    }
-
-    /**
-     * Same as span() but also counts the smallest number of set elements on any path across the span.
-     * 
To replace, count elements, or delete spans, see {@link com.ibm.icu.text.UnicodeSetSpanner UnicodeSetSpanner}.
-     * @param outCount An output-only object (must not be null) for returning the count.
-     * @return the limit (exclusive end) of the span
-     */
-    public int spanAndCount(CharSequence s, int start, SpanCondition spanCondition, OutputInt outCount) {
-        if (outCount == null) {
-            throw new IllegalArgumentException("outCount must not be null");
-        }
-        int end = s.length();
-        if (start < 0) {
-            start = 0;
-        } else if (start >= end) {
-            return end;
-        }
-        if (stringSpan != null) {
-            // We might also have bmpSet != null,
-            // but fully-contained strings are relevant for counting elements.
-            return stringSpan.spanAndCount(s, start, spanCondition, outCount);
-        } else if (bmpSet != null) {
-            return bmpSet.span(s, start, spanCondition, outCount);
-        } else if (!strings.isEmpty()) {
-            int which = spanCondition == SpanCondition.NOT_CONTAINED ? UnicodeSetStringSpan.FWD_UTF16_NOT_CONTAINED
-                    : UnicodeSetStringSpan.FWD_UTF16_CONTAINED;
-            which |= UnicodeSetStringSpan.WITH_COUNT;
-            UnicodeSetStringSpan strSpan = new UnicodeSetStringSpan(this, new ArrayList(strings), which);
-            return strSpan.spanAndCount(s, start, spanCondition, outCount);
-        }
-
-        return spanCodePointsAndCount(s, start, spanCondition, outCount);
-    }
-
-    private int spanCodePointsAndCount(CharSequence s, int start,
-            SpanCondition spanCondition, OutputInt outCount) {
-        // Pin to 0/1 values.
-        boolean spanContained = (spanCondition != SpanCondition.NOT_CONTAINED);
-
-        int c;
-        int next = start;
-        int length = s.length();
-        int count = 0;
-        do {
-            c = Character.codePointAt(s, next);
-            if (spanContained != contains(c)) {
-                break;
-            }
-            ++count;
-            next += Character.charCount(c);
-        } while (next < length);
-        if (outCount != null) { outCount.value = count; }
-        return next;
-    }
-
-    /**
-     * Span a string backwards (from the fromIndex) using this UnicodeSet.
-     * If the fromIndex is less than 0, spanBack will return 0.
-     * If fromIndex is greater than the string length, spanBack will start from the string length.
-     * 
To replace, count elements, or delete spans, see {@link com.ibm.icu.text.UnicodeSetSpanner UnicodeSetSpanner}.
-     * @param s The string to be spanned
-     * @param fromIndex The index of the char (exclusive) that the string should be spanned backwards
-     * @param spanCondition The span condition
-     * @return The string index which starts the span (i.e. inclusive).
-     * @stable ICU 4.4
-     */
-    public int spanBack(CharSequence s, int fromIndex, SpanCondition spanCondition) {
-        if (fromIndex <= 0) {
-            return 0;
-        }
-        if (fromIndex > s.length()) {
-            fromIndex = s.length();
-        }
-        if (bmpSet != null) {
-            // Frozen set without strings, or no string is relevant for spanBack().
-            return bmpSet.spanBack(s, fromIndex, spanCondition);
-        }
-        if (stringSpan != null) {
-            return stringSpan.spanBack(s, fromIndex, spanCondition);
-        } else if (!strings.isEmpty()) {
-            int which = (spanCondition == SpanCondition.NOT_CONTAINED)
-                    ? UnicodeSetStringSpan.BACK_UTF16_NOT_CONTAINED
-                            : UnicodeSetStringSpan.BACK_UTF16_CONTAINED;
-            UnicodeSetStringSpan strSpan = new UnicodeSetStringSpan(this, new ArrayList(strings), which);
-            if (strSpan.needsStringSpanUTF16()) {
-                return strSpan.spanBack(s, fromIndex, spanCondition);
-            }
-        }
-
-        // Pin to 0/1 values.
-        boolean spanContained = (spanCondition != SpanCondition.NOT_CONTAINED);
-
-        int c;
-        int prev = fromIndex;
-        do {
-            c = Character.codePointBefore(s, prev);
-            if (spanContained != contains(c)) {
-                break;
-            }
-            prev -= Character.charCount(c);
-        } while (prev > 0);
-        return prev;
-    }
-
-    /**
-     * Clone a thawed version of this class, according to the Freezable interface.
-     * @return the clone, not frozen
-     * @stable ICU 4.4
-     */
-    public UnicodeSet cloneAsThawed() {
-        UnicodeSet result = new UnicodeSet(this);
-        assert !result.isFrozen();
-        return result;
-    }
-
-    // internal function
-    private void checkFrozen() {
-        if (isFrozen()) {
-            throw new UnsupportedOperationException("Attempt to modify frozen object");
-        }
-    }
-
-    /**
-     * Argument values for whether span() and similar functions continue while the current character is contained vs.
-     * not contained in the set.
-     * 

-     * The functionality is straightforward for sets with only single code points, without strings (which is the common
-     * case):
-     * 
    
-     * 
  • CONTAINED and SIMPLE work the same.
-     * 
  • CONTAINED and SIMPLE are inverses of NOT_CONTAINED.
-     * 
  • span() and spanBack() partition any string the
-     * same way when alternating between span(NOT_CONTAINED) and span(either "contained" condition).
-     * 
  • Using a
-     * complemented (inverted) set and the opposite span conditions yields the same results.
-     * 

-     * When a set contains multi-code point strings, then these statements may not be true, depending on the strings in
-     * the set (for example, whether they overlap with each other) and the string that is processed. For a set with
-     * strings:
-     * 
    
-     * 
  • The complement of the set contains the opposite set of code points, but the same set of strings.
-     * Therefore, complementing both the set and the span conditions may yield different results.
-     * 
  • When starting spans
-     * at different positions in a string (span(s, ...) vs. span(s+1, ...)) the ends of the spans may be different
-     * because a set string may start before the later position.
-     * 
  • span(SIMPLE) may be shorter than
-     * span(CONTAINED) because it will not recursively try all possible paths. For example, with a set which
-     * contains the three strings "xy", "xya" and "ax", span("xyax", CONTAINED) will return 4 but span("xyax",
-     * SIMPLE) will return 3. span(SIMPLE) will never be longer than span(CONTAINED).
-     * 
  • With either "contained" condition, span() and spanBack() may partition a string in different ways. For example,
-     * with a set which contains the two strings "ab" and "ba", and when processing the string "aba", span() will yield
-     * contained/not-contained boundaries of { 0, 2, 3 } while spanBack() will yield boundaries of { 0, 1, 3 }.
-     * 

-     * Note: If it is important to get the same boundaries whether iterating forward or backward through a string, then
-     * either only span() should be used and the boundaries cached for backward operation, or an ICU BreakIterator could
-     * be used.
-     * 

-     * Note: Unpaired surrogates are treated like surrogate code points. Similarly, set strings match only on code point
-     * boundaries, never in the middle of a surrogate pair.
-     *
-     * @stable ICU 4.4
-     */
-    public enum SpanCondition {
-        /**
-         * Continues a span() while there is no set element at the current position.
-         * Increments by one code point at a time.
-         * Stops before the first set element (character or string).
-         * (For code points only, this is like while contains(current)==false).
-         * 

-         * When span() returns, the substring between where it started and the position it returned consists only of
-         * characters that are not in the set, and none of its strings overlap with the span.
-         *
-         * @stable ICU 4.4
-         */
-        NOT_CONTAINED,
-
-        /**
-         * Spans the longest substring that is a concatenation of set elements (characters or strings).
-         * (For characters only, this is like while contains(current)==true).
-         * 

-         * When span() returns, the substring between where it started and the position it returned consists only of set
-         * elements (characters or strings) that are in the set.
-         * 

-         * If a set contains strings, then the span will be the longest substring for which there
-         * exists at least one non-overlapping concatenation of set elements (characters or strings).
-         * This is equivalent to a POSIX regular expression for (OR of each set element)*.
-         * (Java/ICU/Perl regex stops at the first match of an OR.)
-         *
-         * @stable ICU 4.4
-         */
-        CONTAINED,
-
-        /**
-         * Continues a span() while there is a set element at the current position.
-         * Increments by the longest matching element at each position.
-         * (For characters only, this is like while contains(current)==true).
-         * 

-         * When span() returns, the substring between where it started and the position it returned consists only of set
-         * elements (characters or strings) that are in the set.
-         * 

-         * If a set only contains single characters, then this is the same as CONTAINED.
-         * 

-         * If a set contains strings, then the span will be the longest substring with a match at each position with the
-         * longest single set element (character or string).
-         * 

-         * Use this span condition together with other longest-match algorithms, such as ICU converters
-         * (ucnv_getUnicodeSet()).
-         *
-         * @stable ICU 4.4
-         */
-        SIMPLE,
-    }
-
-}
--- /dev/null	2020-01-10 15:58:09.000000000 -0800
+++ new/src/java.base/share/classes/jdk/internal/icu/text/UnicodeSet.java	2020-01-10 15:58:08.000000000 -0800
@@ -0,0 +1,1414 @@
+/*
+ * Copyright (c) 2005, 2020, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.  Oracle designates this
+ * particular file as subject to the "Classpath" exception as provided
+ * by Oracle in the LICENSE file that accompanied this code.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+/*
+ *******************************************************************************
+ * Copyright (C) 1996-2015, International Business Machines Corporation and
+ * others. All Rights Reserved.
+ *******************************************************************************
+ */
+package jdk.internal.icu.text;
+
+import java.text.ParsePosition;
+import java.util.ArrayList;
+import java.util.TreeSet;
+
+import jdk.internal.icu.impl.BMPSet;
+import jdk.internal.icu.impl.UCharacterProperty;
+import jdk.internal.icu.impl.UnicodeSetStringSpan;
+import jdk.internal.icu.impl.Utility;
+import jdk.internal.icu.lang.UCharacter;
+import jdk.internal.icu.util.OutputInt;
+import jdk.internal.icu.util.VersionInfo;
+
+/**
+ * A mutable set of Unicode characters and multicharacter strings.
+ * Objects of this class represent character classes used
+ * in regular expressions. A character specifies a subset of Unicode
+ * code points.  Legal code points are U+0000 to U+10FFFF, inclusive.
+ *
+ * Note: method freeze() will not only make the set immutable, but
+ * also makes important methods much higher performance:
+ * contains(c), containsNone(...), span(...), spanBack(...) etc.
+ * After the object is frozen, any subsequent call that wants to change
+ * the object will throw UnsupportedOperationException.
+ *
+ * 
The UnicodeSet class is not designed to be subclassed.
+ *
+ * 
UnicodeSet supports two APIs. The first is the
+ * operand API that allows the caller to modify the value of
+ * a UnicodeSet object. It conforms to Java 2's
+ * java.util.Set interface, although
+ * UnicodeSet does not actually implement that
+ * interface. All methods of Set are supported, with the
+ * modification that they take a character range or single character
+ * instead of an Object, and they take a
+ * UnicodeSet instead of a Collection.  The
+ * operand API may be thought of in terms of boolean logic: a boolean
+ * OR is implemented by add, a boolean AND is implemented
+ * by retain, a boolean XOR is implemented by
+ * complement taking an argument, and a boolean NOT is
+ * implemented by complement with no argument.  In terms
+ * of traditional set theory function names, add is a
+ * union, retain is an intersection, remove
+ * is an asymmetric difference, and complement with no
+ * argument is a set complement with respect to the superset range
+ * MIN_VALUE-MAX_VALUE
+ *
+ * 
The second API is the
+ * applyPattern()/toPattern() API from the
+ * java.text.Format-derived classes.  Unlike the
+ * methods that add characters, add categories, and control the logic
+ * of the set, the method applyPattern() sets all
+ * attributes of a UnicodeSet at once, based on a
+ * string pattern.
+ *
+ * 
Pattern syntax

+ *
+ * Patterns are accepted by the constructors and the
+ * applyPattern() methods and returned by the
+ * toPattern() method.  These patterns follow a syntax
+ * similar to that employed by version 8 regular expression character
+ * classes.  Here are some simple examples:
+ *
+ * 

+ *   
+ *     
+ *       
+ *       
+ *     
+ *       
+ *       
+ *     
+ *       
+ *       
+ *     
+ *     
+ *       
+ *       
+ *     
+ *     
+ *       
+ *       
+ *     
+ *     
+ *       
+ *       
+ *     
+ *     
+ *       
+ *       
+ *     
+ *   
[]No characters
[a]The character 'a'
[ae]The characters 'a' and 'e'
[a-e]The characters 'a' through 'e' inclusive, in Unicode code
+ *       point order
[\\u4E01]The character U+4E01
[a{ab}{ac}]The character 'a' and the multicharacter strings "ab" and
+ *       "ac"
[\p{Lu}]All characters in the general category Uppercase Letter

+ * 

+ *
+ * Any character may be preceded by a backslash in order to remove any special
+ * meaning.  White space characters, as defined by the Unicode Pattern_White_Space property, are
+ * ignored, unless they are escaped.
+ *
+ * 
Property patterns specify a set of characters having a certain
+ * property as defined by the Unicode standard.  Both the POSIX-like
+ * "[:Lu:]" and the Perl-like syntax "\p{Lu}" are recognized.  For a
+ * complete list of supported property patterns, see the User's Guide
+ * for UnicodeSet at
+ * 
+ * http://www.icu-project.org/userguide/unicodeSet.html.
+ * Actual determination of property data is defined by the underlying
+ * Unicode database as implemented by UCharacter.
+ *
+ * 
Patterns specify individual characters, ranges of characters, and
+ * Unicode property sets.  When elements are concatenated, they
+ * specify their union.  To complement a set, place a '^' immediately
+ * after the opening '['.  Property patterns are inverted by modifying
+ * their delimiters; "[:^foo]" and "\P{foo}".  In any other location,
+ * '^' has no special meaning.
+ *
+ * 
Ranges are indicated by placing two a '-' between two
+ * characters, as in "a-z".  This specifies the range of all
+ * characters from the left to the right, in Unicode order.  If the
+ * left character is greater than or equal to the
+ * right character it is a syntax error.  If a '-' occurs as the first
+ * character after the opening '[' or '[^', or if it occurs as the
+ * last character before the closing ']', then it is taken as a
+ * literal.  Thus "[a\\-b]", "[-ab]", and "[ab-]" all indicate the same
+ * set of three characters, 'a', 'b', and '-'.
+ *
+ * 
Sets may be intersected using the {@literal '&'} operator or the asymmetric
+ * set difference may be taken using the '-' operator, for example,
+ * "{@code [[:L:]&[\\u0000-\\u0FFF]]}" indicates the set of all Unicode letters
+ * with values less than 4096.  Operators ({@literal '&'} and '|') have equal
+ * precedence and bind left-to-right.  Thus
+ * "[[:L:]-[a-z]-[\\u0100-\\u01FF]]" is equivalent to
+ * "[[[:L:]-[a-z]]-[\\u0100-\\u01FF]]".  This only really matters for
+ * difference; intersection is commutative.
+ *
+ * 
+ * 
[a]The set containing 'a'
+ * 
[a-z]The set containing 'a'
+ * through 'z' and all letters in between, in Unicode order
+ * 
[^a-z]The set containing
+ * all characters but 'a' through 'z',
+ * that is, U+0000 through 'a'-1 and 'z'+1 through U+10FFFF
+ * 
[[pat1][pat2]]
+ * The union of sets specified by pat1 and pat2
+ * 
[[pat1]&[pat2]]
+ * The intersection of sets specified by pat1 and pat2
+ * 
[[pat1]-[pat2]]
+ * The asymmetric difference of sets specified by pat1 and
+ * pat2
+ * 
[:Lu:] or \p{Lu}
+ * The set of characters having the specified
+ * Unicode property; in
+ * this case, Unicode uppercase letters
+ * 
[:^Lu:] or \P{Lu}
+ * The set of characters not having the given
+ * Unicode property
+ * 

+ *
+ * 
Warning: you cannot add an empty string ("") to a UnicodeSet.

+ *
+ * 
Formal syntax

+ *
+ * 

+ *   
+ *     
+ *       
+ *       
+ *     
+ *     
+ *       
+ *       
+ *     
+ *     
+ *       
+ *       
+ *     
+ *     
+ *       
+ *       
+ *     
+ *     
+ *       
+ *       
+ *     
+ *     
+ *       
+ *       
+ *     
+ *     
+ *       
+ *       
+ *     
+ *     
+ *       
+ *       
+ *     
+ *   
pattern :=  ('[' '^'? item* ']') |
+ *       property
item :=  char | (char '-' char) | pattern-expr

+ *       
pattern-expr :=  pattern | pattern-expr pattern |
+ *       pattern-expr op pattern

+ *       
op :=  '&' | '-'

+ *       
special :=  '[' | ']' | '-'

+ *       
char :=  any character that is not special

+ *       | ('\\' any character)

+ *       | ('\u' hex hex hex hex)

+ *       
hex :=  any character for which
+ *       Character.digit(c, 16)
+ *       returns a non-negative result
property :=  a Unicode property set pattern

+ *   

+ *   
+ *     
+ *       
+ *     
+ *   
Legend: 
+ *         
+ *           
+ *           
+ *           
+ *         
+ *         
+ *           
+ *           
+ *           
+ *         
+ *         
+ *           
+ *           
+ *           
+ *         
+ *         
+ *           
+ *           
+ *           
+ *         
+ *         
+ *           
+ *           
+ *           
+ *         
+ *       
a := b  a may be replaced by b 
a?zero or one instance of a

+ *           
a*one or more instances of a

+ *           
a | beither a or b

+ *           
'a'the literal string between the quotes 

+ *       

+ * 

+ * 
To iterate over contents of UnicodeSet, the following are available:
+ * 
  • {@link #ranges()} to iterate through the ranges
    • 
+ * 
  • {@link #strings()} to iterate through the strings
    • 
+ * 
  • {@link #iterator()} to iterate through the entire contents in a single loop.
+ * That method is, however, not particularly efficient, since it "boxes" each code point into a String.
+ * 

+ * All of the above can be used in for loops.
+ * The {@link com.ibm.icu.text.UnicodeSetIterator UnicodeSetIterator} can also be used, but not in for loops.
+ * 
To replace, count elements, or delete spans, see {@link com.ibm.icu.text.UnicodeSetSpanner UnicodeSetSpanner}.
+ *
+ * @author Alan Liu
+ * @stable ICU 2.0
+ */
+public class UnicodeSet {
+
+    private static final int LOW = 0x000000; // LOW <= all valid values. ZERO for codepoints
+    private static final int HIGH = 0x110000; // HIGH > all valid values. 10000 for code units.
+    // 110000 for codepoints
+
+    /**
+     * Minimum value that can be stored in a UnicodeSet.
+     * @stable ICU 2.0
+     */
+    public static final int MIN_VALUE = LOW;
+
+    /**
+     * Maximum value that can be stored in a UnicodeSet.
+     * @stable ICU 2.0
+     */
+    public static final int MAX_VALUE = HIGH - 1;
+
+    private int len;      // length used; list may be longer to minimize reallocs
+    private int[] list;   // MUST be terminated with HIGH
+    private int[] rangeList; // internal buffer
+    private int[] buffer; // internal buffer
+
+    // NOTE: normally the field should be of type SortedSet; but that is missing a public clone!!
+    // is not private so that UnicodeSetIterator can get access
+    TreeSet strings = new TreeSet();
+
+    /**
+     * The pattern representation of this set.  This may not be the
+     * most economical pattern.  It is the pattern supplied to
+     * applyPattern(), with variables substituted and whitespace
+     * removed.  For sets constructed without applyPattern(), or
+     * modified using the non-pattern API, this string will be null,
+     * indicating that toPattern() must generate a pattern
+     * representation from the inversion list.
+     */
+
+    private static final int START_EXTRA = 16;         // initial storage. Must be >= 0
+    private static final int GROW_EXTRA = START_EXTRA; // extra amount for growth. Must be >= 0
+
+    private static UnicodeSet INCLUSION = null;
+
+    private volatile BMPSet bmpSet; // The set is frozen if bmpSet or stringSpan is not null.
+    private volatile UnicodeSetStringSpan stringSpan;
+
+    //----------------------------------------------------------------
+    // Public API
+    //----------------------------------------------------------------
+
+    /**
+     * Constructs an empty set.
+     * @stable ICU 2.0
+     */
+    private UnicodeSet() {
+        list = new int[1 + START_EXTRA];
+        list[len++] = HIGH;
+    }
+
+    /**
+     * Constructs a copy of an existing set.
+     * @stable ICU 2.0
+     */
+    private UnicodeSet(UnicodeSet other) {
+        set(other);
+    }
+
+    /**
+     * Constructs a set containing the given range. If end >
+     * start then an empty set is created.
+     *
+     * @param start first character, inclusive, of range
+     * @param end last character, inclusive, of range
+     * @stable ICU 2.0
+     */
+    public UnicodeSet(int start, int end) {
+        this();
+        complement(start, end);
+    }
+
+    /**
+     * Constructs a set from the given pattern.  See the class description
+     * for the syntax of the pattern language.  Whitespace is ignored.
+     * @param pattern a string specifying what characters are in the set
+     * @exception java.lang.IllegalArgumentException if the pattern contains
+     * a syntax error.
+     * @stable ICU 2.0
+     */
+    public UnicodeSet(String pattern) {
+        this();
+        applyPattern(pattern, null);
+    }
+
+    /**
+     * Make this object represent the same set as other.
+     * @param other a UnicodeSet whose value will be
+     * copied to this object
+     * @stable ICU 2.0
+     */
+    public UnicodeSet set(UnicodeSet other) {
+        checkFrozen();
+        list = other.list.clone();
+        len = other.len;
+        strings = new TreeSet(other.strings);
+        return this;
+    }
+
+    /**
+     * Returns the number of elements in this set (its cardinality)
+     * Note than the elements of a set may include both individual
+     * codepoints and strings.
+     *
+     * @return the number of elements in this set (its cardinality).
+     * @stable ICU 2.0
+     */
+    public int size() {
+        int n = 0;
+        int count = getRangeCount();
+        for (int i = 0; i < count; ++i) {
+            n += getRangeEnd(i) - getRangeStart(i) + 1;
+        }
+        return n + strings.size();
+    }
+
+    // for internal use, after checkFrozen has been called
+    private UnicodeSet add_unchecked(int start, int end) {
+        if (start < MIN_VALUE || start > MAX_VALUE) {
+            throw new IllegalArgumentException("Invalid code point U+" + Utility.hex(start, 6));
+        }
+        if (end < MIN_VALUE || end > MAX_VALUE) {
+            throw new IllegalArgumentException("Invalid code point U+" + Utility.hex(end, 6));
+        }
+        if (start < end) {
+            add(range(start, end), 2, 0);
+        } else if (start == end) {
+            add(start);
+        }
+        return this;
+    }
+
+    /**
+     * Adds the specified character to this set if it is not already
+     * present.  If this set already contains the specified character,
+     * the call leaves this set unchanged.
+     * @stable ICU 2.0
+     */
+    public final UnicodeSet add(int c) {
+        checkFrozen();
+        return add_unchecked(c);
+    }
+
+    // for internal use only, after checkFrozen has been called
+    private final UnicodeSet add_unchecked(int c) {
+        if (c < MIN_VALUE || c > MAX_VALUE) {
+            throw new IllegalArgumentException("Invalid code point U+" + Utility.hex(c, 6));
+        }
+
+        // find smallest i such that c < list[i]
+        // if odd, then it is IN the set
+        // if even, then it is OUT of the set
+        int i = findCodePoint(c);
+
+        // already in set?
+        if ((i & 1) != 0) return this;
+
+        // HIGH is 0x110000
+        // assert(list[len-1] == HIGH);
+
+        // empty = [HIGH]
+        // [start_0, limit_0, start_1, limit_1, HIGH]
+
+        // [..., start_k-1, limit_k-1, start_k, limit_k, ..., HIGH]
+        //                             ^
+        //                             list[i]
+
+        // i == 0 means c is before the first range
+
+        if (c == list[i]-1) {
+            // c is before start of next range
+            list[i] = c;
+            // if we touched the HIGH mark, then add a new one
+            if (c == MAX_VALUE) {
+                ensureCapacity(len+1);
+                list[len++] = HIGH;
+            }
+            if (i > 0 && c == list[i-1]) {
+                // collapse adjacent ranges
+
+                // [..., start_k-1, c, c, limit_k, ..., HIGH]
+                //                     ^
+                //                     list[i]
+                System.arraycopy(list, i+1, list, i-1, len-i-1);
+                len -= 2;
+            }
+        }
+
+        else if (i > 0 && c == list[i-1]) {
+            // c is after end of prior range
+            list[i-1]++;
+            // no need to chcek for collapse here
+        }
+
+        else {
+            // At this point we know the new char is not adjacent to
+            // any existing ranges, and it is not 10FFFF.
+
+
+            // [..., start_k-1, limit_k-1, start_k, limit_k, ..., HIGH]
+            //                             ^
+            //                             list[i]
+
+            // [..., start_k-1, limit_k-1, c, c+1, start_k, limit_k, ..., HIGH]
+            //                             ^
+            //                             list[i]
+
+            // Don't use ensureCapacity() to save on copying.
+            // NOTE: This has no measurable impact on performance,
+            // but it might help in some usage patterns.
+            if (len+2 > list.length) {
+                int[] temp = new int[len + 2 + GROW_EXTRA];
+                if (i != 0) System.arraycopy(list, 0, temp, 0, i);
+                System.arraycopy(list, i, temp, i+2, len-i);
+                list = temp;
+            } else {
+                System.arraycopy(list, i, list, i+2, len-i);
+            }
+
+            list[i] = c;
+            list[i+1] = c+1;
+            len += 2;
+        }
+
+        return this;
+    }
+
+    /**
+     * Adds the specified multicharacter to this set if it is not already
+     * present.  If this set already contains the multicharacter,
+     * the call leaves this set unchanged.
+     * Thus {@code "ch" => {"ch"}}
+     * 
Warning: you cannot add an empty string ("") to a UnicodeSet.
+     * @param s the source string
+     * @return this object, for chaining
+     * @stable ICU 2.0
+     */
+    public final UnicodeSet add(CharSequence s) {
+        checkFrozen();
+        int cp = getSingleCP(s);
+        if (cp < 0) {
+            strings.add(s.toString());
+        } else {
+            add_unchecked(cp, cp);
+        }
+        return this;
+    }
+
+    /**
+     * Utility for getting code point from single code point CharSequence.
+     * See the public UTF16.getSingleCodePoint()
+     * @return a code point IF the string consists of a single one.
+     * otherwise returns -1.
+     * @param s to test
+     */
+    private static int getSingleCP(CharSequence s) {
+        if (s.length() < 1) {
+            throw new IllegalArgumentException("Can't use zero-length strings in UnicodeSet");
+        }
+        if (s.length() > 2) return -1;
+        if (s.length() == 1) return s.charAt(0);
+
+        // at this point, len = 2
+        int cp = UTF16.charAt(s, 0);
+        if (cp > 0xFFFF) { // is surrogate pair
+            return cp;
+        }
+        return -1;
+    }
+
+    /**
+     * Complements the specified range in this set.  Any character in
+     * the range will be removed if it is in this set, or will be
+     * added if it is not in this set.  If {@code end > start}
+     * then an empty range is complemented, leaving the set unchanged.
+     *
+     * @param start first character, inclusive, of range to be removed
+     * from this set.
+     * @param end last character, inclusive, of range to be removed
+     * from this set.
+     * @stable ICU 2.0
+     */
+    public UnicodeSet complement(int start, int end) {
+        checkFrozen();
+        if (start < MIN_VALUE || start > MAX_VALUE) {
+            throw new IllegalArgumentException("Invalid code point U+" + Utility.hex(start, 6));
+        }
+        if (end < MIN_VALUE || end > MAX_VALUE) {
+            throw new IllegalArgumentException("Invalid code point U+" + Utility.hex(end, 6));
+        }
+        if (start <= end) {
+            xor(range(start, end), 2, 0);
+        }
+        return this;
+    }
+
+    /**
+     * Returns true if this set contains the given character.
+     * @param c character to be checked for containment
+     * @return true if the test condition is met
+     * @stable ICU 2.0
+     */
+    public boolean contains(int c) {
+        if (c < MIN_VALUE || c > MAX_VALUE) {
+            throw new IllegalArgumentException("Invalid code point U+" + Utility.hex(c, 6));
+        }
+        if (bmpSet != null) {
+            return bmpSet.contains(c);
+        }
+        if (stringSpan != null) {
+            return stringSpan.contains(c);
+        }
+
+        /*
+        // Set i to the index of the start item greater than ch
+        // We know we will terminate without length test!
+        int i = -1;
+        while (true) {
+            if (c < list[++i]) break;
+        }
+         */
+
+        int i = findCodePoint(c);
+
+        return ((i & 1) != 0); // return true if odd
+    }
+
+    /**
+     * Returns the smallest value i such that c < list[i].  Caller
+     * must ensure that c is a legal value or this method will enter
+     * an infinite loop.  This method performs a binary search.
+     * @param c a character in the range MIN_VALUE..MAX_VALUE
+     * inclusive
+     * @return the smallest integer i in the range 0..len-1,
+     * inclusive, such that c < list[i]
+     */
+    private final int findCodePoint(int c) {
+        /* Examples:
+                                           findCodePoint(c)
+           set              list[]         c=0 1 3 4 7 8
+           ===              ==============   ===========
+           []               [110000]         0 0 0 0 0 0
+           [\u0000-\u0003]  [0, 4, 110000]   1 1 1 2 2 2
+           [\u0004-\u0007]  [4, 8, 110000]   0 0 0 1 1 2
+           [:all:]          [0, 110000]      1 1 1 1 1 1
+         */
+
+        // Return the smallest i such that c < list[i].  Assume
+        // list[len - 1] == HIGH and that c is legal (0..HIGH-1).
+        if (c < list[0]) return 0;
+        // High runner test.  c is often after the last range, so an
+        // initial check for this condition pays off.
+        if (len >= 2 && c >= list[len-2]) return len-1;
+        int lo = 0;
+        int hi = len - 1;
+        // invariant: c >= list[lo]
+        // invariant: c < list[hi]
+        for (;;) {
+            int i = (lo + hi) >>> 1;
+        if (i == lo) return hi;
+            if (c < list[i]) {
+                hi = i;
+            } else {
+                lo = i;
+            }
+        }
+    }
+
+    /**
+     * Retains only the elements in this set that are contained in the
+     * specified set.  In other words, removes from this set all of
+     * its elements that are not contained in the specified set.  This
+     * operation effectively modifies this set so that its value is
+     * the intersection of the two sets.
+     *
+     * @param c set that defines which elements this set will retain.
+     * @stable ICU 2.0
+     */
+    public UnicodeSet retainAll(UnicodeSet c) {
+        checkFrozen();
+        retain(c.list, c.len, 0);
+        strings.retainAll(c.strings);
+        return this;
+    }
+
+    /**
+     * Removes all of the elements from this set.  This set will be
+     * empty after this call returns.
+     * @stable ICU 2.0
+     */
+    public UnicodeSet clear() {
+        checkFrozen();
+        list[0] = HIGH;
+        len = 1;
+        strings.clear();
+        return this;
+    }
+
+    /**
+     * Iteration method that returns the number of ranges contained in
+     * this set.
+     * @see #getRangeStart
+     * @see #getRangeEnd
+     * @stable ICU 2.0
+     */
+    public int getRangeCount() {
+        return len/2;
+    }
+
+    /**
+     * Iteration method that returns the first character in the
+     * specified range of this set.
+     * @exception ArrayIndexOutOfBoundsException if index is outside
+     * the range 0..getRangeCount()-1
+     * @see #getRangeCount
+     * @see #getRangeEnd
+     * @stable ICU 2.0
+     */
+    public int getRangeStart(int index) {
+        return list[index*2];
+    }
+
+    /**
+     * Iteration method that returns the last character in the
+     * specified range of this set.
+     * @exception ArrayIndexOutOfBoundsException if index is outside
+     * the range 0..getRangeCount()-1
+     * @see #getRangeStart
+     * @see #getRangeEnd
+     * @stable ICU 2.0
+     */
+    public int getRangeEnd(int index) {
+        return (list[index*2 + 1] - 1);
+    }
+
+    //----------------------------------------------------------------
+    // Implementation: Pattern parsing
+    //----------------------------------------------------------------
+
+    /**
+     * Parses the given pattern, starting at the given position.  The character
+     * at pattern.charAt(pos.getIndex()) must be '[', or the parse fails.
+     * Parsing continues until the corresponding closing ']'.  If a syntax error
+     * is encountered between the opening and closing brace, the parse fails.
+     * Upon return from a successful parse, the ParsePosition is updated to
+     * point to the character following the closing ']', and an inversion
+     * list for the parsed pattern is returned.  This method
+     * calls itself recursively to parse embedded subpatterns.
+     *
+     * @param pattern the string containing the pattern to be parsed.  The
+     * portion of the string from pos.getIndex(), which must be a '[', to the
+     * corresponding closing ']', is parsed.
+     * @param pos upon entry, the position at which to being parsing.  The
+     * character at pattern.charAt(pos.getIndex()) must be a '['.  Upon return
+     * from a successful parse, pos.getIndex() is either the character after the
+     * closing ']' of the parsed pattern, or pattern.length() if the closing ']'
+     * is the last character of the pattern string.
+     * @return an inversion list for the parsed substring
+     * of pattern
+     * @exception java.lang.IllegalArgumentException if the parse fails.
+     */
+    private UnicodeSet applyPattern(String pattern,
+            ParsePosition pos) {
+        if ("[:age=3.2:]".equals(pattern)) {
+            checkFrozen();
+            VersionInfo version = VersionInfo.getInstance("3.2");
+            applyFilter(new VersionFilter(version), UCharacterProperty.SRC_PROPSVEC);
+        } else {
+            throw new IllegalStateException("UnicodeSet.applyPattern(unexpected pattern "
+                          + pattern + ")");
+        }
+
+        return this;
+    }
+
+    //----------------------------------------------------------------
+    // Implementation: Utility methods
+    //----------------------------------------------------------------
+
+    private void ensureCapacity(int newLen) {
+        if (newLen <= list.length) return;
+        int[] temp = new int[newLen + GROW_EXTRA];
+        System.arraycopy(list, 0, temp, 0, len);
+        list = temp;
+    }
+
+    private void ensureBufferCapacity(int newLen) {
+        if (buffer != null && newLen <= buffer.length) return;
+        buffer = new int[newLen + GROW_EXTRA];
+    }
+
+    /**
+     * Assumes start <= end.
+     */
+    private int[] range(int start, int end) {
+        if (rangeList == null) {
+            rangeList = new int[] { start, end+1, HIGH };
+        } else {
+            rangeList[0] = start;
+            rangeList[1] = end+1;
+        }
+        return rangeList;
+    }
+
+    //----------------------------------------------------------------
+    // Implementation: Fundamental operations
+    //----------------------------------------------------------------
+
+    // polarity = 0, 3 is normal: x xor y
+    // polarity = 1, 2: x xor ~y == x === y
+
+    private UnicodeSet xor(int[] other, int otherLen, int polarity) {
+        ensureBufferCapacity(len + otherLen);
+        int i = 0, j = 0, k = 0;
+        int a = list[i++];
+        int b;
+        if (polarity == 1 || polarity == 2) {
+            b = LOW;
+            if (other[j] == LOW) { // skip base if already LOW
+                ++j;
+                b = other[j];
+            }
+        } else {
+            b = other[j++];
+        }
+        // simplest of all the routines
+        // sort the values, discarding identicals!
+        while (true) {
+            if (a < b) {
+                buffer[k++] = a;
+                a = list[i++];
+            } else if (b < a) {
+                buffer[k++] = b;
+                b = other[j++];
+            } else if (a != HIGH) { // at this point, a == b
+                // discard both values!
+                a = list[i++];
+                b = other[j++];
+            } else { // DONE!
+                buffer[k++] = HIGH;
+                len = k;
+                break;
+            }
+        }
+        // swap list and buffer
+        int[] temp = list;
+        list = buffer;
+        buffer = temp;
+        return this;
+    }
+
+    // polarity = 0 is normal: x union y
+    // polarity = 2: x union ~y
+    // polarity = 1: ~x union y
+    // polarity = 3: ~x union ~y
+
+    private UnicodeSet add(int[] other, int otherLen, int polarity) {
+        ensureBufferCapacity(len + otherLen);
+        int i = 0, j = 0, k = 0;
+        int a = list[i++];
+        int b = other[j++];
+        // change from xor is that we have to check overlapping pairs
+        // polarity bit 1 means a is second, bit 2 means b is.
+        main:
+            while (true) {
+                switch (polarity) {
+                case 0: // both first; take lower if unequal
+                    if (a < b) { // take a
+                        // Back up over overlapping ranges in buffer[]
+                        if (k > 0 && a <= buffer[k-1]) {
+                            // Pick latter end value in buffer[] vs. list[]
+                            a = max(list[i], buffer[--k]);
+                        } else {
+                            // No overlap
+                            buffer[k++] = a;
+                            a = list[i];
+                        }
+                        i++; // Common if/else code factored out
+                        polarity ^= 1;
+                    } else if (b < a) { // take b
+                        if (k > 0 && b <= buffer[k-1]) {
+                            b = max(other[j], buffer[--k]);
+                        } else {
+                            buffer[k++] = b;
+                            b = other[j];
+                        }
+                        j++;
+                        polarity ^= 2;
+                    } else { // a == b, take a, drop b
+                        if (a == HIGH) break main;
+                        // This is symmetrical; it doesn't matter if
+                        // we backtrack with a or b. - liu
+                        if (k > 0 && a <= buffer[k-1]) {
+                            a = max(list[i], buffer[--k]);
+                        } else {
+                            // No overlap
+                            buffer[k++] = a;
+                            a = list[i];
+                        }
+                        i++;
+                        polarity ^= 1;
+                        b = other[j++]; polarity ^= 2;
+                    }
+                    break;
+                case 3: // both second; take higher if unequal, and drop other
+                    if (b <= a) { // take a
+                        if (a == HIGH) break main;
+                        buffer[k++] = a;
+                    } else { // take b
+                        if (b == HIGH) break main;
+                        buffer[k++] = b;
+                    }
+                    a = list[i++]; polarity ^= 1;   // factored common code
+                    b = other[j++]; polarity ^= 2;
+                    break;
+                case 1: // a second, b first; if b < a, overlap
+                    if (a < b) { // no overlap, take a
+                        buffer[k++] = a; a = list[i++]; polarity ^= 1;
+                    } else if (b < a) { // OVERLAP, drop b
+                        b = other[j++]; polarity ^= 2;
+                    } else { // a == b, drop both!
+                        if (a == HIGH) break main;
+                        a = list[i++]; polarity ^= 1;
+                        b = other[j++]; polarity ^= 2;
+                    }
+                    break;
+                case 2: // a first, b second; if a < b, overlap
+                    if (b < a) { // no overlap, take b
+                        buffer[k++] = b; b = other[j++]; polarity ^= 2;
+                    } else  if (a < b) { // OVERLAP, drop a
+                        a = list[i++]; polarity ^= 1;
+                    } else { // a == b, drop both!
+                        if (a == HIGH) break main;
+                        a = list[i++]; polarity ^= 1;
+                        b = other[j++]; polarity ^= 2;
+                    }
+                    break;
+                }
+            }
+        buffer[k++] = HIGH;    // terminate
+        len = k;
+        // swap list and buffer
+        int[] temp = list;
+        list = buffer;
+        buffer = temp;
+        return this;
+    }
+
+    // polarity = 0 is normal: x intersect y
+    // polarity = 2: x intersect ~y == set-minus
+    // polarity = 1: ~x intersect y
+    // polarity = 3: ~x intersect ~y
+
+    private UnicodeSet retain(int[] other, int otherLen, int polarity) {
+        ensureBufferCapacity(len + otherLen);
+        int i = 0, j = 0, k = 0;
+        int a = list[i++];
+        int b = other[j++];
+        // change from xor is that we have to check overlapping pairs
+        // polarity bit 1 means a is second, bit 2 means b is.
+        main:
+            while (true) {
+                switch (polarity) {
+                case 0: // both first; drop the smaller
+                    if (a < b) { // drop a
+                        a = list[i++]; polarity ^= 1;
+                    } else if (b < a) { // drop b
+                        b = other[j++]; polarity ^= 2;
+                    } else { // a == b, take one, drop other
+                        if (a == HIGH) break main;
+                        buffer[k++] = a; a = list[i++]; polarity ^= 1;
+                        b = other[j++]; polarity ^= 2;
+                    }
+                    break;
+                case 3: // both second; take lower if unequal
+                    if (a < b) { // take a
+                        buffer[k++] = a; a = list[i++]; polarity ^= 1;
+                    } else if (b < a) { // take b
+                        buffer[k++] = b; b = other[j++]; polarity ^= 2;
+                    } else { // a == b, take one, drop other
+                        if (a == HIGH) break main;
+                        buffer[k++] = a; a = list[i++]; polarity ^= 1;
+                        b = other[j++]; polarity ^= 2;
+                    }
+                    break;
+                case 1: // a second, b first;
+                    if (a < b) { // NO OVERLAP, drop a
+                        a = list[i++]; polarity ^= 1;
+                    } else if (b < a) { // OVERLAP, take b
+                        buffer[k++] = b; b = other[j++]; polarity ^= 2;
+                    } else { // a == b, drop both!
+                        if (a == HIGH) break main;
+                        a = list[i++]; polarity ^= 1;
+                        b = other[j++]; polarity ^= 2;
+                    }
+                    break;
+                case 2: // a first, b second; if a < b, overlap
+                    if (b < a) { // no overlap, drop b
+                        b = other[j++]; polarity ^= 2;
+                    } else  if (a < b) { // OVERLAP, take a
+                        buffer[k++] = a; a = list[i++]; polarity ^= 1;
+                    } else { // a == b, drop both!
+                        if (a == HIGH) break main;
+                        a = list[i++]; polarity ^= 1;
+                        b = other[j++]; polarity ^= 2;
+                    }
+                    break;
+                }
+            }
+        buffer[k++] = HIGH;    // terminate
+        len = k;
+        // swap list and buffer
+        int[] temp = list;
+        list = buffer;
+        buffer = temp;
+        return this;
+    }
+
+    private static final int max(int a, int b) {
+        return (a > b) ? a : b;
+    }
+
+    //----------------------------------------------------------------
+    // Generic filter-based scanning code
+    //----------------------------------------------------------------
+
+    private static interface Filter {
+        boolean contains(int codePoint);
+    }
+
+    private static final VersionInfo NO_VERSION = VersionInfo.getInstance(0, 0, 0, 0);
+
+    private static class VersionFilter implements Filter {
+        VersionInfo version;
+        VersionFilter(VersionInfo version) { this.version = version; }
+        public boolean contains(int ch) {
+            VersionInfo v = UCharacter.getAge(ch);
+            // Reference comparison ok; VersionInfo caches and reuses
+            // unique objects.
+            return v != NO_VERSION &&
+                    v.compareTo(version) <= 0;
+        }
+    }
+
+    private static synchronized UnicodeSet getInclusions(int src) {
+        if (src != UCharacterProperty.SRC_PROPSVEC) {
+            throw new IllegalStateException("UnicodeSet.getInclusions(unknown src "+src+")");
+        }
+
+        if (INCLUSION == null) {
+            UnicodeSet incl = new UnicodeSet();
+            UCharacterProperty.INSTANCE.upropsvec_addPropertyStarts(incl);
+            INCLUSION = incl;
+        }
+        return INCLUSION;
+    }
+
+    /**
+     * Generic filter-based scanning code for UCD property UnicodeSets.
+     */
+    private UnicodeSet applyFilter(Filter filter, int src) {
+        // Logically, walk through all Unicode characters, noting the start
+        // and end of each range for which filter.contain(c) is
+        // true.  Add each range to a set.
+        //
+        // To improve performance, use an inclusions set which
+        // encodes information about character ranges that are known
+        // to have identical properties.
+        // getInclusions(src) contains exactly the first characters of
+        // same-value ranges for the given properties "source".
+
+        clear();
+
+        int startHasProperty = -1;
+        UnicodeSet inclusions = getInclusions(src);
+        int limitRange = inclusions.getRangeCount();
+
+        for (int j=0; j= 0) {
+                    add_unchecked(startHasProperty, ch-1);
+                    startHasProperty = -1;
+                }
+            }
+        }
+        if (startHasProperty >= 0) {
+            add_unchecked(startHasProperty, 0x10FFFF);
+        }
+
+        return this;
+    }
+
+    /**
+     * Is this frozen, according to the Freezable interface?
+     *
+     * @return value
+     * @stable ICU 3.8
+     */
+    public boolean isFrozen() {
+        return (bmpSet != null || stringSpan != null);
+    }
+
+    /**
+     * Freeze this class, according to the Freezable interface.
+     *
+     * @return this
+     * @stable ICU 4.4
+     */
+    public UnicodeSet freeze() {
+        if (!isFrozen()) {
+            // Do most of what compact() does before freezing because
+            // compact() will not work when the set is frozen.
+            // Small modification: Don't shrink if the savings would be tiny (<=GROW_EXTRA).
+
+            // Delete buffer first to defragment memory less.
+            buffer = null;
+            if (list.length > (len + GROW_EXTRA)) {
+                // Make the capacity equal to len or 1.
+                // We don't want to realloc of 0 size.
+                int capacity = (len == 0) ? 1 : len;
+                int[] oldList = list;
+                list = new int[capacity];
+                for (int i = capacity; i-- > 0;) {
+                    list[i] = oldList[i];
+                }
+            }
+
+            // Optimize contains() and span() and similar functions.
+            if (!strings.isEmpty()) {
+                stringSpan = new UnicodeSetStringSpan(this, new ArrayList(strings), UnicodeSetStringSpan.ALL);
+            }
+            if (stringSpan == null || !stringSpan.needsStringSpanUTF16()) {
+                // Optimize for code point spans.
+                // There are no strings, or
+                // all strings are irrelevant for span() etc. because
+                // all of each string's code points are contained in this set.
+                // However, fully contained strings are relevant for spanAndCount(),
+                // so we create both objects.
+                bmpSet = new BMPSet(list, len);
+            }
+        }
+        return this;
+    }
+
+    /**
+     * Span a string using this UnicodeSet.
+     * 
To replace, count elements, or delete spans, see {@link com.ibm.icu.text.UnicodeSetSpanner UnicodeSetSpanner}.
+     * @param s The string to be spanned
+     * @param spanCondition The span condition
+     * @return the length of the span
+     * @stable ICU 4.4
+     */
+    public int span(CharSequence s, SpanCondition spanCondition) {
+        return span(s, 0, spanCondition);
+    }
+
+    /**
+     * Span a string using this UnicodeSet.
+     *   If the start index is less than 0, span will start from 0.
+     *   If the start index is greater than the string length, span returns the string length.
+     * 
To replace, count elements, or delete spans, see {@link com.ibm.icu.text.UnicodeSetSpanner UnicodeSetSpanner}.
+     * @param s The string to be spanned
+     * @param start The start index that the span begins
+     * @param spanCondition The span condition
+     * @return the string index which ends the span (i.e. exclusive)
+     * @stable ICU 4.4
+     */
+    public int span(CharSequence s, int start, SpanCondition spanCondition) {
+        int end = s.length();
+        if (start < 0) {
+            start = 0;
+        } else if (start >= end) {
+            return end;
+        }
+        if (bmpSet != null) {
+            // Frozen set without strings, or no string is relevant for span().
+            return bmpSet.span(s, start, spanCondition, null);
+        }
+        if (stringSpan != null) {
+            return stringSpan.span(s, start, spanCondition);
+        } else if (!strings.isEmpty()) {
+            int which = spanCondition == SpanCondition.NOT_CONTAINED ? UnicodeSetStringSpan.FWD_UTF16_NOT_CONTAINED
+                    : UnicodeSetStringSpan.FWD_UTF16_CONTAINED;
+            UnicodeSetStringSpan strSpan = new UnicodeSetStringSpan(this, new ArrayList(strings), which);
+            if (strSpan.needsStringSpanUTF16()) {
+                return strSpan.span(s, start, spanCondition);
+            }
+        }
+
+        return spanCodePointsAndCount(s, start, spanCondition, null);
+    }
+
+    /**
+     * Same as span() but also counts the smallest number of set elements on any path across the span.
+     * 
To replace, count elements, or delete spans, see {@link com.ibm.icu.text.UnicodeSetSpanner UnicodeSetSpanner}.
+     * @param outCount An output-only object (must not be null) for returning the count.
+     * @return the limit (exclusive end) of the span
+     */
+    public int spanAndCount(CharSequence s, int start, SpanCondition spanCondition, OutputInt outCount) {
+        if (outCount == null) {
+            throw new IllegalArgumentException("outCount must not be null");
+        }
+        int end = s.length();
+        if (start < 0) {
+            start = 0;
+        } else if (start >= end) {
+            return end;
+        }
+        if (stringSpan != null) {
+            // We might also have bmpSet != null,
+            // but fully-contained strings are relevant for counting elements.
+            return stringSpan.spanAndCount(s, start, spanCondition, outCount);
+        } else if (bmpSet != null) {
+            return bmpSet.span(s, start, spanCondition, outCount);
+        } else if (!strings.isEmpty()) {
+            int which = spanCondition == SpanCondition.NOT_CONTAINED ? UnicodeSetStringSpan.FWD_UTF16_NOT_CONTAINED
+                    : UnicodeSetStringSpan.FWD_UTF16_CONTAINED;
+            which |= UnicodeSetStringSpan.WITH_COUNT;
+            UnicodeSetStringSpan strSpan = new UnicodeSetStringSpan(this, new ArrayList(strings), which);
+            return strSpan.spanAndCount(s, start, spanCondition, outCount);
+        }
+
+        return spanCodePointsAndCount(s, start, spanCondition, outCount);
+    }
+
+    private int spanCodePointsAndCount(CharSequence s, int start,
+            SpanCondition spanCondition, OutputInt outCount) {
+        // Pin to 0/1 values.
+        boolean spanContained = (spanCondition != SpanCondition.NOT_CONTAINED);
+
+        int c;
+        int next = start;
+        int length = s.length();
+        int count = 0;
+        do {
+            c = Character.codePointAt(s, next);
+            if (spanContained != contains(c)) {
+                break;
+            }
+            ++count;
+            next += Character.charCount(c);
+        } while (next < length);
+        if (outCount != null) { outCount.value = count; }
+        return next;
+    }
+
+    /**
+     * Span a string backwards (from the fromIndex) using this UnicodeSet.
+     * If the fromIndex is less than 0, spanBack will return 0.
+     * If fromIndex is greater than the string length, spanBack will start from the string length.
+     * 
To replace, count elements, or delete spans, see {@link com.ibm.icu.text.UnicodeSetSpanner UnicodeSetSpanner}.
+     * @param s The string to be spanned
+     * @param fromIndex The index of the char (exclusive) that the string should be spanned backwards
+     * @param spanCondition The span condition
+     * @return The string index which starts the span (i.e. inclusive).
+     * @stable ICU 4.4
+     */
+    public int spanBack(CharSequence s, int fromIndex, SpanCondition spanCondition) {
+        if (fromIndex <= 0) {
+            return 0;
+        }
+        if (fromIndex > s.length()) {
+            fromIndex = s.length();
+        }
+        if (bmpSet != null) {
+            // Frozen set without strings, or no string is relevant for spanBack().
+            return bmpSet.spanBack(s, fromIndex, spanCondition);
+        }
+        if (stringSpan != null) {
+            return stringSpan.spanBack(s, fromIndex, spanCondition);
+        } else if (!strings.isEmpty()) {
+            int which = (spanCondition == SpanCondition.NOT_CONTAINED)
+                    ? UnicodeSetStringSpan.BACK_UTF16_NOT_CONTAINED
+                            : UnicodeSetStringSpan.BACK_UTF16_CONTAINED;
+            UnicodeSetStringSpan strSpan = new UnicodeSetStringSpan(this, new ArrayList(strings), which);
+            if (strSpan.needsStringSpanUTF16()) {
+                return strSpan.spanBack(s, fromIndex, spanCondition);
+            }
+        }
+
+        // Pin to 0/1 values.
+        boolean spanContained = (spanCondition != SpanCondition.NOT_CONTAINED);
+
+        int c;
+        int prev = fromIndex;
+        do {
+            c = Character.codePointBefore(s, prev);
+            if (spanContained != contains(c)) {
+                break;
+            }
+            prev -= Character.charCount(c);
+        } while (prev > 0);
+        return prev;
+    }
+
+    /**
+     * Clone a thawed version of this class, according to the Freezable interface.
+     * @return the clone, not frozen
+     * @stable ICU 4.4
+     */
+    public UnicodeSet cloneAsThawed() {
+        UnicodeSet result = new UnicodeSet(this);
+        assert !result.isFrozen();
+        return result;
+    }
+
+    // internal function
+    private void checkFrozen() {
+        if (isFrozen()) {
+            throw new UnsupportedOperationException("Attempt to modify frozen object");
+        }
+    }
+
+    /**
+     * Argument values for whether span() and similar functions continue while the current character is contained vs.
+     * not contained in the set.
+     * 

+     * The functionality is straightforward for sets with only single code points, without strings (which is the common
+     * case):
+     * 
    
+     * 
  • CONTAINED and SIMPLE work the same.
+     * 
  • CONTAINED and SIMPLE are inverses of NOT_CONTAINED.
+     * 
  • span() and spanBack() partition any string the
+     * same way when alternating between span(NOT_CONTAINED) and span(either "contained" condition).
+     * 
  • Using a
+     * complemented (inverted) set and the opposite span conditions yields the same results.
+     * 

+     * When a set contains multi-code point strings, then these statements may not be true, depending on the strings in
+     * the set (for example, whether they overlap with each other) and the string that is processed. For a set with
+     * strings:
+     * 
    
+     * 
  • The complement of the set contains the opposite set of code points, but the same set of strings.
+     * Therefore, complementing both the set and the span conditions may yield different results.
+     * 
  • When starting spans
+     * at different positions in a string (span(s, ...) vs. span(s+1, ...)) the ends of the spans may be different
+     * because a set string may start before the later position.
+     * 
  • span(SIMPLE) may be shorter than
+     * span(CONTAINED) because it will not recursively try all possible paths. For example, with a set which
+     * contains the three strings "xy", "xya" and "ax", span("xyax", CONTAINED) will return 4 but span("xyax",
+     * SIMPLE) will return 3. span(SIMPLE) will never be longer than span(CONTAINED).
+     * 
  • With either "contained" condition, span() and spanBack() may partition a string in different ways. For example,
+     * with a set which contains the two strings "ab" and "ba", and when processing the string "aba", span() will yield
+     * contained/not-contained boundaries of { 0, 2, 3 } while spanBack() will yield boundaries of { 0, 1, 3 }.
+     * 

+     * Note: If it is important to get the same boundaries whether iterating forward or backward through a string, then
+     * either only span() should be used and the boundaries cached for backward operation, or an ICU BreakIterator could
+     * be used.
+     * 

+     * Note: Unpaired surrogates are treated like surrogate code points. Similarly, set strings match only on code point
+     * boundaries, never in the middle of a surrogate pair.
+     *
+     * @stable ICU 4.4
+     */
+    public enum SpanCondition {
+        /**
+         * Continues a span() while there is no set element at the current position.
+         * Increments by one code point at a time.
+         * Stops before the first set element (character or string).
+         * (For code points only, this is like while contains(current)==false).
+         * 

+         * When span() returns, the substring between where it started and the position it returned consists only of
+         * characters that are not in the set, and none of its strings overlap with the span.
+         *
+         * @stable ICU 4.4
+         */
+        NOT_CONTAINED,
+
+        /**
+         * Spans the longest substring that is a concatenation of set elements (characters or strings).
+         * (For characters only, this is like while contains(current)==true).
+         * 

+         * When span() returns, the substring between where it started and the position it returned consists only of set
+         * elements (characters or strings) that are in the set.
+         * 

+         * If a set contains strings, then the span will be the longest substring for which there
+         * exists at least one non-overlapping concatenation of set elements (characters or strings).
+         * This is equivalent to a POSIX regular expression for (OR of each set element)*.
+         * (Java/ICU/Perl regex stops at the first match of an OR.)
+         *
+         * @stable ICU 4.4
+         */
+        CONTAINED,
+
+        /**
+         * Continues a span() while there is a set element at the current position.
+         * Increments by the longest matching element at each position.
+         * (For characters only, this is like while contains(current)==true).
+         * 

+         * When span() returns, the substring between where it started and the position it returned consists only of set
+         * elements (characters or strings) that are in the set.
+         * 

+         * If a set only contains single characters, then this is the same as CONTAINED.
+         * 

+         * If a set contains strings, then the span will be the longest substring with a match at each position with the
+         * longest single set element (character or string).
+         * 

+         * Use this span condition together with other longest-match algorithms, such as ICU converters
+         * (ucnv_getUnicodeSet()).
+         *
+         * @stable ICU 4.4
+         */
+        SIMPLE,
+    }
+
+}
--- old/src/java.base/share/classes/sun/text/normalizer/CodePointMap.java	2020-01-10 15:58:10.000000000 -0800
+++ /dev/null	2020-01-10 15:58:10.000000000 -0800
@@ -1,501 +0,0 @@
-/*
- * Copyright (c) 2019, Oracle and/or its affiliates. All rights reserved.
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.  Oracle designates this
- * particular file as subject to the "Classpath" exception as provided
- * by Oracle in the LICENSE file that accompanied this code.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- */
-// (c) 2018 and later: Unicode, Inc. and others.
-// License & terms of use: http://www.unicode.org/copyright.html#License
-
-// created: 2018may10 Markus W. Scherer
-
-package sun.text.normalizer;
-
-import java.util.Iterator;
-import java.util.NoSuchElementException;
-
-/**
- * Abstract map from Unicode code points (U+0000..U+10FFFF) to integer values.
- * This does not implement java.util.Map.
- *
- * @draft ICU 63
- * @provisional This API might change or be removed in a future release.
- */
-public abstract class CodePointMap implements Iterable {
-    /**
-     * Selectors for how getRange() should report value ranges overlapping with surrogates.
-     * Most users should use NORMAL.
-     *
-     * @see #getRange
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    public enum RangeOption {
-        /**
-         * getRange() enumerates all same-value ranges as stored in the map.
-         * Most users should use this option.
-         *
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        NORMAL,
-        /**
-         * getRange() enumerates all same-value ranges as stored in the map,
-         * except that lead surrogates (U+D800..U+DBFF) are treated as having the
-         * surrogateValue, which is passed to getRange() as a separate parameter.
-         * The surrogateValue is not transformed via filter().
-         * See {@link Character#isHighSurrogate}.
-         *
-         * 
Most users should use NORMAL instead.
-         *
-         * 
This option is useful for maps that map surrogate code *units* to
-         * special values optimized for UTF-16 string processing
-         * or for special error behavior for unpaired surrogates,
-         * but those values are not to be associated with the lead surrogate code *points*.
-         *
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        FIXED_LEAD_SURROGATES,
-        /**
-         * getRange() enumerates all same-value ranges as stored in the map,
-         * except that all surrogates (U+D800..U+DFFF) are treated as having the
-         * surrogateValue, which is passed to getRange() as a separate parameter.
-         * The surrogateValue is not transformed via filter().
-         * See {@link Character#isSurrogate}.
-         *
-         * 
Most users should use NORMAL instead.
-         *
-         * 
This option is useful for maps that map surrogate code *units* to
-         * special values optimized for UTF-16 string processing
-         * or for special error behavior for unpaired surrogates,
-         * but those values are not to be associated with the lead surrogate code *points*.
-         *
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        FIXED_ALL_SURROGATES
-    }
-
-    /**
-     * Callback function interface: Modifies a map value.
-     * Optionally called by getRange().
-     * The modified value will be returned by the getRange() function.
-     *
-     * 
Can be used to ignore some of the value bits,
-     * make a filter for one of several values,
-     * return a value index computed from the map value, etc.
-     *
-     * @see #getRange
-     * @see #iterator
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    public interface ValueFilter {
-        /**
-         * Modifies the map value.
-         *
-         * @param value map value
-         * @return modified value
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        public int apply(int value);
-    }
-
-    /**
-     * Range iteration result data.
-     * Code points from start to end map to the same value.
-     * The value may have been modified by {@link ValueFilter#apply(int)},
-     * or it may be the surrogateValue if a RangeOption other than "normal" was used.
-     *
-     * @see #getRange
-     * @see #iterator
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    public static final class Range {
-        private int start;
-        private int end;
-        private int value;
-
-        /**
-         * Constructor. Sets start and end to -1 and value to 0.
-         *
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        public Range() {
-            start = end = -1;
-            value = 0;
-        }
-
-        /**
-         * @return the start code point
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        public int getStart() { return start; }
-        /**
-         * @return the (inclusive) end code point
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        public int getEnd() { return end; }
-        /**
-         * @return the range value
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        public int getValue() { return value; }
-        /**
-         * Sets the range. When using {@link #iterator()},
-         * iteration will resume after the newly set end.
-         *
-         * @param start new start code point
-         * @param end new end code point
-         * @param value new value
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        public void set(int start, int end, int value) {
-            this.start = start;
-            this.end = end;
-            this.value = value;
-        }
-    }
-
-    private final class RangeIterator implements Iterator {
-        private Range range = new Range();
-
-        @Override
-        public boolean hasNext() {
-            return -1 <= range.end && range.end < 0x10ffff;
-        }
-
-        @Override
-        public Range next() {
-            if (getRange(range.end + 1, null, range)) {
-                return range;
-            } else {
-                throw new NoSuchElementException();
-            }
-        }
-
-        @Override
-        public final void remove() {
-            throw new UnsupportedOperationException();
-        }
-    }
-
-    /**
-     * Iterates over code points of a string and fetches map values.
-     * This does not implement java.util.Iterator.
-     *
-     * 
-     * void onString(CodePointMap map, CharSequence s, int start) {
-     *     CodePointMap.StringIterator iter = map.stringIterator(s, start);
-     *     while (iter.next()) {
-     *         int end = iter.getIndex();  // code point from between start and end
-     *         useValue(s, start, end, iter.getCodePoint(), iter.getValue());
-     *         start = end;
-     *     }
-     * }
-     * 

-     *
-     * 
This class is not intended for public subclassing.
-     *
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    public class StringIterator {
-        /**
-         * @internal
-         * @deprecated This API is ICU internal only.
-         */
-        @Deprecated
-        protected CharSequence s;
-        /**
-         * @internal
-         * @deprecated This API is ICU internal only.
-         */
-        @Deprecated
-        protected int sIndex;
-        /**
-         * @internal
-         * @deprecated This API is ICU internal only.
-         */
-        @Deprecated
-        protected int c;
-        /**
-         * @internal
-         * @deprecated This API is ICU internal only.
-         */
-        @Deprecated
-        protected int value;
-
-        /**
-         * @internal
-         * @deprecated This API is ICU internal only.
-         */
-        @Deprecated
-        protected StringIterator(CharSequence s, int sIndex) {
-            this.s = s;
-            this.sIndex = sIndex;
-            c = -1;
-            value = 0;
-        }
-
-        /**
-         * Resets the iterator to a new string and/or a new string index.
-         *
-         * @param s string to iterate over
-         * @param sIndex string index where the iteration will start
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        public void reset(CharSequence s, int sIndex) {
-            this.s = s;
-            this.sIndex = sIndex;
-            c = -1;
-            value = 0;
-        }
-
-        /**
-         * Reads the next code point, post-increments the string index,
-         * and gets a value from the map.
-         * Sets an implementation-defined error value if the code point is an unpaired surrogate.
-         *
-         * @return true if the string index was not yet at the end of the string;
-         *         otherwise the iterator did not advance
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        public boolean next() {
-            if (sIndex >= s.length()) {
-                return false;
-            }
-            c = Character.codePointAt(s, sIndex);
-            sIndex += Character.charCount(c);
-            value = get(c);
-            return true;
-        }
-
-        /**
-         * Reads the previous code point, pre-decrements the string index,
-         * and gets a value from the map.
-         * Sets an implementation-defined error value if the code point is an unpaired surrogate.
-         *
-         * @return true if the string index was not yet at the start of the string;
-         *         otherwise the iterator did not advance
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        public boolean previous() {
-            if (sIndex <= 0) {
-                return false;
-            }
-            c = Character.codePointBefore(s, sIndex);
-            sIndex -= Character.charCount(c);
-            value = get(c);
-            return true;
-        }
-        /**
-         * @return the string index
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        public final int getIndex() { return sIndex; }
-        /**
-         * @return the code point
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        public final int getCodePoint() { return c; }
-        /**
-         * @return the map value,
-         *         or an implementation-defined error value if
-         *         the code point is an unpaired surrogate
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        public final int getValue() { return value; }
-    }
-
-    /**
-     * Protected no-args constructor.
-     *
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    protected CodePointMap() {
-    }
-
-    /**
-     * Returns the value for a code point as stored in the map, with range checking.
-     * Returns an implementation-defined error value if c is not in the range 0..U+10FFFF.
-     *
-     * @param c the code point
-     * @return the map value,
-     *         or an implementation-defined error value if
-     *         the code point is not in the range 0..U+10FFFF
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    public abstract int get(int c);
-
-    /**
-     * Sets the range object to a range of code points beginning with the start parameter.
-     * The range start is the same as the start input parameter
-     * (even if there are preceding code points that have the same value).
-     * The range end is the last code point such that
-     * all those from start to there have the same value.
-     * Returns false if start is not 0..U+10FFFF.
-     * Can be used to efficiently iterate over all same-value ranges in a map.
-     * (This is normally faster than iterating over code points and get()ting each value,
-     * but may be much slower than a data structure that stores ranges directly.)
-     *
-     * 
If the {@link ValueFilter} parameter is not null, then
-     * the value to be delivered is passed through that filter, and the return value is the end
-     * of the range where all values are modified to the same actual value.
-     * The value is unchanged if that parameter is null.
-     *
-     * 
Example:
-     * 
-     * int start = 0;
-     * CodePointMap.Range range = new CodePointMap.Range();
-     * while (map.getRange(start, null, range)) {
-     *     int end = range.getEnd();
-     *     int value = range.getValue();
-     *     // Work with the range start..end and its value.
-     *     start = end + 1;
-     * }
-     * 

-     *
-     * @param start range start
-     * @param filter an object that may modify the map data value,
-     *     or null if the values from the map are to be used unmodified
-     * @param range the range object that will be set to the code point range and value
-     * @return true if start is 0..U+10FFFF; otherwise no new range is fetched
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    public abstract boolean getRange(int start, ValueFilter filter, Range range);
-
-    /**
-     * Sets the range object to a range of code points beginning with the start parameter.
-     * The range start is the same as the start input parameter
-     * (even if there are preceding code points that have the same value).
-     * The range end is the last code point such that
-     * all those from start to there have the same value.
-     * Returns false if start is not 0..U+10FFFF.
-     *
-     * 
Same as the simpler {@link #getRange(int, ValueFilter, Range)} but optionally
-     * modifies the range if it overlaps with surrogate code points.
-     *
-     * @param start range start
-     * @param option defines whether surrogates are treated normally,
-     *               or as having the surrogateValue; usually {@link RangeOption#NORMAL}
-     * @param surrogateValue value for surrogates; ignored if option=={@link RangeOption#NORMAL}
-     * @param filter an object that may modify the map data value,
-     *     or null if the values from the map are to be used unmodified
-     * @param range the range object that will be set to the code point range and value
-     * @return true if start is 0..U+10FFFF; otherwise no new range is fetched
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    public boolean getRange(int start, RangeOption option, int surrogateValue,
-            ValueFilter filter, Range range) {
-        assert option != null;
-        if (!getRange(start, filter, range)) {
-            return false;
-        }
-        if (option == RangeOption.NORMAL) {
-            return true;
-        }
-        int surrEnd = option == RangeOption.FIXED_ALL_SURROGATES ? 0xdfff : 0xdbff;
-        int end = range.end;
-        if (end < 0xd7ff || start > surrEnd) {
-            return true;
-        }
-        // The range overlaps with surrogates, or ends just before the first one.
-        if (range.value == surrogateValue) {
-            if (end >= surrEnd) {
-                // Surrogates followed by a non-surrValue range,
-                // or surrogates are part of a larger surrValue range.
-                return true;
-            }
-        } else {
-            if (start <= 0xd7ff) {
-                range.end = 0xd7ff;  // Non-surrValue range ends before surrValue surrogates.
-                return true;
-            }
-            // Start is a surrogate with a non-surrValue code *unit* value.
-            // Return a surrValue code *point* range.
-            range.value = surrogateValue;
-            if (end > surrEnd) {
-                range.end = surrEnd;  // Surrogate range ends before non-surrValue rest of range.
-                return true;
-            }
-        }
-        // See if the surrValue surrogate range can be merged with
-        // an immediately following range.
-        if (getRange(surrEnd + 1, filter, range) && range.value == surrogateValue) {
-            range.start = start;
-            return true;
-        }
-        range.start = start;
-        range.end = surrEnd;
-        range.value = surrogateValue;
-        return true;
-    }
-
-    /**
-     * Convenience iterator over same-map-value code point ranges.
-     * Same as looping over all ranges with {@link #getRange(int, ValueFilter, Range)}
-     * without filtering.
-     * Adjacent ranges have different map values.
-     *
-     * 
The iterator always returns the same Range object.
-     *
-     * @return a Range iterator
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    @Override
-    public Iterator iterator() {
-        return new RangeIterator();
-    }
-
-    /**
-     * Returns an iterator (not a java.util.Iterator) over code points of a string
-     * for fetching map values.
-     *
-     * @param s string to iterate over
-     * @param sIndex string index where the iteration will start
-     * @return the iterator
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    public StringIterator stringIterator(CharSequence s, int sIndex) {
-        return new StringIterator(s, sIndex);
-    }
-}
--- /dev/null	2020-01-10 15:58:10.000000000 -0800
+++ new/src/java.base/share/classes/jdk/internal/icu/util/CodePointMap.java	2020-01-10 15:58:10.000000000 -0800
@@ -0,0 +1,501 @@
+/*
+ * Copyright (c) 2019, 2020, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.  Oracle designates this
+ * particular file as subject to the "Classpath" exception as provided
+ * by Oracle in the LICENSE file that accompanied this code.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+// (c) 2018 and later: Unicode, Inc. and others.
+// License & terms of use: http://www.unicode.org/copyright.html#License
+
+// created: 2018may10 Markus W. Scherer
+
+package jdk.internal.icu.util;
+
+import java.util.Iterator;
+import java.util.NoSuchElementException;
+
+/**
+ * Abstract map from Unicode code points (U+0000..U+10FFFF) to integer values.
+ * This does not implement java.util.Map.
+ *
+ * @draft ICU 63
+ * @provisional This API might change or be removed in a future release.
+ */
+public abstract class CodePointMap implements Iterable {
+    /**
+     * Selectors for how getRange() should report value ranges overlapping with surrogates.
+     * Most users should use NORMAL.
+     *
+     * @see #getRange
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    public enum RangeOption {
+        /**
+         * getRange() enumerates all same-value ranges as stored in the map.
+         * Most users should use this option.
+         *
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        NORMAL,
+        /**
+         * getRange() enumerates all same-value ranges as stored in the map,
+         * except that lead surrogates (U+D800..U+DBFF) are treated as having the
+         * surrogateValue, which is passed to getRange() as a separate parameter.
+         * The surrogateValue is not transformed via filter().
+         * See {@link Character#isHighSurrogate}.
+         *
+         * 
Most users should use NORMAL instead.
+         *
+         * 
This option is useful for maps that map surrogate code *units* to
+         * special values optimized for UTF-16 string processing
+         * or for special error behavior for unpaired surrogates,
+         * but those values are not to be associated with the lead surrogate code *points*.
+         *
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        FIXED_LEAD_SURROGATES,
+        /**
+         * getRange() enumerates all same-value ranges as stored in the map,
+         * except that all surrogates (U+D800..U+DFFF) are treated as having the
+         * surrogateValue, which is passed to getRange() as a separate parameter.
+         * The surrogateValue is not transformed via filter().
+         * See {@link Character#isSurrogate}.
+         *
+         * 
Most users should use NORMAL instead.
+         *
+         * 
This option is useful for maps that map surrogate code *units* to
+         * special values optimized for UTF-16 string processing
+         * or for special error behavior for unpaired surrogates,
+         * but those values are not to be associated with the lead surrogate code *points*.
+         *
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        FIXED_ALL_SURROGATES
+    }
+
+    /**
+     * Callback function interface: Modifies a map value.
+     * Optionally called by getRange().
+     * The modified value will be returned by the getRange() function.
+     *
+     * 
Can be used to ignore some of the value bits,
+     * make a filter for one of several values,
+     * return a value index computed from the map value, etc.
+     *
+     * @see #getRange
+     * @see #iterator
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    public interface ValueFilter {
+        /**
+         * Modifies the map value.
+         *
+         * @param value map value
+         * @return modified value
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        public int apply(int value);
+    }
+
+    /**
+     * Range iteration result data.
+     * Code points from start to end map to the same value.
+     * The value may have been modified by {@link ValueFilter#apply(int)},
+     * or it may be the surrogateValue if a RangeOption other than "normal" was used.
+     *
+     * @see #getRange
+     * @see #iterator
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    public static final class Range {
+        private int start;
+        private int end;
+        private int value;
+
+        /**
+         * Constructor. Sets start and end to -1 and value to 0.
+         *
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        public Range() {
+            start = end = -1;
+            value = 0;
+        }
+
+        /**
+         * @return the start code point
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        public int getStart() { return start; }
+        /**
+         * @return the (inclusive) end code point
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        public int getEnd() { return end; }
+        /**
+         * @return the range value
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        public int getValue() { return value; }
+        /**
+         * Sets the range. When using {@link #iterator()},
+         * iteration will resume after the newly set end.
+         *
+         * @param start new start code point
+         * @param end new end code point
+         * @param value new value
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        public void set(int start, int end, int value) {
+            this.start = start;
+            this.end = end;
+            this.value = value;
+        }
+    }
+
+    private final class RangeIterator implements Iterator {
+        private Range range = new Range();
+
+        @Override
+        public boolean hasNext() {
+            return -1 <= range.end && range.end < 0x10ffff;
+        }
+
+        @Override
+        public Range next() {
+            if (getRange(range.end + 1, null, range)) {
+                return range;
+            } else {
+                throw new NoSuchElementException();
+            }
+        }
+
+        @Override
+        public final void remove() {
+            throw new UnsupportedOperationException();
+        }
+    }
+
+    /**
+     * Iterates over code points of a string and fetches map values.
+     * This does not implement java.util.Iterator.
+     *
+     * 
+     * void onString(CodePointMap map, CharSequence s, int start) {
+     *     CodePointMap.StringIterator iter = map.stringIterator(s, start);
+     *     while (iter.next()) {
+     *         int end = iter.getIndex();  // code point from between start and end
+     *         useValue(s, start, end, iter.getCodePoint(), iter.getValue());
+     *         start = end;
+     *     }
+     * }
+     * 

+     *
+     * 
This class is not intended for public subclassing.
+     *
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    public class StringIterator {
+        /**
+         * @internal
+         * @deprecated This API is ICU internal only.
+         */
+        @Deprecated
+        protected CharSequence s;
+        /**
+         * @internal
+         * @deprecated This API is ICU internal only.
+         */
+        @Deprecated
+        protected int sIndex;
+        /**
+         * @internal
+         * @deprecated This API is ICU internal only.
+         */
+        @Deprecated
+        protected int c;
+        /**
+         * @internal
+         * @deprecated This API is ICU internal only.
+         */
+        @Deprecated
+        protected int value;
+
+        /**
+         * @internal
+         * @deprecated This API is ICU internal only.
+         */
+        @Deprecated
+        protected StringIterator(CharSequence s, int sIndex) {
+            this.s = s;
+            this.sIndex = sIndex;
+            c = -1;
+            value = 0;
+        }
+
+        /**
+         * Resets the iterator to a new string and/or a new string index.
+         *
+         * @param s string to iterate over
+         * @param sIndex string index where the iteration will start
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        public void reset(CharSequence s, int sIndex) {
+            this.s = s;
+            this.sIndex = sIndex;
+            c = -1;
+            value = 0;
+        }
+
+        /**
+         * Reads the next code point, post-increments the string index,
+         * and gets a value from the map.
+         * Sets an implementation-defined error value if the code point is an unpaired surrogate.
+         *
+         * @return true if the string index was not yet at the end of the string;
+         *         otherwise the iterator did not advance
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        public boolean next() {
+            if (sIndex >= s.length()) {
+                return false;
+            }
+            c = Character.codePointAt(s, sIndex);
+            sIndex += Character.charCount(c);
+            value = get(c);
+            return true;
+        }
+
+        /**
+         * Reads the previous code point, pre-decrements the string index,
+         * and gets a value from the map.
+         * Sets an implementation-defined error value if the code point is an unpaired surrogate.
+         *
+         * @return true if the string index was not yet at the start of the string;
+         *         otherwise the iterator did not advance
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        public boolean previous() {
+            if (sIndex <= 0) {
+                return false;
+            }
+            c = Character.codePointBefore(s, sIndex);
+            sIndex -= Character.charCount(c);
+            value = get(c);
+            return true;
+        }
+        /**
+         * @return the string index
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        public final int getIndex() { return sIndex; }
+        /**
+         * @return the code point
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        public final int getCodePoint() { return c; }
+        /**
+         * @return the map value,
+         *         or an implementation-defined error value if
+         *         the code point is an unpaired surrogate
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        public final int getValue() { return value; }
+    }
+
+    /**
+     * Protected no-args constructor.
+     *
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    protected CodePointMap() {
+    }
+
+    /**
+     * Returns the value for a code point as stored in the map, with range checking.
+     * Returns an implementation-defined error value if c is not in the range 0..U+10FFFF.
+     *
+     * @param c the code point
+     * @return the map value,
+     *         or an implementation-defined error value if
+     *         the code point is not in the range 0..U+10FFFF
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    public abstract int get(int c);
+
+    /**
+     * Sets the range object to a range of code points beginning with the start parameter.
+     * The range start is the same as the start input parameter
+     * (even if there are preceding code points that have the same value).
+     * The range end is the last code point such that
+     * all those from start to there have the same value.
+     * Returns false if start is not 0..U+10FFFF.
+     * Can be used to efficiently iterate over all same-value ranges in a map.
+     * (This is normally faster than iterating over code points and get()ting each value,
+     * but may be much slower than a data structure that stores ranges directly.)
+     *
+     * 
If the {@link ValueFilter} parameter is not null, then
+     * the value to be delivered is passed through that filter, and the return value is the end
+     * of the range where all values are modified to the same actual value.
+     * The value is unchanged if that parameter is null.
+     *
+     * 
Example:
+     * 
+     * int start = 0;
+     * CodePointMap.Range range = new CodePointMap.Range();
+     * while (map.getRange(start, null, range)) {
+     *     int end = range.getEnd();
+     *     int value = range.getValue();
+     *     // Work with the range start..end and its value.
+     *     start = end + 1;
+     * }
+     * 

+     *
+     * @param start range start
+     * @param filter an object that may modify the map data value,
+     *     or null if the values from the map are to be used unmodified
+     * @param range the range object that will be set to the code point range and value
+     * @return true if start is 0..U+10FFFF; otherwise no new range is fetched
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    public abstract boolean getRange(int start, ValueFilter filter, Range range);
+
+    /**
+     * Sets the range object to a range of code points beginning with the start parameter.
+     * The range start is the same as the start input parameter
+     * (even if there are preceding code points that have the same value).
+     * The range end is the last code point such that
+     * all those from start to there have the same value.
+     * Returns false if start is not 0..U+10FFFF.
+     *
+     * 
Same as the simpler {@link #getRange(int, ValueFilter, Range)} but optionally
+     * modifies the range if it overlaps with surrogate code points.
+     *
+     * @param start range start
+     * @param option defines whether surrogates are treated normally,
+     *               or as having the surrogateValue; usually {@link RangeOption#NORMAL}
+     * @param surrogateValue value for surrogates; ignored if option=={@link RangeOption#NORMAL}
+     * @param filter an object that may modify the map data value,
+     *     or null if the values from the map are to be used unmodified
+     * @param range the range object that will be set to the code point range and value
+     * @return true if start is 0..U+10FFFF; otherwise no new range is fetched
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    public boolean getRange(int start, RangeOption option, int surrogateValue,
+            ValueFilter filter, Range range) {
+        assert option != null;
+        if (!getRange(start, filter, range)) {
+            return false;
+        }
+        if (option == RangeOption.NORMAL) {
+            return true;
+        }
+        int surrEnd = option == RangeOption.FIXED_ALL_SURROGATES ? 0xdfff : 0xdbff;
+        int end = range.end;
+        if (end < 0xd7ff || start > surrEnd) {
+            return true;
+        }
+        // The range overlaps with surrogates, or ends just before the first one.
+        if (range.value == surrogateValue) {
+            if (end >= surrEnd) {
+                // Surrogates followed by a non-surrValue range,
+                // or surrogates are part of a larger surrValue range.
+                return true;
+            }
+        } else {
+            if (start <= 0xd7ff) {
+                range.end = 0xd7ff;  // Non-surrValue range ends before surrValue surrogates.
+                return true;
+            }
+            // Start is a surrogate with a non-surrValue code *unit* value.
+            // Return a surrValue code *point* range.
+            range.value = surrogateValue;
+            if (end > surrEnd) {
+                range.end = surrEnd;  // Surrogate range ends before non-surrValue rest of range.
+                return true;
+            }
+        }
+        // See if the surrValue surrogate range can be merged with
+        // an immediately following range.
+        if (getRange(surrEnd + 1, filter, range) && range.value == surrogateValue) {
+            range.start = start;
+            return true;
+        }
+        range.start = start;
+        range.end = surrEnd;
+        range.value = surrogateValue;
+        return true;
+    }
+
+    /**
+     * Convenience iterator over same-map-value code point ranges.
+     * Same as looping over all ranges with {@link #getRange(int, ValueFilter, Range)}
+     * without filtering.
+     * Adjacent ranges have different map values.
+     *
+     * 
The iterator always returns the same Range object.
+     *
+     * @return a Range iterator
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    @Override
+    public Iterator iterator() {
+        return new RangeIterator();
+    }
+
+    /**
+     * Returns an iterator (not a java.util.Iterator) over code points of a string
+     * for fetching map values.
+     *
+     * @param s string to iterate over
+     * @param sIndex string index where the iteration will start
+     * @return the iterator
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    public StringIterator stringIterator(CharSequence s, int sIndex) {
+        return new StringIterator(s, sIndex);
+    }
+}
--- old/src/java.base/share/classes/sun/text/normalizer/CodePointTrie.java	2020-01-10 15:58:11.000000000 -0800
+++ /dev/null	2020-01-10 15:58:11.000000000 -0800
@@ -1,1310 +0,0 @@
-/*
- * Copyright (c) 2019, Oracle and/or its affiliates. All rights reserved.
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.  Oracle designates this
- * particular file as subject to the "Classpath" exception as provided
- * by Oracle in the LICENSE file that accompanied this code.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- */
-// (c) 2018 and later: Unicode, Inc. and others.
-// License & terms of use: http://www.unicode.org/copyright.html#License
-
-// created: 2018may04 Markus W. Scherer
-
-package sun.text.normalizer;
-
-import java.io.DataOutputStream;
-import java.io.IOException;
-import java.io.UncheckedIOException;
-import java.io.OutputStream;
-import java.nio.ByteBuffer;
-import java.nio.ByteOrder;
-
-import static sun.text.normalizer.NormalizerImpl.UTF16Plus;
-
-/**
- * Immutable Unicode code point trie.
- * Fast, reasonably compact, map from Unicode code points (U+0000..U+10FFFF) to integer values.
- * For details see http://site.icu-project.org/design/struct/utrie
- *
- * 
This class is not intended for public subclassing.
- *
- * @see MutableCodePointTrie
- * @draft ICU 63
- * @provisional This API might change or be removed in a future release.
- */
-@SuppressWarnings("deprecation")
-public abstract class CodePointTrie extends CodePointMap {
-    /**
-     * Selectors for the type of a CodePointTrie.
-     * Different trade-offs for size vs. speed.
-     *
-     * 
Use null for {@link #fromBinary} to accept any type;
-     * {@link #getType} will return the actual type.
-     *
-     * @see MutableCodePointTrie#buildImmutable(CodePointTrie.Type, CodePointTrie.ValueWidth)
-     * @see #fromBinary
-     * @see #getType
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    public enum Type {
-        /**
-         * Fast/simple/larger BMP data structure.
-         * The {@link Fast} subclasses have additional functions for lookup for BMP and supplementary code points.
-         *
-         * @see Fast
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        FAST,
-        /**
-         * Small/slower BMP data structure.
-         *
-         * @see Small
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        SMALL
-    }
-
-    /**
-     * Selectors for the number of bits in a CodePointTrie data value.
-     *
-     * 
Use null for {@link #fromBinary} to accept any data value width;
-     * {@link #getValueWidth} will return the actual data value width.
-     *
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    public enum ValueWidth {
-        /**
-         * The trie stores 16 bits per data value.
-         * It returns them as unsigned values 0..0xffff=65535.
-         *
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        BITS_16,
-        /**
-         * The trie stores 32 bits per data value.
-         *
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        BITS_32,
-        /**
-         * The trie stores 8 bits per data value.
-         * It returns them as unsigned values 0..0xff=255.
-         *
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        BITS_8
-    }
-
-    private CodePointTrie(char[] index, Data data, int highStart,
-            int index3NullOffset, int dataNullOffset) {
-        this.ascii = new int[ASCII_LIMIT];
-        this.index = index;
-        this.data = data;
-        this.dataLength = data.getDataLength();
-        this.highStart = highStart;
-        this.index3NullOffset = index3NullOffset;
-        this.dataNullOffset = dataNullOffset;
-
-        for (int c = 0; c < ASCII_LIMIT; ++c) {
-            ascii[c] = data.getFromIndex(c);
-        }
-
-        int nullValueOffset = dataNullOffset;
-        if (nullValueOffset >= dataLength) {
-            nullValueOffset = dataLength - HIGH_VALUE_NEG_DATA_OFFSET;
-        }
-        nullValue = data.getFromIndex(nullValueOffset);
-    }
-
-    /**
-     * Creates a trie from its binary form,
-     * stored in the ByteBuffer starting at the current position.
-     * Advances the buffer position to just after the trie data.
-     * Inverse of {@link #toBinary(OutputStream)}.
-     *
-     * 
The data is copied from the buffer;
-     * later modification of the buffer will not affect the trie.
-     *
-     * @param type selects the trie type; this method throws an exception
-     *             if the type does not match the binary data;
-     *             use null to accept any type
-     * @param valueWidth selects the number of bits in a data value; this method throws an exception
-     *                  if the valueWidth does not match the binary data;
-     *                  use null to accept any data value width
-     * @param bytes a buffer containing the binary data of a CodePointTrie
-     * @return the trie
-     * @see MutableCodePointTrie#MutableCodePointTrie(int, int)
-     * @see MutableCodePointTrie#buildImmutable(CodePointTrie.Type, CodePointTrie.ValueWidth)
-     * @see #toBinary(OutputStream)
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    public static CodePointTrie fromBinary(Type type, ValueWidth valueWidth, ByteBuffer bytes) {
-        ByteOrder outerByteOrder = bytes.order();
-        try {
-            // Enough data for a trie header?
-            if (bytes.remaining() < 16 /* sizeof(UCPTrieHeader) */) {
-                throw new InternalError("Buffer too short for a CodePointTrie header");
-            }
-
-            // struct UCPTrieHeader
-            /** "Tri3" in big-endian US-ASCII (0x54726933) */
-            int signature = bytes.getInt();
-
-            // Check the signature.
-            switch (signature) {
-            case 0x54726933:
-                // The buffer is already set to the trie data byte order.
-                break;
-            case 0x33697254:
-                // Temporarily reverse the byte order.
-                boolean isBigEndian = outerByteOrder == ByteOrder.BIG_ENDIAN;
-                bytes.order(isBigEndian ? ByteOrder.LITTLE_ENDIAN : ByteOrder.BIG_ENDIAN);
-                signature = 0x54726933;
-                break;
-            default:
-                throw new InternalError("Buffer does not contain a serialized CodePointTrie");
-            }
-
-            // struct UCPTrieHeader continued
-            /**
-             * Options bit field:
-             * Bits 15..12: Data length bits 19..16.
-             * Bits 11..8: Data null block offset bits 19..16.
-             * Bits 7..6: UCPTrieType
-             * Bits 5..3: Reserved (0).
-             * Bits 2..0: UCPTrieValueWidth
-             */
-            int options = bytes.getChar();
-
-            /** Total length of the index tables. */
-            int indexLength = bytes.getChar();
-
-            /** Data length bits 15..0. */
-            int dataLength = bytes.getChar();
-
-            /** Index-3 null block offset, 0x7fff or 0xffff if none. */
-            int index3NullOffset = bytes.getChar();
-
-            /** Data null block offset bits 15..0, 0xfffff if none. */
-            int dataNullOffset = bytes.getChar();
-
-            /**
-             * First code point of the single-value range ending with U+10ffff,
-             * rounded up and then shifted right by SHIFT_2.
-             */
-            int shiftedHighStart = bytes.getChar();
-            // struct UCPTrieHeader end
-
-            int typeInt = (options >> 6) & 3;
-            Type actualType;
-            switch (typeInt) {
-            case 0: actualType = Type.FAST; break;
-            case 1: actualType = Type.SMALL; break;
-            default:
-                throw new InternalError("CodePointTrie data header has an unsupported type");
-            }
-
-            int valueWidthInt = options & OPTIONS_VALUE_BITS_MASK;
-            ValueWidth actualValueWidth;
-            switch (valueWidthInt) {
-            case 0: actualValueWidth = ValueWidth.BITS_16; break;
-            case 1: actualValueWidth = ValueWidth.BITS_32; break;
-            case 2: actualValueWidth = ValueWidth.BITS_8; break;
-            default:
-                throw new InternalError("CodePointTrie data header has an unsupported value width");
-            }
-
-            if ((options & OPTIONS_RESERVED_MASK) != 0) {
-                throw new InternalError("CodePointTrie data header has unsupported options");
-            }
-
-            if (type == null) {
-                type = actualType;
-            }
-            if (valueWidth == null) {
-                valueWidth = actualValueWidth;
-            }
-            if (type != actualType || valueWidth != actualValueWidth) {
-                throw new InternalError("CodePointTrie data header has a different type or value width than required");
-            }
-
-            // Get the length values and offsets.
-            dataLength |= ((options & OPTIONS_DATA_LENGTH_MASK) << 4);
-            dataNullOffset |= ((options & OPTIONS_DATA_NULL_OFFSET_MASK) << 8);
-
-            int highStart = shiftedHighStart << SHIFT_2;
-
-            // Calculate the actual length, minus the header.
-            int actualLength = indexLength * 2;
-            if (valueWidth == ValueWidth.BITS_16) {
-                actualLength += dataLength * 2;
-            } else if (valueWidth == ValueWidth.BITS_32) {
-                actualLength += dataLength * 4;
-            } else {
-                actualLength += dataLength;
-            }
-            if (bytes.remaining() < actualLength) {
-                throw new InternalError("Buffer too short for the CodePointTrie data");
-            }
-
-            char[] index = ICUBinary.getChars(bytes, indexLength, 0);
-            switch (valueWidth) {
-            case BITS_16: {
-                char[] data16 = ICUBinary.getChars(bytes, dataLength, 0);
-                return type == Type.FAST ?
-                        new Fast16(index, data16, highStart, index3NullOffset, dataNullOffset) :
-                            new Small16(index, data16, highStart, index3NullOffset, dataNullOffset);
-            }
-            case BITS_32: {
-                int[] data32 = ICUBinary.getInts(bytes, dataLength, 0);
-                return type == Type.FAST ?
-                        new Fast32(index, data32, highStart, index3NullOffset, dataNullOffset) :
-                            new Small32(index, data32, highStart, index3NullOffset, dataNullOffset);
-            }
-            case BITS_8: {
-                byte[] data8 = ICUBinary.getBytes(bytes, dataLength, 0);
-                return type == Type.FAST ?
-                        new Fast8(index, data8, highStart, index3NullOffset, dataNullOffset) :
-                            new Small8(index, data8, highStart, index3NullOffset, dataNullOffset);
-            }
-            default:
-                throw new AssertionError("should be unreachable");
-            }
-        } finally {
-            bytes.order(outerByteOrder);
-        }
-    }
-
-    /**
-     * Returns the trie type.
-     *
-     * @return the trie type
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    public abstract Type getType();
-    /**
-     * Returns the number of bits in a trie data value.
-     *
-     * @return the number of bits in a trie data value
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    public final ValueWidth getValueWidth() { return data.getValueWidth(); }
-
-    /**
-     * {@inheritDoc}
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    @Override
-    public int get(int c) {
-        return data.getFromIndex(cpIndex(c));
-    }
-
-    /**
-     * Returns a trie value for an ASCII code point, without range checking.
-     *
-     * @param c the input code point; must be U+0000..U+007F
-     * @return The ASCII code point's trie value.
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    public final int asciiGet(int c) {
-        return ascii[c];
-    }
-
-    private static final int MAX_UNICODE = 0x10ffff;
-
-    private static final int ASCII_LIMIT = 0x80;
-
-    private static final int maybeFilterValue(int value, int trieNullValue, int nullValue,
-            ValueFilter filter) {
-        if (value == trieNullValue) {
-            value = nullValue;
-        } else if (filter != null) {
-            value = filter.apply(value);
-        }
-        return value;
-    }
-
-    /**
-     * {@inheritDoc}
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    @Override
-    public final boolean getRange(int start, ValueFilter filter, Range range) {
-        if (start < 0 || MAX_UNICODE < start) {
-            return false;
-        }
-        if (start >= highStart) {
-            int di = dataLength - HIGH_VALUE_NEG_DATA_OFFSET;
-            int value = data.getFromIndex(di);
-            if (filter != null) { value = filter.apply(value); }
-            range.set(start, MAX_UNICODE, value);
-            return true;
-        }
-
-        int nullValue = this.nullValue;
-        if (filter != null) { nullValue = filter.apply(nullValue); }
-        Type type = getType();
-
-        int prevI3Block = -1;
-        int prevBlock = -1;
-        int c = start;
-        // Initialize to make compiler happy. Real value when haveValue is true.
-        int trieValue = 0, value = 0;
-        boolean haveValue = false;
-        do {
-            int i3Block;
-            int i3;
-            int i3BlockLength;
-            int dataBlockLength;
-            if (c <= 0xffff && (type == Type.FAST || c <= SMALL_MAX)) {
-                i3Block = 0;
-                i3 = c >> FAST_SHIFT;
-                i3BlockLength = type == Type.FAST ? BMP_INDEX_LENGTH : SMALL_INDEX_LENGTH;
-                dataBlockLength = FAST_DATA_BLOCK_LENGTH;
-            } else {
-                // Use the multi-stage index.
-                int i1 = c >> SHIFT_1;
-                if (type == Type.FAST) {
-                    assert(0xffff < c && c < highStart);
-                    i1 += BMP_INDEX_LENGTH - OMITTED_BMP_INDEX_1_LENGTH;
-                } else {
-                    assert(c < highStart && highStart > SMALL_LIMIT);
-                    i1 += SMALL_INDEX_LENGTH;
-                }
-                i3Block = index[index[i1] + ((c >> SHIFT_2) & INDEX_2_MASK)];
-                if (i3Block == prevI3Block && (c - start) >= CP_PER_INDEX_2_ENTRY) {
-                    // The index-3 block is the same as the previous one, and filled with value.
-                    assert((c & (CP_PER_INDEX_2_ENTRY - 1)) == 0);
-                    c += CP_PER_INDEX_2_ENTRY;
-                    continue;
-                }
-                prevI3Block = i3Block;
-                if (i3Block == index3NullOffset) {
-                    // This is the index-3 null block.
-                    if (haveValue) {
-                        if (nullValue != value) {
-                            range.set(start, c - 1, value);
-                            return true;
-                        }
-                    } else {
-                        trieValue = this.nullValue;
-                        value = nullValue;
-                        haveValue = true;
-                    }
-                    prevBlock = dataNullOffset;
-                    c = (c + CP_PER_INDEX_2_ENTRY) & ~(CP_PER_INDEX_2_ENTRY - 1);
-                    continue;
-                }
-                i3 = (c >> SHIFT_3) & INDEX_3_MASK;
-                i3BlockLength = INDEX_3_BLOCK_LENGTH;
-                dataBlockLength = SMALL_DATA_BLOCK_LENGTH;
-            }
-            // Enumerate data blocks for one index-3 block.
-            do {
-                int block;
-                if ((i3Block & 0x8000) == 0) {
-                    block = index[i3Block + i3];
-                } else {
-                    // 18-bit indexes stored in groups of 9 entries per 8 indexes.
-                    int group = (i3Block & 0x7fff) + (i3 & ~7) + (i3 >> 3);
-                    int gi = i3 & 7;
-                    block = (index[group++] << (2 + (2 * gi))) & 0x30000;
-                    block |= index[group + gi];
-                }
-                if (block == prevBlock && (c - start) >= dataBlockLength) {
-                    // The block is the same as the previous one, and filled with value.
-                    assert((c & (dataBlockLength - 1)) == 0);
-                    c += dataBlockLength;
-                } else {
-                    int dataMask = dataBlockLength - 1;
-                    prevBlock = block;
-                    if (block == dataNullOffset) {
-                        // This is the data null block.
-                        if (haveValue) {
-                            if (nullValue != value) {
-                                range.set(start, c - 1, value);
-                                return true;
-                            }
-                        } else {
-                            trieValue = this.nullValue;
-                            value = nullValue;
-                            haveValue = true;
-                        }
-                        c = (c + dataBlockLength) & ~dataMask;
-                    } else {
-                        int di = block + (c & dataMask);
-                        int trieValue2 = data.getFromIndex(di);
-                        if (haveValue) {
-                            if (trieValue2 != trieValue) {
-                                if (filter == null ||
-                                        maybeFilterValue(trieValue2, this.nullValue, nullValue,
-                                                filter) != value) {
-                                    range.set(start, c - 1, value);
-                                    return true;
-                                }
-                                trieValue = trieValue2;  // may or may not help
-                            }
-                        } else {
-                            trieValue = trieValue2;
-                            value = maybeFilterValue(trieValue2, this.nullValue, nullValue, filter);
-                            haveValue = true;
-                        }
-                        while ((++c & dataMask) != 0) {
-                            trieValue2 = data.getFromIndex(++di);
-                            if (trieValue2 != trieValue) {
-                                if (filter == null ||
-                                        maybeFilterValue(trieValue2, this.nullValue, nullValue,
-                                                filter) != value) {
-                                    range.set(start, c - 1, value);
-                                    return true;
-                                }
-                                trieValue = trieValue2;  // may or may not help
-                            }
-                        }
-                    }
-                }
-            } while (++i3 < i3BlockLength);
-        } while (c < highStart);
-        assert(haveValue);
-        int di = dataLength - HIGH_VALUE_NEG_DATA_OFFSET;
-        int highValue = data.getFromIndex(di);
-        if (maybeFilterValue(highValue, this.nullValue, nullValue, filter) != value) {
-            --c;
-        } else {
-            c = MAX_UNICODE;
-        }
-        range.set(start, c, value);
-        return true;
-    }
-
-    /**
-     * Writes a representation of the trie to the output stream.
-     * Inverse of {@link #fromBinary}.
-     *
-     * @param os the output stream
-     * @return the number of bytes written
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    public final int toBinary(OutputStream os) {
-        try {
-            DataOutputStream dos = new DataOutputStream(os);
-
-            // Write the UCPTrieHeader
-            dos.writeInt(0x54726933);  // signature="Tri3"
-            dos.writeChar(  // options
-                ((dataLength & 0xf0000) >> 4) |
-                ((dataNullOffset & 0xf0000) >> 8) |
-                (getType().ordinal() << 6) |
-                getValueWidth().ordinal());
-            dos.writeChar(index.length);
-            dos.writeChar(dataLength);
-            dos.writeChar(index3NullOffset);
-            dos.writeChar(dataNullOffset);
-            dos.writeChar(highStart >> SHIFT_2);  // shiftedHighStart
-            int length = 16;  // sizeof(UCPTrieHeader)
-
-            for (char i : index) { dos.writeChar(i); }
-            length += index.length * 2;
-            length += data.write(dos);
-            return length;
-        } catch (IOException e) {
-            throw new UncheckedIOException(e);
-        }
-    }
-
-    /** @internal */
-    static final int FAST_SHIFT = 6;
-
-    /** Number of entries in a data block for code points below the fast limit. 64=0x40 @internal */
-    static final int FAST_DATA_BLOCK_LENGTH = 1 << FAST_SHIFT;
-
-    /** Mask for getting the lower bits for the in-fast-data-block offset. @internal */
-    private static final int FAST_DATA_MASK = FAST_DATA_BLOCK_LENGTH - 1;
-
-    /** @internal */
-    private static final int SMALL_MAX = 0xfff;
-
-    /**
-     * Offset from dataLength (to be subtracted) for fetching the
-     * value returned for out-of-range code points and ill-formed UTF-8/16.
-     * @internal
-     */
-    private static final int ERROR_VALUE_NEG_DATA_OFFSET = 1;
-    /**
-     * Offset from dataLength (to be subtracted) for fetching the
-     * value returned for code points highStart..U+10FFFF.
-     * @internal
-     */
-    private static final int HIGH_VALUE_NEG_DATA_OFFSET = 2;
-
-    // ucptrie_impl.h
-
-    /** The length of the BMP index table. 1024=0x400 */
-    private static final int BMP_INDEX_LENGTH = 0x10000 >> FAST_SHIFT;
-
-    static final int SMALL_LIMIT = 0x1000;
-    private static final int SMALL_INDEX_LENGTH = SMALL_LIMIT >> FAST_SHIFT;
-
-    /** Shift size for getting the index-3 table offset. */
-    static final int SHIFT_3 = 4;
-
-    /** Shift size for getting the index-2 table offset. */
-    private static final int SHIFT_2 = 5 + SHIFT_3;
-
-    /** Shift size for getting the index-1 table offset. */
-    private static final int SHIFT_1 = 5 + SHIFT_2;
-
-    /**
-     * Difference between two shift sizes,
-     * for getting an index-2 offset from an index-3 offset. 5=9-4
-     */
-    static final int SHIFT_2_3 = SHIFT_2 - SHIFT_3;
-
-    /**
-     * Difference between two shift sizes,
-     * for getting an index-1 offset from an index-2 offset. 5=14-9
-     */
-    static final int SHIFT_1_2 = SHIFT_1 - SHIFT_2;
-
-    /**
-     * Number of index-1 entries for the BMP. (4)
-     * This part of the index-1 table is omitted from the serialized form.
-     */
-    private static final int OMITTED_BMP_INDEX_1_LENGTH = 0x10000 >> SHIFT_1;
-
-    /** Number of entries in an index-2 block. 32=0x20 */
-    static final int INDEX_2_BLOCK_LENGTH = 1 << SHIFT_1_2;
-
-    /** Mask for getting the lower bits for the in-index-2-block offset. */
-    static final int INDEX_2_MASK = INDEX_2_BLOCK_LENGTH - 1;
-
-    /** Number of code points per index-2 table entry. 512=0x200 */
-    static final int CP_PER_INDEX_2_ENTRY = 1 << SHIFT_2;
-
-    /** Number of entries in an index-3 block. 32=0x20 */
-    static final int INDEX_3_BLOCK_LENGTH = 1 << SHIFT_2_3;
-
-    /** Mask for getting the lower bits for the in-index-3-block offset. */
-    private static final int INDEX_3_MASK = INDEX_3_BLOCK_LENGTH - 1;
-
-    /** Number of entries in a small data block. 16=0x10 */
-    static final int SMALL_DATA_BLOCK_LENGTH = 1 << SHIFT_3;
-
-    /** Mask for getting the lower bits for the in-small-data-block offset. */
-    static final int SMALL_DATA_MASK = SMALL_DATA_BLOCK_LENGTH - 1;
-
-    // ucptrie_impl.h: Constants for use with UCPTrieHeader.options.
-    private static final int OPTIONS_DATA_LENGTH_MASK = 0xf000;
-    private static final int OPTIONS_DATA_NULL_OFFSET_MASK = 0xf00;
-    private static final int OPTIONS_RESERVED_MASK = 0x38;
-    private static final int OPTIONS_VALUE_BITS_MASK = 7;
-    /**
-     * Value for index3NullOffset which indicates that there is no index-3 null block.
-     * Bit 15 is unused for this value because this bit is used if the index-3 contains
-     * 18-bit indexes.
-     */
-    static final int NO_INDEX3_NULL_OFFSET = 0x7fff;
-    static final int NO_DATA_NULL_OFFSET = 0xfffff;
-
-    private static abstract class Data {
-        abstract ValueWidth getValueWidth();
-        abstract int getDataLength();
-        abstract int getFromIndex(int index);
-        abstract int write(DataOutputStream dos) throws IOException;
-    }
-
-    private static final class Data16 extends Data {
-        char[] array;
-        Data16(char[] a) { array = a; }
-        @Override ValueWidth getValueWidth() { return ValueWidth.BITS_16; }
-        @Override int getDataLength() { return array.length; }
-        @Override int getFromIndex(int index) { return array[index]; }
-        @Override int write(DataOutputStream dos) throws IOException {
-            for (char v : array) { dos.writeChar(v); }
-            return array.length * 2;
-        }
-    }
-
-    private static final class Data32 extends Data {
-        int[] array;
-        Data32(int[] a) { array = a; }
-        @Override ValueWidth getValueWidth() { return ValueWidth.BITS_32; }
-        @Override int getDataLength() { return array.length; }
-        @Override int getFromIndex(int index) { return array[index]; }
-        @Override int write(DataOutputStream dos) throws IOException {
-            for (int v : array) { dos.writeInt(v); }
-            return array.length * 4;
-        }
-    }
-
-    private static final class Data8 extends Data {
-        byte[] array;
-        Data8(byte[] a) { array = a; }
-        @Override ValueWidth getValueWidth() { return ValueWidth.BITS_8; }
-        @Override int getDataLength() { return array.length; }
-        @Override int getFromIndex(int index) { return array[index] & 0xff; }
-        @Override int write(DataOutputStream dos) throws IOException {
-            for (byte v : array) { dos.writeByte(v); }
-            return array.length;
-        }
-    }
-
-    /** @internal */
-    private final int[] ascii;
-
-    /** @internal */
-    private final char[] index;
-
-    /**
-     * @internal
-     * @deprecated This API is ICU internal only.
-     */
-    @Deprecated
-    protected final Data data;
-    /**
-     * @internal
-     * @deprecated This API is ICU internal only.
-     */
-    @Deprecated
-    protected final int dataLength;
-    /**
-     * Start of the last range which ends at U+10FFFF.
-     * @internal
-     * @deprecated This API is ICU internal only.
-     */
-    @Deprecated
-    protected final int highStart;
-
-    /**
-     * Internal index-3 null block offset.
-     * Set to an impossibly high value (e.g., 0xffff) if there is no dedicated index-3 null block.
-     * @internal
-     */
-    private final int index3NullOffset;
-    /**
-     * Internal data null block offset, not shifted.
-     * Set to an impossibly high value (e.g., 0xfffff) if there is no dedicated data null block.
-     * @internal
-     */
-    private final int dataNullOffset;
-    /** @internal */
-    private final int nullValue;
-
-    /**
-     * @internal
-     * @deprecated This API is ICU internal only.
-     */
-    @Deprecated
-    protected final int fastIndex(int c) {
-        return index[c >> FAST_SHIFT] + (c & FAST_DATA_MASK);
-    }
-
-    /**
-     * @internal
-     * @deprecated This API is ICU internal only.
-     */
-    @Deprecated
-    protected final int smallIndex(Type type, int c) {
-        // Split into two methods to make this part inline-friendly.
-        // In C, this part is a macro.
-        if (c >= highStart) {
-            return dataLength - HIGH_VALUE_NEG_DATA_OFFSET;
-        }
-        return internalSmallIndex(type, c);
-    }
-
-    private final int internalSmallIndex(Type type, int c) {
-        int i1 = c >> SHIFT_1;
-        if (type == Type.FAST) {
-            assert(0xffff < c && c < highStart);
-            i1 += BMP_INDEX_LENGTH - OMITTED_BMP_INDEX_1_LENGTH;
-        } else {
-            assert(0 <= c && c < highStart && highStart > SMALL_LIMIT);
-            i1 += SMALL_INDEX_LENGTH;
-        }
-        int i3Block = index[index[i1] + ((c >> SHIFT_2) & INDEX_2_MASK)];
-        int i3 = (c >> SHIFT_3) & INDEX_3_MASK;
-        int dataBlock;
-        if ((i3Block & 0x8000) == 0) {
-            // 16-bit indexes
-            dataBlock = index[i3Block + i3];
-        } else {
-            // 18-bit indexes stored in groups of 9 entries per 8 indexes.
-            i3Block = (i3Block & 0x7fff) + (i3 & ~7) + (i3 >> 3);
-            i3 &= 7;
-            dataBlock = (index[i3Block++] << (2 + (2 * i3))) & 0x30000;
-            dataBlock |= index[i3Block + i3];
-        }
-        return dataBlock + (c & SMALL_DATA_MASK);
-    }
-
-    /**
-     * @internal
-     * @deprecated This API is ICU internal only.
-     */
-    @Deprecated
-    protected abstract int cpIndex(int c);
-
-    /**
-     * A CodePointTrie with {@link Type#FAST}.
-     *
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    public static abstract class Fast extends CodePointTrie {
-        private Fast(char[] index, Data data, int highStart,
-                int index3NullOffset, int dataNullOffset) {
-            super(index, data, highStart, index3NullOffset, dataNullOffset);
-        }
-
-        /**
-         * Creates a trie from its binary form.
-         * Same as {@link CodePointTrie#fromBinary(Type, ValueWidth, ByteBuffer)}
-         * with {@link Type#FAST}.
-         *
-         * @param valueWidth selects the number of bits in a data value; this method throws an exception
-         *                  if the valueWidth does not match the binary data;
-         *                  use null to accept any data value width
-         * @param bytes a buffer containing the binary data of a CodePointTrie
-         * @return the trie
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        public static Fast fromBinary(ValueWidth valueWidth, ByteBuffer bytes) {
-            return (Fast) CodePointTrie.fromBinary(Type.FAST, valueWidth, bytes);
-        }
-
-        /**
-         * @return {@link Type#FAST}
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        @Override
-        public final Type getType() { return Type.FAST; }
-
-        /**
-         * Returns a trie value for a BMP code point (U+0000..U+FFFF), without range checking.
-         * Can be used to look up a value for a UTF-16 code unit if other parts of
-         * the string processing check for surrogates.
-         *
-         * @param c the input code point, must be U+0000..U+FFFF
-         * @return The BMP code point's trie value.
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        public abstract int bmpGet(int c);
-
-        /**
-         * Returns a trie value for a supplementary code point (U+10000..U+10FFFF),
-         * without range checking.
-         *
-         * @param c the input code point, must be U+10000..U+10FFFF
-         * @return The supplementary code point's trie value.
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        public abstract int suppGet(int c);
-
-        /**
-         * @internal
-         * @deprecated This API is ICU internal only.
-         */
-        @Deprecated
-        @Override
-        protected final int cpIndex(int c) {
-            if (c >= 0) {
-                if (c <= 0xffff) {
-                    return fastIndex(c);
-                } else if (c <= 0x10ffff) {
-                    return smallIndex(Type.FAST, c);
-                }
-            }
-            return dataLength - ERROR_VALUE_NEG_DATA_OFFSET;
-        }
-
-        /**
-         * {@inheritDoc}
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        @Override
-        public final StringIterator stringIterator(CharSequence s, int sIndex) {
-            return new FastStringIterator(s, sIndex);
-        }
-
-        private final class FastStringIterator extends StringIterator {
-            private FastStringIterator(CharSequence s, int sIndex) {
-                super(s, sIndex);
-            }
-
-            @Override
-            public boolean next() {
-                if (sIndex >= s.length()) {
-                    return false;
-                }
-                char lead = s.charAt(sIndex++);
-                c = lead;
-                int dataIndex;
-                if (!Character.isSurrogate(lead)) {
-                    dataIndex = fastIndex(c);
-                } else {
-                    char trail;
-                    if (UTF16Plus.isSurrogateLead(lead) && sIndex < s.length() &&
-                            Character.isLowSurrogate(trail = s.charAt(sIndex))) {
-                        ++sIndex;
-                        c = Character.toCodePoint(lead, trail);
-                        dataIndex = smallIndex(Type.FAST, c);
-                    } else {
-                        dataIndex = dataLength - ERROR_VALUE_NEG_DATA_OFFSET;
-                    }
-                }
-                value = data.getFromIndex(dataIndex);
-                return true;
-            }
-
-            @Override
-            public boolean previous() {
-                if (sIndex <= 0) {
-                    return false;
-                }
-                char trail = s.charAt(--sIndex);
-                c = trail;
-                int dataIndex;
-                if (!Character.isSurrogate(trail)) {
-                    dataIndex = fastIndex(c);
-                } else {
-                    char lead;
-                    if (!UTF16Plus.isSurrogateLead(trail) && sIndex > 0 &&
-                            Character.isHighSurrogate(lead = s.charAt(sIndex - 1))) {
-                        --sIndex;
-                        c = Character.toCodePoint(lead, trail);
-                        dataIndex = smallIndex(Type.FAST, c);
-                    } else {
-                        dataIndex = dataLength - ERROR_VALUE_NEG_DATA_OFFSET;
-                    }
-                }
-                value = data.getFromIndex(dataIndex);
-                return true;
-            }
-        }
-    }
-
-    /**
-     * A CodePointTrie with {@link Type#SMALL}.
-     *
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    public static abstract class Small extends CodePointTrie {
-        private Small(char[] index, Data data, int highStart,
-                int index3NullOffset, int dataNullOffset) {
-            super(index, data, highStart, index3NullOffset, dataNullOffset);
-        }
-
-        /**
-         * Creates a trie from its binary form.
-         * Same as {@link CodePointTrie#fromBinary(Type, ValueWidth, ByteBuffer)}
-         * with {@link Type#SMALL}.
-         *
-         * @param valueWidth selects the number of bits in a data value; this method throws an exception
-         *                  if the valueWidth does not match the binary data;
-         *                  use null to accept any data value width
-         * @param bytes a buffer containing the binary data of a CodePointTrie
-         * @return the trie
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        public static Small fromBinary(ValueWidth valueWidth, ByteBuffer bytes) {
-            return (Small) CodePointTrie.fromBinary(Type.SMALL, valueWidth, bytes);
-        }
-
-        /**
-         * @return {@link Type#SMALL}
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        @Override
-        public final Type getType() { return Type.SMALL; }
-
-        /**
-         * @internal
-         * @deprecated This API is ICU internal only.
-         */
-        @Deprecated
-        @Override
-        protected final int cpIndex(int c) {
-            if (c >= 0) {
-                if (c <= SMALL_MAX) {
-                    return fastIndex(c);
-                } else if (c <= 0x10ffff) {
-                    return smallIndex(Type.SMALL, c);
-                }
-            }
-            return dataLength - ERROR_VALUE_NEG_DATA_OFFSET;
-        }
-
-        /**
-         * {@inheritDoc}
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        @Override
-        public final StringIterator stringIterator(CharSequence s, int sIndex) {
-            return new SmallStringIterator(s, sIndex);
-        }
-
-        private final class SmallStringIterator extends StringIterator {
-            private SmallStringIterator(CharSequence s, int sIndex) {
-                super(s, sIndex);
-            }
-
-            @Override
-            public boolean next() {
-                if (sIndex >= s.length()) {
-                    return false;
-                }
-                char lead = s.charAt(sIndex++);
-                c = lead;
-                int dataIndex;
-                if (!Character.isSurrogate(lead)) {
-                    dataIndex = cpIndex(c);
-                } else {
-                    char trail;
-                    if (UTF16Plus.isSurrogateLead(lead) && sIndex < s.length() &&
-                            Character.isLowSurrogate(trail = s.charAt(sIndex))) {
-                        ++sIndex;
-                        c = Character.toCodePoint(lead, trail);
-                        dataIndex = smallIndex(Type.SMALL, c);
-                    } else {
-                        dataIndex = dataLength - ERROR_VALUE_NEG_DATA_OFFSET;
-                    }
-                }
-                value = data.getFromIndex(dataIndex);
-                return true;
-            }
-
-            @Override
-            public boolean previous() {
-                if (sIndex <= 0) {
-                    return false;
-                }
-                char trail = s.charAt(--sIndex);
-                c = trail;
-                int dataIndex;
-                if (!Character.isSurrogate(trail)) {
-                    dataIndex = cpIndex(c);
-                } else {
-                    char lead;
-                    if (!UTF16Plus.isSurrogateLead(trail) && sIndex > 0 &&
-                            Character.isHighSurrogate(lead = s.charAt(sIndex - 1))) {
-                        --sIndex;
-                        c = Character.toCodePoint(lead, trail);
-                        dataIndex = smallIndex(Type.SMALL, c);
-                    } else {
-                        dataIndex = dataLength - ERROR_VALUE_NEG_DATA_OFFSET;
-                    }
-                }
-                value = data.getFromIndex(dataIndex);
-                return true;
-            }
-        }
-    }
-
-    /**
-     * A CodePointTrie with {@link Type#FAST} and {@link ValueWidth#BITS_16}.
-     *
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    public static final class Fast16 extends Fast {
-        private final char[] dataArray;
-
-        Fast16(char[] index, char[] data16, int highStart,
-                int index3NullOffset, int dataNullOffset) {
-            super(index, new Data16(data16), highStart, index3NullOffset, dataNullOffset);
-            this.dataArray = data16;
-        }
-
-        /**
-         * Creates a trie from its binary form.
-         * Same as {@link CodePointTrie#fromBinary(Type, ValueWidth, ByteBuffer)}
-         * with {@link Type#FAST} and {@link ValueWidth#BITS_16}.
-         *
-         * @param bytes a buffer containing the binary data of a CodePointTrie
-         * @return the trie
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        public static Fast16 fromBinary(ByteBuffer bytes) {
-            return (Fast16) CodePointTrie.fromBinary(Type.FAST, ValueWidth.BITS_16, bytes);
-        }
-
-        /**
-         * {@inheritDoc}
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        @Override
-        public final int get(int c) {
-            return dataArray[cpIndex(c)];
-        }
-
-        /**
-         * {@inheritDoc}
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        @Override
-        public final int bmpGet(int c) {
-            assert 0 <= c && c <= 0xffff;
-            return dataArray[fastIndex(c)];
-        }
-
-        /**
-         * {@inheritDoc}
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        @Override
-        public final int suppGet(int c) {
-            assert 0x10000 <= c && c <= 0x10ffff;
-            return dataArray[smallIndex(Type.FAST, c)];
-        }
-    }
-
-    /**
-     * A CodePointTrie with {@link Type#FAST} and {@link ValueWidth#BITS_32}.
-     *
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    public static final class Fast32 extends Fast {
-        private final int[] dataArray;
-
-        Fast32(char[] index, int[] data32, int highStart,
-                int index3NullOffset, int dataNullOffset) {
-            super(index, new Data32(data32), highStart, index3NullOffset, dataNullOffset);
-            this.dataArray = data32;
-        }
-
-        /**
-         * Creates a trie from its binary form.
-         * Same as {@link CodePointTrie#fromBinary(Type, ValueWidth, ByteBuffer)}
-         * with {@link Type#FAST} and {@link ValueWidth#BITS_32}.
-         *
-         * @param bytes a buffer containing the binary data of a CodePointTrie
-         * @return the trie
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        public static Fast32 fromBinary(ByteBuffer bytes) {
-            return (Fast32) CodePointTrie.fromBinary(Type.FAST, ValueWidth.BITS_32, bytes);
-        }
-
-        /**
-         * {@inheritDoc}
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        @Override
-        public final int get(int c) {
-            return dataArray[cpIndex(c)];
-        }
-
-        /**
-         * {@inheritDoc}
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        @Override
-        public final int bmpGet(int c) {
-            assert 0 <= c && c <= 0xffff;
-            return dataArray[fastIndex(c)];
-        }
-
-        /**
-         * {@inheritDoc}
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        @Override
-        public final int suppGet(int c) {
-            assert 0x10000 <= c && c <= 0x10ffff;
-            return dataArray[smallIndex(Type.FAST, c)];
-        }
-    }
-
-    /**
-     * A CodePointTrie with {@link Type#FAST} and {@link ValueWidth#BITS_8}.
-     *
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    public static final class Fast8 extends Fast {
-        private final byte[] dataArray;
-
-        Fast8(char[] index, byte[] data8, int highStart,
-                int index3NullOffset, int dataNullOffset) {
-            super(index, new Data8(data8), highStart, index3NullOffset, dataNullOffset);
-            this.dataArray = data8;
-        }
-
-        /**
-         * Creates a trie from its binary form.
-         * Same as {@link CodePointTrie#fromBinary(Type, ValueWidth, ByteBuffer)}
-         * with {@link Type#FAST} and {@link ValueWidth#BITS_8}.
-         *
-         * @param bytes a buffer containing the binary data of a CodePointTrie
-         * @return the trie
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        public static Fast8 fromBinary(ByteBuffer bytes) {
-            return (Fast8) CodePointTrie.fromBinary(Type.FAST, ValueWidth.BITS_8, bytes);
-        }
-
-        /**
-         * {@inheritDoc}
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        @Override
-        public final int get(int c) {
-            return dataArray[cpIndex(c)] & 0xff;
-        }
-
-        /**
-         * {@inheritDoc}
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        @Override
-        public final int bmpGet(int c) {
-            assert 0 <= c && c <= 0xffff;
-            return dataArray[fastIndex(c)] & 0xff;
-        }
-
-        /**
-         * {@inheritDoc}
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        @Override
-        public final int suppGet(int c) {
-            assert 0x10000 <= c && c <= 0x10ffff;
-            return dataArray[smallIndex(Type.FAST, c)] & 0xff;
-        }
-    }
-
-    /**
-     * A CodePointTrie with {@link Type#SMALL} and {@link ValueWidth#BITS_16}.
-     *
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    public static final class Small16 extends Small {
-        Small16(char[] index, char[] data16, int highStart,
-                int index3NullOffset, int dataNullOffset) {
-            super(index, new Data16(data16), highStart, index3NullOffset, dataNullOffset);
-        }
-
-        /**
-         * Creates a trie from its binary form.
-         * Same as {@link CodePointTrie#fromBinary(Type, ValueWidth, ByteBuffer)}
-         * with {@link Type#SMALL} and {@link ValueWidth#BITS_16}.
-         *
-         * @param bytes a buffer containing the binary data of a CodePointTrie
-         * @return the trie
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        public static Small16 fromBinary(ByteBuffer bytes) {
-            return (Small16) CodePointTrie.fromBinary(Type.SMALL, ValueWidth.BITS_16, bytes);
-        }
-    }
-
-    /**
-     * A CodePointTrie with {@link Type#SMALL} and {@link ValueWidth#BITS_32}.
-     *
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    public static final class Small32 extends Small {
-        Small32(char[] index, int[] data32, int highStart,
-                int index3NullOffset, int dataNullOffset) {
-            super(index, new Data32(data32), highStart, index3NullOffset, dataNullOffset);
-        }
-
-        /**
-         * Creates a trie from its binary form.
-         * Same as {@link CodePointTrie#fromBinary(Type, ValueWidth, ByteBuffer)}
-         * with {@link Type#SMALL} and {@link ValueWidth#BITS_32}.
-         *
-         * @param bytes a buffer containing the binary data of a CodePointTrie
-         * @return the trie
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        public static Small32 fromBinary(ByteBuffer bytes) {
-            return (Small32) CodePointTrie.fromBinary(Type.SMALL, ValueWidth.BITS_32, bytes);
-        }
-    }
-
-    /**
-     * A CodePointTrie with {@link Type#SMALL} and {@link ValueWidth#BITS_8}.
-     *
-     * @draft ICU 63
-     * @provisional This API might change or be removed in a future release.
-     */
-    public static final class Small8 extends Small {
-        Small8(char[] index, byte[] data8, int highStart,
-                int index3NullOffset, int dataNullOffset) {
-            super(index, new Data8(data8), highStart, index3NullOffset, dataNullOffset);
-        }
-
-        /**
-         * Creates a trie from its binary form.
-         * Same as {@link CodePointTrie#fromBinary(Type, ValueWidth, ByteBuffer)}
-         * with {@link Type#SMALL} and {@link ValueWidth#BITS_8}.
-         *
-         * @param bytes a buffer containing the binary data of a CodePointTrie
-         * @return the trie
-         * @draft ICU 63
-         * @provisional This API might change or be removed in a future release.
-         */
-        public static Small8 fromBinary(ByteBuffer bytes) {
-            return (Small8) CodePointTrie.fromBinary(Type.SMALL, ValueWidth.BITS_8, bytes);
-        }
-    }
-}
--- /dev/null	2020-01-10 15:58:11.000000000 -0800
+++ new/src/java.base/share/classes/jdk/internal/icu/util/CodePointTrie.java	2020-01-10 15:58:11.000000000 -0800
@@ -0,0 +1,1312 @@
+/*
+ * Copyright (c) 2019, 2020, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.  Oracle designates this
+ * particular file as subject to the "Classpath" exception as provided
+ * by Oracle in the LICENSE file that accompanied this code.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+// (c) 2018 and later: Unicode, Inc. and others.
+// License & terms of use: http://www.unicode.org/copyright.html#License
+
+// created: 2018may04 Markus W. Scherer
+
+package jdk.internal.icu.util;
+
+import jdk.internal.icu.impl.ICUBinary;
+
+import java.io.DataOutputStream;
+import java.io.IOException;
+import java.io.UncheckedIOException;
+import java.io.OutputStream;
+import java.nio.ByteBuffer;
+import java.nio.ByteOrder;
+
+import static jdk.internal.icu.impl.NormalizerImpl.UTF16Plus;
+
+/**
+ * Immutable Unicode code point trie.
+ * Fast, reasonably compact, map from Unicode code points (U+0000..U+10FFFF) to integer values.
+ * For details see http://site.icu-project.org/design/struct/utrie
+ *
+ * 
This class is not intended for public subclassing.
+ *
+ * @see MutableCodePointTrie
+ * @draft ICU 63
+ * @provisional This API might change or be removed in a future release.
+ */
+@SuppressWarnings("deprecation")
+public abstract class CodePointTrie extends CodePointMap {
+    /**
+     * Selectors for the type of a CodePointTrie.
+     * Different trade-offs for size vs. speed.
+     *
+     * 
Use null for {@link #fromBinary} to accept any type;
+     * {@link #getType} will return the actual type.
+     *
+     * @see MutableCodePointTrie#buildImmutable(CodePointTrie.Type, CodePointTrie.ValueWidth)
+     * @see #fromBinary
+     * @see #getType
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    public enum Type {
+        /**
+         * Fast/simple/larger BMP data structure.
+         * The {@link Fast} subclasses have additional functions for lookup for BMP and supplementary code points.
+         *
+         * @see Fast
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        FAST,
+        /**
+         * Small/slower BMP data structure.
+         *
+         * @see Small
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        SMALL
+    }
+
+    /**
+     * Selectors for the number of bits in a CodePointTrie data value.
+     *
+     * 
Use null for {@link #fromBinary} to accept any data value width;
+     * {@link #getValueWidth} will return the actual data value width.
+     *
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    public enum ValueWidth {
+        /**
+         * The trie stores 16 bits per data value.
+         * It returns them as unsigned values 0..0xffff=65535.
+         *
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        BITS_16,
+        /**
+         * The trie stores 32 bits per data value.
+         *
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        BITS_32,
+        /**
+         * The trie stores 8 bits per data value.
+         * It returns them as unsigned values 0..0xff=255.
+         *
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        BITS_8
+    }
+
+    private CodePointTrie(char[] index, Data data, int highStart,
+            int index3NullOffset, int dataNullOffset) {
+        this.ascii = new int[ASCII_LIMIT];
+        this.index = index;
+        this.data = data;
+        this.dataLength = data.getDataLength();
+        this.highStart = highStart;
+        this.index3NullOffset = index3NullOffset;
+        this.dataNullOffset = dataNullOffset;
+
+        for (int c = 0; c < ASCII_LIMIT; ++c) {
+            ascii[c] = data.getFromIndex(c);
+        }
+
+        int nullValueOffset = dataNullOffset;
+        if (nullValueOffset >= dataLength) {
+            nullValueOffset = dataLength - HIGH_VALUE_NEG_DATA_OFFSET;
+        }
+        nullValue = data.getFromIndex(nullValueOffset);
+    }
+
+    /**
+     * Creates a trie from its binary form,
+     * stored in the ByteBuffer starting at the current position.
+     * Advances the buffer position to just after the trie data.
+     * Inverse of {@link #toBinary(OutputStream)}.
+     *
+     * 
The data is copied from the buffer;
+     * later modification of the buffer will not affect the trie.
+     *
+     * @param type selects the trie type; this method throws an exception
+     *             if the type does not match the binary data;
+     *             use null to accept any type
+     * @param valueWidth selects the number of bits in a data value; this method throws an exception
+     *                  if the valueWidth does not match the binary data;
+     *                  use null to accept any data value width
+     * @param bytes a buffer containing the binary data of a CodePointTrie
+     * @return the trie
+     * @see MutableCodePointTrie#MutableCodePointTrie(int, int)
+     * @see MutableCodePointTrie#buildImmutable(CodePointTrie.Type, CodePointTrie.ValueWidth)
+     * @see #toBinary(OutputStream)
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    public static CodePointTrie fromBinary(Type type, ValueWidth valueWidth, ByteBuffer bytes) {
+        ByteOrder outerByteOrder = bytes.order();
+        try {
+            // Enough data for a trie header?
+            if (bytes.remaining() < 16 /* sizeof(UCPTrieHeader) */) {
+                throw new InternalError("Buffer too short for a CodePointTrie header");
+            }
+
+            // struct UCPTrieHeader
+            /** "Tri3" in big-endian US-ASCII (0x54726933) */
+            int signature = bytes.getInt();
+
+            // Check the signature.
+            switch (signature) {
+            case 0x54726933:
+                // The buffer is already set to the trie data byte order.
+                break;
+            case 0x33697254:
+                // Temporarily reverse the byte order.
+                boolean isBigEndian = outerByteOrder == ByteOrder.BIG_ENDIAN;
+                bytes.order(isBigEndian ? ByteOrder.LITTLE_ENDIAN : ByteOrder.BIG_ENDIAN);
+                signature = 0x54726933;
+                break;
+            default:
+                throw new InternalError("Buffer does not contain a serialized CodePointTrie");
+            }
+
+            // struct UCPTrieHeader continued
+            /**
+             * Options bit field:
+             * Bits 15..12: Data length bits 19..16.
+             * Bits 11..8: Data null block offset bits 19..16.
+             * Bits 7..6: UCPTrieType
+             * Bits 5..3: Reserved (0).
+             * Bits 2..0: UCPTrieValueWidth
+             */
+            int options = bytes.getChar();
+
+            /** Total length of the index tables. */
+            int indexLength = bytes.getChar();
+
+            /** Data length bits 15..0. */
+            int dataLength = bytes.getChar();
+
+            /** Index-3 null block offset, 0x7fff or 0xffff if none. */
+            int index3NullOffset = bytes.getChar();
+
+            /** Data null block offset bits 15..0, 0xfffff if none. */
+            int dataNullOffset = bytes.getChar();
+
+            /**
+             * First code point of the single-value range ending with U+10ffff,
+             * rounded up and then shifted right by SHIFT_2.
+             */
+            int shiftedHighStart = bytes.getChar();
+            // struct UCPTrieHeader end
+
+            int typeInt = (options >> 6) & 3;
+            Type actualType;
+            switch (typeInt) {
+            case 0: actualType = Type.FAST; break;
+            case 1: actualType = Type.SMALL; break;
+            default:
+                throw new InternalError("CodePointTrie data header has an unsupported type");
+            }
+
+            int valueWidthInt = options & OPTIONS_VALUE_BITS_MASK;
+            ValueWidth actualValueWidth;
+            switch (valueWidthInt) {
+            case 0: actualValueWidth = ValueWidth.BITS_16; break;
+            case 1: actualValueWidth = ValueWidth.BITS_32; break;
+            case 2: actualValueWidth = ValueWidth.BITS_8; break;
+            default:
+                throw new InternalError("CodePointTrie data header has an unsupported value width");
+            }
+
+            if ((options & OPTIONS_RESERVED_MASK) != 0) {
+                throw new InternalError("CodePointTrie data header has unsupported options");
+            }
+
+            if (type == null) {
+                type = actualType;
+            }
+            if (valueWidth == null) {
+                valueWidth = actualValueWidth;
+            }
+            if (type != actualType || valueWidth != actualValueWidth) {
+                throw new InternalError("CodePointTrie data header has a different type or value width than required");
+            }
+
+            // Get the length values and offsets.
+            dataLength |= ((options & OPTIONS_DATA_LENGTH_MASK) << 4);
+            dataNullOffset |= ((options & OPTIONS_DATA_NULL_OFFSET_MASK) << 8);
+
+            int highStart = shiftedHighStart << SHIFT_2;
+
+            // Calculate the actual length, minus the header.
+            int actualLength = indexLength * 2;
+            if (valueWidth == ValueWidth.BITS_16) {
+                actualLength += dataLength * 2;
+            } else if (valueWidth == ValueWidth.BITS_32) {
+                actualLength += dataLength * 4;
+            } else {
+                actualLength += dataLength;
+            }
+            if (bytes.remaining() < actualLength) {
+                throw new InternalError("Buffer too short for the CodePointTrie data");
+            }
+
+            char[] index = ICUBinary.getChars(bytes, indexLength, 0);
+            switch (valueWidth) {
+            case BITS_16: {
+                char[] data16 = ICUBinary.getChars(bytes, dataLength, 0);
+                return type == Type.FAST ?
+                        new Fast16(index, data16, highStart, index3NullOffset, dataNullOffset) :
+                            new Small16(index, data16, highStart, index3NullOffset, dataNullOffset);
+            }
+            case BITS_32: {
+                int[] data32 = ICUBinary.getInts(bytes, dataLength, 0);
+                return type == Type.FAST ?
+                        new Fast32(index, data32, highStart, index3NullOffset, dataNullOffset) :
+                            new Small32(index, data32, highStart, index3NullOffset, dataNullOffset);
+            }
+            case BITS_8: {
+                byte[] data8 = ICUBinary.getBytes(bytes, dataLength, 0);
+                return type == Type.FAST ?
+                        new Fast8(index, data8, highStart, index3NullOffset, dataNullOffset) :
+                            new Small8(index, data8, highStart, index3NullOffset, dataNullOffset);
+            }
+            default:
+                throw new AssertionError("should be unreachable");
+            }
+        } finally {
+            bytes.order(outerByteOrder);
+        }
+    }
+
+    /**
+     * Returns the trie type.
+     *
+     * @return the trie type
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    public abstract Type getType();
+    /**
+     * Returns the number of bits in a trie data value.
+     *
+     * @return the number of bits in a trie data value
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    public final ValueWidth getValueWidth() { return data.getValueWidth(); }
+
+    /**
+     * {@inheritDoc}
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    @Override
+    public int get(int c) {
+        return data.getFromIndex(cpIndex(c));
+    }
+
+    /**
+     * Returns a trie value for an ASCII code point, without range checking.
+     *
+     * @param c the input code point; must be U+0000..U+007F
+     * @return The ASCII code point's trie value.
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    public final int asciiGet(int c) {
+        return ascii[c];
+    }
+
+    private static final int MAX_UNICODE = 0x10ffff;
+
+    private static final int ASCII_LIMIT = 0x80;
+
+    private static final int maybeFilterValue(int value, int trieNullValue, int nullValue,
+            ValueFilter filter) {
+        if (value == trieNullValue) {
+            value = nullValue;
+        } else if (filter != null) {
+            value = filter.apply(value);
+        }
+        return value;
+    }
+
+    /**
+     * {@inheritDoc}
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    @Override
+    public final boolean getRange(int start, ValueFilter filter, Range range) {
+        if (start < 0 || MAX_UNICODE < start) {
+            return false;
+        }
+        if (start >= highStart) {
+            int di = dataLength - HIGH_VALUE_NEG_DATA_OFFSET;
+            int value = data.getFromIndex(di);
+            if (filter != null) { value = filter.apply(value); }
+            range.set(start, MAX_UNICODE, value);
+            return true;
+        }
+
+        int nullValue = this.nullValue;
+        if (filter != null) { nullValue = filter.apply(nullValue); }
+        Type type = getType();
+
+        int prevI3Block = -1;
+        int prevBlock = -1;
+        int c = start;
+        // Initialize to make compiler happy. Real value when haveValue is true.
+        int trieValue = 0, value = 0;
+        boolean haveValue = false;
+        do {
+            int i3Block;
+            int i3;
+            int i3BlockLength;
+            int dataBlockLength;
+            if (c <= 0xffff && (type == Type.FAST || c <= SMALL_MAX)) {
+                i3Block = 0;
+                i3 = c >> FAST_SHIFT;
+                i3BlockLength = type == Type.FAST ? BMP_INDEX_LENGTH : SMALL_INDEX_LENGTH;
+                dataBlockLength = FAST_DATA_BLOCK_LENGTH;
+            } else {
+                // Use the multi-stage index.
+                int i1 = c >> SHIFT_1;
+                if (type == Type.FAST) {
+                    assert(0xffff < c && c < highStart);
+                    i1 += BMP_INDEX_LENGTH - OMITTED_BMP_INDEX_1_LENGTH;
+                } else {
+                    assert(c < highStart && highStart > SMALL_LIMIT);
+                    i1 += SMALL_INDEX_LENGTH;
+                }
+                i3Block = index[index[i1] + ((c >> SHIFT_2) & INDEX_2_MASK)];
+                if (i3Block == prevI3Block && (c - start) >= CP_PER_INDEX_2_ENTRY) {
+                    // The index-3 block is the same as the previous one, and filled with value.
+                    assert((c & (CP_PER_INDEX_2_ENTRY - 1)) == 0);
+                    c += CP_PER_INDEX_2_ENTRY;
+                    continue;
+                }
+                prevI3Block = i3Block;
+                if (i3Block == index3NullOffset) {
+                    // This is the index-3 null block.
+                    if (haveValue) {
+                        if (nullValue != value) {
+                            range.set(start, c - 1, value);
+                            return true;
+                        }
+                    } else {
+                        trieValue = this.nullValue;
+                        value = nullValue;
+                        haveValue = true;
+                    }
+                    prevBlock = dataNullOffset;
+                    c = (c + CP_PER_INDEX_2_ENTRY) & ~(CP_PER_INDEX_2_ENTRY - 1);
+                    continue;
+                }
+                i3 = (c >> SHIFT_3) & INDEX_3_MASK;
+                i3BlockLength = INDEX_3_BLOCK_LENGTH;
+                dataBlockLength = SMALL_DATA_BLOCK_LENGTH;
+            }
+            // Enumerate data blocks for one index-3 block.
+            do {
+                int block;
+                if ((i3Block & 0x8000) == 0) {
+                    block = index[i3Block + i3];
+                } else {
+                    // 18-bit indexes stored in groups of 9 entries per 8 indexes.
+                    int group = (i3Block & 0x7fff) + (i3 & ~7) + (i3 >> 3);
+                    int gi = i3 & 7;
+                    block = (index[group++] << (2 + (2 * gi))) & 0x30000;
+                    block |= index[group + gi];
+                }
+                if (block == prevBlock && (c - start) >= dataBlockLength) {
+                    // The block is the same as the previous one, and filled with value.
+                    assert((c & (dataBlockLength - 1)) == 0);
+                    c += dataBlockLength;
+                } else {
+                    int dataMask = dataBlockLength - 1;
+                    prevBlock = block;
+                    if (block == dataNullOffset) {
+                        // This is the data null block.
+                        if (haveValue) {
+                            if (nullValue != value) {
+                                range.set(start, c - 1, value);
+                                return true;
+                            }
+                        } else {
+                            trieValue = this.nullValue;
+                            value = nullValue;
+                            haveValue = true;
+                        }
+                        c = (c + dataBlockLength) & ~dataMask;
+                    } else {
+                        int di = block + (c & dataMask);
+                        int trieValue2 = data.getFromIndex(di);
+                        if (haveValue) {
+                            if (trieValue2 != trieValue) {
+                                if (filter == null ||
+                                        maybeFilterValue(trieValue2, this.nullValue, nullValue,
+                                                filter) != value) {
+                                    range.set(start, c - 1, value);
+                                    return true;
+                                }
+                                trieValue = trieValue2;  // may or may not help
+                            }
+                        } else {
+                            trieValue = trieValue2;
+                            value = maybeFilterValue(trieValue2, this.nullValue, nullValue, filter);
+                            haveValue = true;
+                        }
+                        while ((++c & dataMask) != 0) {
+                            trieValue2 = data.getFromIndex(++di);
+                            if (trieValue2 != trieValue) {
+                                if (filter == null ||
+                                        maybeFilterValue(trieValue2, this.nullValue, nullValue,
+                                                filter) != value) {
+                                    range.set(start, c - 1, value);
+                                    return true;
+                                }
+                                trieValue = trieValue2;  // may or may not help
+                            }
+                        }
+                    }
+                }
+            } while (++i3 < i3BlockLength);
+        } while (c < highStart);
+        assert(haveValue);
+        int di = dataLength - HIGH_VALUE_NEG_DATA_OFFSET;
+        int highValue = data.getFromIndex(di);
+        if (maybeFilterValue(highValue, this.nullValue, nullValue, filter) != value) {
+            --c;
+        } else {
+            c = MAX_UNICODE;
+        }
+        range.set(start, c, value);
+        return true;
+    }
+
+    /**
+     * Writes a representation of the trie to the output stream.
+     * Inverse of {@link #fromBinary}.
+     *
+     * @param os the output stream
+     * @return the number of bytes written
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    public final int toBinary(OutputStream os) {
+        try {
+            DataOutputStream dos = new DataOutputStream(os);
+
+            // Write the UCPTrieHeader
+            dos.writeInt(0x54726933);  // signature="Tri3"
+            dos.writeChar(  // options
+                ((dataLength & 0xf0000) >> 4) |
+                ((dataNullOffset & 0xf0000) >> 8) |
+                (getType().ordinal() << 6) |
+                getValueWidth().ordinal());
+            dos.writeChar(index.length);
+            dos.writeChar(dataLength);
+            dos.writeChar(index3NullOffset);
+            dos.writeChar(dataNullOffset);
+            dos.writeChar(highStart >> SHIFT_2);  // shiftedHighStart
+            int length = 16;  // sizeof(UCPTrieHeader)
+
+            for (char i : index) { dos.writeChar(i); }
+            length += index.length * 2;
+            length += data.write(dos);
+            return length;
+        } catch (IOException e) {
+            throw new UncheckedIOException(e);
+        }
+    }
+
+    /** @internal */
+    static final int FAST_SHIFT = 6;
+
+    /** Number of entries in a data block for code points below the fast limit. 64=0x40 @internal */
+    static final int FAST_DATA_BLOCK_LENGTH = 1 << FAST_SHIFT;
+
+    /** Mask for getting the lower bits for the in-fast-data-block offset. @internal */
+    private static final int FAST_DATA_MASK = FAST_DATA_BLOCK_LENGTH - 1;
+
+    /** @internal */
+    private static final int SMALL_MAX = 0xfff;
+
+    /**
+     * Offset from dataLength (to be subtracted) for fetching the
+     * value returned for out-of-range code points and ill-formed UTF-8/16.
+     * @internal
+     */
+    private static final int ERROR_VALUE_NEG_DATA_OFFSET = 1;
+    /**
+     * Offset from dataLength (to be subtracted) for fetching the
+     * value returned for code points highStart..U+10FFFF.
+     * @internal
+     */
+    private static final int HIGH_VALUE_NEG_DATA_OFFSET = 2;
+
+    // ucptrie_impl.h
+
+    /** The length of the BMP index table. 1024=0x400 */
+    private static final int BMP_INDEX_LENGTH = 0x10000 >> FAST_SHIFT;
+
+    static final int SMALL_LIMIT = 0x1000;
+    private static final int SMALL_INDEX_LENGTH = SMALL_LIMIT >> FAST_SHIFT;
+
+    /** Shift size for getting the index-3 table offset. */
+    static final int SHIFT_3 = 4;
+
+    /** Shift size for getting the index-2 table offset. */
+    private static final int SHIFT_2 = 5 + SHIFT_3;
+
+    /** Shift size for getting the index-1 table offset. */
+    private static final int SHIFT_1 = 5 + SHIFT_2;
+
+    /**
+     * Difference between two shift sizes,
+     * for getting an index-2 offset from an index-3 offset. 5=9-4
+     */
+    static final int SHIFT_2_3 = SHIFT_2 - SHIFT_3;
+
+    /**
+     * Difference between two shift sizes,
+     * for getting an index-1 offset from an index-2 offset. 5=14-9
+     */
+    static final int SHIFT_1_2 = SHIFT_1 - SHIFT_2;
+
+    /**
+     * Number of index-1 entries for the BMP. (4)
+     * This part of the index-1 table is omitted from the serialized form.
+     */
+    private static final int OMITTED_BMP_INDEX_1_LENGTH = 0x10000 >> SHIFT_1;
+
+    /** Number of entries in an index-2 block. 32=0x20 */
+    static final int INDEX_2_BLOCK_LENGTH = 1 << SHIFT_1_2;
+
+    /** Mask for getting the lower bits for the in-index-2-block offset. */
+    static final int INDEX_2_MASK = INDEX_2_BLOCK_LENGTH - 1;
+
+    /** Number of code points per index-2 table entry. 512=0x200 */
+    static final int CP_PER_INDEX_2_ENTRY = 1 << SHIFT_2;
+
+    /** Number of entries in an index-3 block. 32=0x20 */
+    static final int INDEX_3_BLOCK_LENGTH = 1 << SHIFT_2_3;
+
+    /** Mask for getting the lower bits for the in-index-3-block offset. */
+    private static final int INDEX_3_MASK = INDEX_3_BLOCK_LENGTH - 1;
+
+    /** Number of entries in a small data block. 16=0x10 */
+    static final int SMALL_DATA_BLOCK_LENGTH = 1 << SHIFT_3;
+
+    /** Mask for getting the lower bits for the in-small-data-block offset. */
+    static final int SMALL_DATA_MASK = SMALL_DATA_BLOCK_LENGTH - 1;
+
+    // ucptrie_impl.h: Constants for use with UCPTrieHeader.options.
+    private static final int OPTIONS_DATA_LENGTH_MASK = 0xf000;
+    private static final int OPTIONS_DATA_NULL_OFFSET_MASK = 0xf00;
+    private static final int OPTIONS_RESERVED_MASK = 0x38;
+    private static final int OPTIONS_VALUE_BITS_MASK = 7;
+    /**
+     * Value for index3NullOffset which indicates that there is no index-3 null block.
+     * Bit 15 is unused for this value because this bit is used if the index-3 contains
+     * 18-bit indexes.
+     */
+    static final int NO_INDEX3_NULL_OFFSET = 0x7fff;
+    static final int NO_DATA_NULL_OFFSET = 0xfffff;
+
+    private static abstract class Data {
+        abstract ValueWidth getValueWidth();
+        abstract int getDataLength();
+        abstract int getFromIndex(int index);
+        abstract int write(DataOutputStream dos) throws IOException;
+    }
+
+    private static final class Data16 extends Data {
+        char[] array;
+        Data16(char[] a) { array = a; }
+        @Override ValueWidth getValueWidth() { return ValueWidth.BITS_16; }
+        @Override int getDataLength() { return array.length; }
+        @Override int getFromIndex(int index) { return array[index]; }
+        @Override int write(DataOutputStream dos) throws IOException {
+            for (char v : array) { dos.writeChar(v); }
+            return array.length * 2;
+        }
+    }
+
+    private static final class Data32 extends Data {
+        int[] array;
+        Data32(int[] a) { array = a; }
+        @Override ValueWidth getValueWidth() { return ValueWidth.BITS_32; }
+        @Override int getDataLength() { return array.length; }
+        @Override int getFromIndex(int index) { return array[index]; }
+        @Override int write(DataOutputStream dos) throws IOException {
+            for (int v : array) { dos.writeInt(v); }
+            return array.length * 4;
+        }
+    }
+
+    private static final class Data8 extends Data {
+        byte[] array;
+        Data8(byte[] a) { array = a; }
+        @Override ValueWidth getValueWidth() { return ValueWidth.BITS_8; }
+        @Override int getDataLength() { return array.length; }
+        @Override int getFromIndex(int index) { return array[index] & 0xff; }
+        @Override int write(DataOutputStream dos) throws IOException {
+            for (byte v : array) { dos.writeByte(v); }
+            return array.length;
+        }
+    }
+
+    /** @internal */
+    private final int[] ascii;
+
+    /** @internal */
+    private final char[] index;
+
+    /**
+     * @internal
+     * @deprecated This API is ICU internal only.
+     */
+    @Deprecated
+    protected final Data data;
+    /**
+     * @internal
+     * @deprecated This API is ICU internal only.
+     */
+    @Deprecated
+    protected final int dataLength;
+    /**
+     * Start of the last range which ends at U+10FFFF.
+     * @internal
+     * @deprecated This API is ICU internal only.
+     */
+    @Deprecated
+    protected final int highStart;
+
+    /**
+     * Internal index-3 null block offset.
+     * Set to an impossibly high value (e.g., 0xffff) if there is no dedicated index-3 null block.
+     * @internal
+     */
+    private final int index3NullOffset;
+    /**
+     * Internal data null block offset, not shifted.
+     * Set to an impossibly high value (e.g., 0xfffff) if there is no dedicated data null block.
+     * @internal
+     */
+    private final int dataNullOffset;
+    /** @internal */
+    private final int nullValue;
+
+    /**
+     * @internal
+     * @deprecated This API is ICU internal only.
+     */
+    @Deprecated
+    protected final int fastIndex(int c) {
+        return index[c >> FAST_SHIFT] + (c & FAST_DATA_MASK);
+    }
+
+    /**
+     * @internal
+     * @deprecated This API is ICU internal only.
+     */
+    @Deprecated
+    protected final int smallIndex(Type type, int c) {
+        // Split into two methods to make this part inline-friendly.
+        // In C, this part is a macro.
+        if (c >= highStart) {
+            return dataLength - HIGH_VALUE_NEG_DATA_OFFSET;
+        }
+        return internalSmallIndex(type, c);
+    }
+
+    private final int internalSmallIndex(Type type, int c) {
+        int i1 = c >> SHIFT_1;
+        if (type == Type.FAST) {
+            assert(0xffff < c && c < highStart);
+            i1 += BMP_INDEX_LENGTH - OMITTED_BMP_INDEX_1_LENGTH;
+        } else {
+            assert(0 <= c && c < highStart && highStart > SMALL_LIMIT);
+            i1 += SMALL_INDEX_LENGTH;
+        }
+        int i3Block = index[index[i1] + ((c >> SHIFT_2) & INDEX_2_MASK)];
+        int i3 = (c >> SHIFT_3) & INDEX_3_MASK;
+        int dataBlock;
+        if ((i3Block & 0x8000) == 0) {
+            // 16-bit indexes
+            dataBlock = index[i3Block + i3];
+        } else {
+            // 18-bit indexes stored in groups of 9 entries per 8 indexes.
+            i3Block = (i3Block & 0x7fff) + (i3 & ~7) + (i3 >> 3);
+            i3 &= 7;
+            dataBlock = (index[i3Block++] << (2 + (2 * i3))) & 0x30000;
+            dataBlock |= index[i3Block + i3];
+        }
+        return dataBlock + (c & SMALL_DATA_MASK);
+    }
+
+    /**
+     * @internal
+     * @deprecated This API is ICU internal only.
+     */
+    @Deprecated
+    protected abstract int cpIndex(int c);
+
+    /**
+     * A CodePointTrie with {@link Type#FAST}.
+     *
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    public static abstract class Fast extends CodePointTrie {
+        private Fast(char[] index, Data data, int highStart,
+                int index3NullOffset, int dataNullOffset) {
+            super(index, data, highStart, index3NullOffset, dataNullOffset);
+        }
+
+        /**
+         * Creates a trie from its binary form.
+         * Same as {@link CodePointTrie#fromBinary(Type, ValueWidth, ByteBuffer)}
+         * with {@link Type#FAST}.
+         *
+         * @param valueWidth selects the number of bits in a data value; this method throws an exception
+         *                  if the valueWidth does not match the binary data;
+         *                  use null to accept any data value width
+         * @param bytes a buffer containing the binary data of a CodePointTrie
+         * @return the trie
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        public static Fast fromBinary(ValueWidth valueWidth, ByteBuffer bytes) {
+            return (Fast) CodePointTrie.fromBinary(Type.FAST, valueWidth, bytes);
+        }
+
+        /**
+         * @return {@link Type#FAST}
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        @Override
+        public final Type getType() { return Type.FAST; }
+
+        /**
+         * Returns a trie value for a BMP code point (U+0000..U+FFFF), without range checking.
+         * Can be used to look up a value for a UTF-16 code unit if other parts of
+         * the string processing check for surrogates.
+         *
+         * @param c the input code point, must be U+0000..U+FFFF
+         * @return The BMP code point's trie value.
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        public abstract int bmpGet(int c);
+
+        /**
+         * Returns a trie value for a supplementary code point (U+10000..U+10FFFF),
+         * without range checking.
+         *
+         * @param c the input code point, must be U+10000..U+10FFFF
+         * @return The supplementary code point's trie value.
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        public abstract int suppGet(int c);
+
+        /**
+         * @internal
+         * @deprecated This API is ICU internal only.
+         */
+        @Deprecated
+        @Override
+        protected final int cpIndex(int c) {
+            if (c >= 0) {
+                if (c <= 0xffff) {
+                    return fastIndex(c);
+                } else if (c <= 0x10ffff) {
+                    return smallIndex(Type.FAST, c);
+                }
+            }
+            return dataLength - ERROR_VALUE_NEG_DATA_OFFSET;
+        }
+
+        /**
+         * {@inheritDoc}
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        @Override
+        public final StringIterator stringIterator(CharSequence s, int sIndex) {
+            return new FastStringIterator(s, sIndex);
+        }
+
+        private final class FastStringIterator extends StringIterator {
+            private FastStringIterator(CharSequence s, int sIndex) {
+                super(s, sIndex);
+            }
+
+            @Override
+            public boolean next() {
+                if (sIndex >= s.length()) {
+                    return false;
+                }
+                char lead = s.charAt(sIndex++);
+                c = lead;
+                int dataIndex;
+                if (!Character.isSurrogate(lead)) {
+                    dataIndex = fastIndex(c);
+                } else {
+                    char trail;
+                    if (UTF16Plus.isSurrogateLead(lead) && sIndex < s.length() &&
+                            Character.isLowSurrogate(trail = s.charAt(sIndex))) {
+                        ++sIndex;
+                        c = Character.toCodePoint(lead, trail);
+                        dataIndex = smallIndex(Type.FAST, c);
+                    } else {
+                        dataIndex = dataLength - ERROR_VALUE_NEG_DATA_OFFSET;
+                    }
+                }
+                value = data.getFromIndex(dataIndex);
+                return true;
+            }
+
+            @Override
+            public boolean previous() {
+                if (sIndex <= 0) {
+                    return false;
+                }
+                char trail = s.charAt(--sIndex);
+                c = trail;
+                int dataIndex;
+                if (!Character.isSurrogate(trail)) {
+                    dataIndex = fastIndex(c);
+                } else {
+                    char lead;
+                    if (!UTF16Plus.isSurrogateLead(trail) && sIndex > 0 &&
+                            Character.isHighSurrogate(lead = s.charAt(sIndex - 1))) {
+                        --sIndex;
+                        c = Character.toCodePoint(lead, trail);
+                        dataIndex = smallIndex(Type.FAST, c);
+                    } else {
+                        dataIndex = dataLength - ERROR_VALUE_NEG_DATA_OFFSET;
+                    }
+                }
+                value = data.getFromIndex(dataIndex);
+                return true;
+            }
+        }
+    }
+
+    /**
+     * A CodePointTrie with {@link Type#SMALL}.
+     *
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    public static abstract class Small extends CodePointTrie {
+        private Small(char[] index, Data data, int highStart,
+                int index3NullOffset, int dataNullOffset) {
+            super(index, data, highStart, index3NullOffset, dataNullOffset);
+        }
+
+        /**
+         * Creates a trie from its binary form.
+         * Same as {@link CodePointTrie#fromBinary(Type, ValueWidth, ByteBuffer)}
+         * with {@link Type#SMALL}.
+         *
+         * @param valueWidth selects the number of bits in a data value; this method throws an exception
+         *                  if the valueWidth does not match the binary data;
+         *                  use null to accept any data value width
+         * @param bytes a buffer containing the binary data of a CodePointTrie
+         * @return the trie
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        public static Small fromBinary(ValueWidth valueWidth, ByteBuffer bytes) {
+            return (Small) CodePointTrie.fromBinary(Type.SMALL, valueWidth, bytes);
+        }
+
+        /**
+         * @return {@link Type#SMALL}
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        @Override
+        public final Type getType() { return Type.SMALL; }
+
+        /**
+         * @internal
+         * @deprecated This API is ICU internal only.
+         */
+        @Deprecated
+        @Override
+        protected final int cpIndex(int c) {
+            if (c >= 0) {
+                if (c <= SMALL_MAX) {
+                    return fastIndex(c);
+                } else if (c <= 0x10ffff) {
+                    return smallIndex(Type.SMALL, c);
+                }
+            }
+            return dataLength - ERROR_VALUE_NEG_DATA_OFFSET;
+        }
+
+        /**
+         * {@inheritDoc}
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        @Override
+        public final StringIterator stringIterator(CharSequence s, int sIndex) {
+            return new SmallStringIterator(s, sIndex);
+        }
+
+        private final class SmallStringIterator extends StringIterator {
+            private SmallStringIterator(CharSequence s, int sIndex) {
+                super(s, sIndex);
+            }
+
+            @Override
+            public boolean next() {
+                if (sIndex >= s.length()) {
+                    return false;
+                }
+                char lead = s.charAt(sIndex++);
+                c = lead;
+                int dataIndex;
+                if (!Character.isSurrogate(lead)) {
+                    dataIndex = cpIndex(c);
+                } else {
+                    char trail;
+                    if (UTF16Plus.isSurrogateLead(lead) && sIndex < s.length() &&
+                            Character.isLowSurrogate(trail = s.charAt(sIndex))) {
+                        ++sIndex;
+                        c = Character.toCodePoint(lead, trail);
+                        dataIndex = smallIndex(Type.SMALL, c);
+                    } else {
+                        dataIndex = dataLength - ERROR_VALUE_NEG_DATA_OFFSET;
+                    }
+                }
+                value = data.getFromIndex(dataIndex);
+                return true;
+            }
+
+            @Override
+            public boolean previous() {
+                if (sIndex <= 0) {
+                    return false;
+                }
+                char trail = s.charAt(--sIndex);
+                c = trail;
+                int dataIndex;
+                if (!Character.isSurrogate(trail)) {
+                    dataIndex = cpIndex(c);
+                } else {
+                    char lead;
+                    if (!UTF16Plus.isSurrogateLead(trail) && sIndex > 0 &&
+                            Character.isHighSurrogate(lead = s.charAt(sIndex - 1))) {
+                        --sIndex;
+                        c = Character.toCodePoint(lead, trail);
+                        dataIndex = smallIndex(Type.SMALL, c);
+                    } else {
+                        dataIndex = dataLength - ERROR_VALUE_NEG_DATA_OFFSET;
+                    }
+                }
+                value = data.getFromIndex(dataIndex);
+                return true;
+            }
+        }
+    }
+
+    /**
+     * A CodePointTrie with {@link Type#FAST} and {@link ValueWidth#BITS_16}.
+     *
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    public static final class Fast16 extends Fast {
+        private final char[] dataArray;
+
+        Fast16(char[] index, char[] data16, int highStart,
+                int index3NullOffset, int dataNullOffset) {
+            super(index, new Data16(data16), highStart, index3NullOffset, dataNullOffset);
+            this.dataArray = data16;
+        }
+
+        /**
+         * Creates a trie from its binary form.
+         * Same as {@link CodePointTrie#fromBinary(Type, ValueWidth, ByteBuffer)}
+         * with {@link Type#FAST} and {@link ValueWidth#BITS_16}.
+         *
+         * @param bytes a buffer containing the binary data of a CodePointTrie
+         * @return the trie
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        public static Fast16 fromBinary(ByteBuffer bytes) {
+            return (Fast16) CodePointTrie.fromBinary(Type.FAST, ValueWidth.BITS_16, bytes);
+        }
+
+        /**
+         * {@inheritDoc}
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        @Override
+        public final int get(int c) {
+            return dataArray[cpIndex(c)];
+        }
+
+        /**
+         * {@inheritDoc}
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        @Override
+        public final int bmpGet(int c) {
+            assert 0 <= c && c <= 0xffff;
+            return dataArray[fastIndex(c)];
+        }
+
+        /**
+         * {@inheritDoc}
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        @Override
+        public final int suppGet(int c) {
+            assert 0x10000 <= c && c <= 0x10ffff;
+            return dataArray[smallIndex(Type.FAST, c)];
+        }
+    }
+
+    /**
+     * A CodePointTrie with {@link Type#FAST} and {@link ValueWidth#BITS_32}.
+     *
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    public static final class Fast32 extends Fast {
+        private final int[] dataArray;
+
+        Fast32(char[] index, int[] data32, int highStart,
+                int index3NullOffset, int dataNullOffset) {
+            super(index, new Data32(data32), highStart, index3NullOffset, dataNullOffset);
+            this.dataArray = data32;
+        }
+
+        /**
+         * Creates a trie from its binary form.
+         * Same as {@link CodePointTrie#fromBinary(Type, ValueWidth, ByteBuffer)}
+         * with {@link Type#FAST} and {@link ValueWidth#BITS_32}.
+         *
+         * @param bytes a buffer containing the binary data of a CodePointTrie
+         * @return the trie
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        public static Fast32 fromBinary(ByteBuffer bytes) {
+            return (Fast32) CodePointTrie.fromBinary(Type.FAST, ValueWidth.BITS_32, bytes);
+        }
+
+        /**
+         * {@inheritDoc}
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        @Override
+        public final int get(int c) {
+            return dataArray[cpIndex(c)];
+        }
+
+        /**
+         * {@inheritDoc}
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        @Override
+        public final int bmpGet(int c) {
+            assert 0 <= c && c <= 0xffff;
+            return dataArray[fastIndex(c)];
+        }
+
+        /**
+         * {@inheritDoc}
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        @Override
+        public final int suppGet(int c) {
+            assert 0x10000 <= c && c <= 0x10ffff;
+            return dataArray[smallIndex(Type.FAST, c)];
+        }
+    }
+
+    /**
+     * A CodePointTrie with {@link Type#FAST} and {@link ValueWidth#BITS_8}.
+     *
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    public static final class Fast8 extends Fast {
+        private final byte[] dataArray;
+
+        Fast8(char[] index, byte[] data8, int highStart,
+                int index3NullOffset, int dataNullOffset) {
+            super(index, new Data8(data8), highStart, index3NullOffset, dataNullOffset);
+            this.dataArray = data8;
+        }
+
+        /**
+         * Creates a trie from its binary form.
+         * Same as {@link CodePointTrie#fromBinary(Type, ValueWidth, ByteBuffer)}
+         * with {@link Type#FAST} and {@link ValueWidth#BITS_8}.
+         *
+         * @param bytes a buffer containing the binary data of a CodePointTrie
+         * @return the trie
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        public static Fast8 fromBinary(ByteBuffer bytes) {
+            return (Fast8) CodePointTrie.fromBinary(Type.FAST, ValueWidth.BITS_8, bytes);
+        }
+
+        /**
+         * {@inheritDoc}
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        @Override
+        public final int get(int c) {
+            return dataArray[cpIndex(c)] & 0xff;
+        }
+
+        /**
+         * {@inheritDoc}
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        @Override
+        public final int bmpGet(int c) {
+            assert 0 <= c && c <= 0xffff;
+            return dataArray[fastIndex(c)] & 0xff;
+        }
+
+        /**
+         * {@inheritDoc}
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        @Override
+        public final int suppGet(int c) {
+            assert 0x10000 <= c && c <= 0x10ffff;
+            return dataArray[smallIndex(Type.FAST, c)] & 0xff;
+        }
+    }
+
+    /**
+     * A CodePointTrie with {@link Type#SMALL} and {@link ValueWidth#BITS_16}.
+     *
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    public static final class Small16 extends Small {
+        Small16(char[] index, char[] data16, int highStart,
+                int index3NullOffset, int dataNullOffset) {
+            super(index, new Data16(data16), highStart, index3NullOffset, dataNullOffset);
+        }
+
+        /**
+         * Creates a trie from its binary form.
+         * Same as {@link CodePointTrie#fromBinary(Type, ValueWidth, ByteBuffer)}
+         * with {@link Type#SMALL} and {@link ValueWidth#BITS_16}.
+         *
+         * @param bytes a buffer containing the binary data of a CodePointTrie
+         * @return the trie
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        public static Small16 fromBinary(ByteBuffer bytes) {
+            return (Small16) CodePointTrie.fromBinary(Type.SMALL, ValueWidth.BITS_16, bytes);
+        }
+    }
+
+    /**
+     * A CodePointTrie with {@link Type#SMALL} and {@link ValueWidth#BITS_32}.
+     *
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    public static final class Small32 extends Small {
+        Small32(char[] index, int[] data32, int highStart,
+                int index3NullOffset, int dataNullOffset) {
+            super(index, new Data32(data32), highStart, index3NullOffset, dataNullOffset);
+        }
+
+        /**
+         * Creates a trie from its binary form.
+         * Same as {@link CodePointTrie#fromBinary(Type, ValueWidth, ByteBuffer)}
+         * with {@link Type#SMALL} and {@link ValueWidth#BITS_32}.
+         *
+         * @param bytes a buffer containing the binary data of a CodePointTrie
+         * @return the trie
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        public static Small32 fromBinary(ByteBuffer bytes) {
+            return (Small32) CodePointTrie.fromBinary(Type.SMALL, ValueWidth.BITS_32, bytes);
+        }
+    }
+
+    /**
+     * A CodePointTrie with {@link Type#SMALL} and {@link ValueWidth#BITS_8}.
+     *
+     * @draft ICU 63
+     * @provisional This API might change or be removed in a future release.
+     */
+    public static final class Small8 extends Small {
+        Small8(char[] index, byte[] data8, int highStart,
+                int index3NullOffset, int dataNullOffset) {
+            super(index, new Data8(data8), highStart, index3NullOffset, dataNullOffset);
+        }
+
+        /**
+         * Creates a trie from its binary form.
+         * Same as {@link CodePointTrie#fromBinary(Type, ValueWidth, ByteBuffer)}
+         * with {@link Type#SMALL} and {@link ValueWidth#BITS_8}.
+         *
+         * @param bytes a buffer containing the binary data of a CodePointTrie
+         * @return the trie
+         * @draft ICU 63
+         * @provisional This API might change or be removed in a future release.
+         */
+        public static Small8 fromBinary(ByteBuffer bytes) {
+            return (Small8) CodePointTrie.fromBinary(Type.SMALL, ValueWidth.BITS_8, bytes);
+        }
+    }
+}
--- old/src/java.base/share/classes/sun/text/normalizer/OutputInt.java	2020-01-10 15:58:13.000000000 -0800
+++ /dev/null	2020-01-10 15:58:13.000000000 -0800
@@ -1,50 +0,0 @@
-/*
- * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.  Oracle designates this
- * particular file as subject to the "Classpath" exception as provided
- * by Oracle in the LICENSE file that accompanied this code.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- */
-
-/*
- *******************************************************************************
- * Copyright (C) 2014, International Business Machines Corporation and
- * others. All Rights Reserved.
- *******************************************************************************
- */
-package sun.text.normalizer;
-
-/**
- * Simple struct-like class for int output parameters.
- * Like Output<Integer> but without auto-boxing.
- *
- * @internal but could become public
- * deprecated This API is ICU internal only.
- */
-class OutputInt {
-
-    /**
-     * The value field.
-     *
-     * @internal
-     * deprecated This API is ICU internal only.
-     */
-    public int value;
-}
--- /dev/null	2020-01-10 15:58:13.000000000 -0800
+++ new/src/java.base/share/classes/jdk/internal/icu/util/OutputInt.java	2020-01-10 15:58:12.000000000 -0800
@@ -0,0 +1,50 @@
+/*
+ * Copyright (c) 2015, 2020, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.  Oracle designates this
+ * particular file as subject to the "Classpath" exception as provided
+ * by Oracle in the LICENSE file that accompanied this code.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+/*
+ *******************************************************************************
+ * Copyright (C) 2014, International Business Machines Corporation and
+ * others. All Rights Reserved.
+ *******************************************************************************
+ */
+package jdk.internal.icu.util;
+
+/**
+ * Simple struct-like class for int output parameters.
+ * Like Output<Integer> but without auto-boxing.
+ *
+ * @internal but could become public
+ * deprecated This API is ICU internal only.
+ */
+public class OutputInt {
+
+    /**
+     * The value field.
+     *
+     * @internal
+     * deprecated This API is ICU internal only.
+     */
+    public int value;
+}
--- old/src/java.base/share/classes/sun/text/normalizer/VersionInfo.java	2020-01-10 15:58:14.000000000 -0800
+++ /dev/null	2020-01-10 15:58:14.000000000 -0800
@@ -1,185 +0,0 @@
-/*
- * Copyright (c) 2005, 2011, Oracle and/or its affiliates. All rights reserved.
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.  Oracle designates this
- * particular file as subject to the "Classpath" exception as provided
- * by Oracle in the LICENSE file that accompanied this code.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- */
-/*
- *******************************************************************************
- * (C) Copyright IBM Corp. and others, 1996-2009 - All Rights Reserved         *
- *                                                                             *
- * The original version of this source code and documentation is copyrighted   *
- * and owned by IBM, These materials are provided under terms of a License     *
- * Agreement between IBM and Sun. This technology is protected by multiple     *
- * US and International patents. This notice and attribution to IBM may not    *
- * to removed.                                                                 *
- *******************************************************************************
- */
-
-package sun.text.normalizer;
-
-import java.util.HashMap;
-
-/**
- * Class to store version numbers of the form major.minor.milli.micro.
- * @author synwee
- * @stable ICU 2.6
- */
-public final class VersionInfo
-{
-
-    // public methods ------------------------------------------------------
-
-    /**
-     * Returns an instance of VersionInfo with the argument version.
-     * @param version version String in the format of "major.minor.milli.micro"
-     *                or "major.minor.milli" or "major.minor" or "major",
-     *                where major, minor, milli, micro are non-negative numbers
-     *                {@literal <=} 255. If the trailing version numbers are
-     *                not specified they are taken as 0s. E.g. Version "3.1" is
-     *                equivalent to "3.1.0.0".
-     * @return an instance of VersionInfo with the argument version.
-     * @exception throws an IllegalArgumentException when the argument version
-     *                is not in the right format
-     * @stable ICU 2.6
-     */
-    public static VersionInfo getInstance(String version)
-    {
-        int length  = version.length();
-        int array[] = {0, 0, 0, 0};
-        int count   = 0;
-        int index   = 0;
-
-        while (count < 4 && index < length) {
-            char c = version.charAt(index);
-            if (c == '.') {
-                count ++;
-            }
-            else {
-                c -= '0';
-                if (c < 0 || c > 9) {
-                    throw new IllegalArgumentException(INVALID_VERSION_NUMBER_);
-                }
-                array[count] *= 10;
-                array[count] += c;
-            }
-            index ++;
-        }
-        if (index != length) {
-            throw new IllegalArgumentException(
-                                               "Invalid version number: String '" + version + "' exceeds version format");
-        }
-        for (int i = 0; i < 4; i ++) {
-            if (array[i] < 0 || array[i] > 255) {
-                throw new IllegalArgumentException(INVALID_VERSION_NUMBER_);
-            }
-        }
-
-        return getInstance(array[0], array[1], array[2], array[3]);
-    }
-
-    /**
-     * Returns an instance of VersionInfo with the argument version.
-     * @param major major version, non-negative number {@literal <=} 255.
-     * @param minor minor version, non-negative number {@literal <=} 255.
-     * @param milli milli version, non-negative number {@literal <=} 255.
-     * @param micro micro version, non-negative number {@literal <=} 255.
-     * @exception throws an IllegalArgumentException when either arguments are
-     *                                     negative or {@literal >} 255
-     * @stable ICU 2.6
-     */
-    public static VersionInfo getInstance(int major, int minor, int milli,
-                                          int micro)
-    {
-        // checks if it is in the hashmap
-        // else
-        if (major < 0 || major > 255 || minor < 0 || minor > 255 ||
-            milli < 0 || milli > 255 || micro < 0 || micro > 255) {
-            throw new IllegalArgumentException(INVALID_VERSION_NUMBER_);
-        }
-        int     version = getInt(major, minor, milli, micro);
-        Integer key     = Integer.valueOf(version);
-        Object  result  = MAP_.get(key);
-        if (result == null) {
-            result = new VersionInfo(version);
-            MAP_.put(key, result);
-        }
-        return (VersionInfo)result;
-    }
-
-    /**
-     * Compares other with this VersionInfo.
-     * @param other VersionInfo to be compared
-     * @return 0 if the argument is a VersionInfo object that has version
-     *           information equals to this object.
-     *           Less than 0 if the argument is a VersionInfo object that has
-     *           version information greater than this object.
-     *           Greater than 0 if the argument is a VersionInfo object that
-     *           has version information less than this object.
-     * @stable ICU 2.6
-     */
-    public int compareTo(VersionInfo other)
-    {
-        return m_version_ - other.m_version_;
-    }
-
-    // private data members ----------------------------------------------
-
-    /**
-     * Version number stored as a byte for each of the major, minor, milli and
-     * micro numbers in the 32 bit int.
-     * Most significant for the major and the least significant contains the
-     * micro numbers.
-     */
-    private int m_version_;
-    /**
-     * Map of singletons
-     */
-    private static final HashMap MAP_ = new HashMap<>();
-    /**
-     * Error statement string
-     */
-    private static final String INVALID_VERSION_NUMBER_ =
-        "Invalid version number: Version number may be negative or greater than 255";
-
-    // private constructor -----------------------------------------------
-
-    /**
-     * Constructor with int
-     * @param compactversion a 32 bit int with each byte representing a number
-     */
-    private VersionInfo(int compactversion)
-    {
-        m_version_ = compactversion;
-    }
-
-    /**
-     * Gets the int from the version numbers
-     * @param major non-negative version number
-     * @param minor non-negativeversion number
-     * @param milli non-negativeversion number
-     * @param micro non-negativeversion number
-     */
-    private static int getInt(int major, int minor, int milli, int micro)
-    {
-        return (major << 24) | (minor << 16) | (milli << 8) | micro;
-    }
-}
--- /dev/null	2020-01-10 15:58:14.000000000 -0800
+++ new/src/java.base/share/classes/jdk/internal/icu/util/VersionInfo.java	2020-01-10 15:58:14.000000000 -0800
@@ -0,0 +1,195 @@
+/*
+ * Copyright (c) 2005, 2020, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.  Oracle designates this
+ * particular file as subject to the "Classpath" exception as provided
+ * by Oracle in the LICENSE file that accompanied this code.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+/*
+ *******************************************************************************
+ * (C) Copyright IBM Corp. and others, 1996-2009 - All Rights Reserved         *
+ *                                                                             *
+ * The original version of this source code and documentation is copyrighted   *
+ * and owned by IBM, These materials are provided under terms of a License     *
+ * Agreement between IBM and Sun. This technology is protected by multiple     *
+ * US and International patents. This notice and attribution to IBM may not    *
+ * to removed.                                                                 *
+ *******************************************************************************
+ */
+
+package jdk.internal.icu.util;
+
+import java.util.HashMap;
+
+/**
+ * Class to store version numbers of the form major.minor.milli.micro.
+ * @author synwee
+ * @stable ICU 2.6
+ */
+public final class VersionInfo
+{
+    // public data members -------------------------------------------------
+
+    /**
+     * Data version string for ICU's internal data.
+     * Used for appending to data path (e.g. icudt43b)
+     * @internal
+     * @deprecated This API is ICU internal only.
+     */
+    @Deprecated
+    public static final String ICU_DATA_VERSION_PATH = "64b";
+
+    // public methods ------------------------------------------------------
+
+    /**
+     * Returns an instance of VersionInfo with the argument version.
+     * @param version version String in the format of "major.minor.milli.micro"
+     *                or "major.minor.milli" or "major.minor" or "major",
+     *                where major, minor, milli, micro are non-negative numbers
+     *                {@literal <=} 255. If the trailing version numbers are
+     *                not specified they are taken as 0s. E.g. Version "3.1" is
+     *                equivalent to "3.1.0.0".
+     * @return an instance of VersionInfo with the argument version.
+     * @exception throws an IllegalArgumentException when the argument version
+     *                is not in the right format
+     * @stable ICU 2.6
+     */
+    public static VersionInfo getInstance(String version)
+    {
+        int length  = version.length();
+        int array[] = {0, 0, 0, 0};
+        int count   = 0;
+        int index   = 0;
+
+        while (count < 4 && index < length) {
+            char c = version.charAt(index);
+            if (c == '.') {
+                count ++;
+            }
+            else {
+                c -= '0';
+                if (c < 0 || c > 9) {
+                    throw new IllegalArgumentException(INVALID_VERSION_NUMBER_);
+                }
+                array[count] *= 10;
+                array[count] += c;
+            }
+            index ++;
+        }
+        if (index != length) {
+            throw new IllegalArgumentException(
+                                               "Invalid version number: String '" + version + "' exceeds version format");
+        }
+        for (int i = 0; i < 4; i ++) {
+            if (array[i] < 0 || array[i] > 255) {
+                throw new IllegalArgumentException(INVALID_VERSION_NUMBER_);
+            }
+        }
+
+        return getInstance(array[0], array[1], array[2], array[3]);
+    }
+
+    /**
+     * Returns an instance of VersionInfo with the argument version.
+     * @param major major version, non-negative number {@literal <=} 255.
+     * @param minor minor version, non-negative number {@literal <=} 255.
+     * @param milli milli version, non-negative number {@literal <=} 255.
+     * @param micro micro version, non-negative number {@literal <=} 255.
+     * @exception throws an IllegalArgumentException when either arguments are
+     *                                     negative or {@literal >} 255
+     * @stable ICU 2.6
+     */
+    public static VersionInfo getInstance(int major, int minor, int milli,
+                                          int micro)
+    {
+        // checks if it is in the hashmap
+        // else
+        if (major < 0 || major > 255 || minor < 0 || minor > 255 ||
+            milli < 0 || milli > 255 || micro < 0 || micro > 255) {
+            throw new IllegalArgumentException(INVALID_VERSION_NUMBER_);
+        }
+        int     version = getInt(major, minor, milli, micro);
+        Integer key     = Integer.valueOf(version);
+        Object  result  = MAP_.get(key);
+        if (result == null) {
+            result = new VersionInfo(version);
+            MAP_.put(key, result);
+        }
+        return (VersionInfo)result;
+    }
+
+    /**
+     * Compares other with this VersionInfo.
+     * @param other VersionInfo to be compared
+     * @return 0 if the argument is a VersionInfo object that has version
+     *           information equals to this object.
+     *           Less than 0 if the argument is a VersionInfo object that has
+     *           version information greater than this object.
+     *           Greater than 0 if the argument is a VersionInfo object that
+     *           has version information less than this object.
+     * @stable ICU 2.6
+     */
+    public int compareTo(VersionInfo other)
+    {
+        return m_version_ - other.m_version_;
+    }
+
+    // private data members ----------------------------------------------
+
+    /**
+     * Version number stored as a byte for each of the major, minor, milli and
+     * micro numbers in the 32 bit int.
+     * Most significant for the major and the least significant contains the
+     * micro numbers.
+     */
+    private int m_version_;
+    /**
+     * Map of singletons
+     */
+    private static final HashMap MAP_ = new HashMap<>();
+    /**
+     * Error statement string
+     */
+    private static final String INVALID_VERSION_NUMBER_ =
+        "Invalid version number: Version number may be negative or greater than 255";
+
+    // private constructor -----------------------------------------------
+
+    /**
+     * Constructor with int
+     * @param compactversion a 32 bit int with each byte representing a number
+     */
+    private VersionInfo(int compactversion)
+    {
+        m_version_ = compactversion;
+    }
+
+    /**
+     * Gets the int from the version numbers
+     * @param major non-negative version number
+     * @param minor non-negativeversion number
+     * @param milli non-negativeversion number
+     * @param micro non-negativeversion number
+     */
+    private static int getInt(int major, int minor, int milli, int micro)
+    {
+        return (major << 24) | (minor << 16) | (milli << 8) | micro;
+    }
+}