1 /*
   2  * Copyright (c) 2003, 2013, Oracle and/or its affiliates. All rights reserved.
   3  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   4  *
   5  * This code is free software; you can redistribute it and/or modify it
   6  * under the terms of the GNU General Public License version 2 only, as
   7  * published by the Free Software Foundation.  Oracle designates this
   8  * particular file as subject to the "Classpath" exception as provided
   9  * by Oracle in the LICENSE file that accompanied this code.
  10  *
  11  * This code is distributed in the hope that it will be useful, but WITHOUT
  12  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  13  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
  14  * version 2 for more details (a copy is included in the LICENSE file that
  15  * accompanied this code).
  16  *
  17  * You should have received a copy of the GNU General Public License version
  18  * 2 along with this work; if not, write to the Free Software Foundation,
  19  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
  20  *
  21  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
  22  * or visit www.oracle.com if you need additional information or have any
  23  * questions.
  24  */
  25 
  26 package sun.font;
  27 
  28 import java.nio.ByteBuffer;
  29 import java.nio.CharBuffer;
  30 import java.nio.IntBuffer;
  31 import java.util.Locale;
  32 import java.nio.charset.*;
  33 
  34 /*
  35  * A tt font has a CMAP table which is in turn made up of sub-tables which
  36  * describe the char to glyph mapping in (possibly) multiple ways.
  37  * CMAP subtables are described by 3 values.
  38  * 1. Platform ID (eg 3=Microsoft, which is the id we look for in JDK)
  39  * 2. Encoding (eg 0=symbol, 1=unicode)
  40  * 3. TrueType subtable format (how the char->glyph mapping for the encoding
  41  * is stored in the subtable). See the TrueType spec. Format 4 is required
  42  * by MS in fonts for windows. Its uses segmented mapping to delta values.
  43  * Most typically we see are (3,1,4) :
  44  * CMAP Platform ID=3 is what we use.
  45  * Encodings that are used in practice by JDK on Solaris are
  46  *  symbol (3,0)
  47  *  unicode (3,1)
  48  *  GBK (3,5) (note that solaris zh fonts report 3,4 but are really 3,5)
  49  * The format for almost all subtables is 4. However the solaris (3,5)
  50  * encodings are typically in format 2.
  51  */
  52 abstract class CMap {
  53 
  54 //     static char WingDings_b2c[] = {
  55 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
  56 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
  57 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
  58 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
  59 //         0xfffd, 0xfffd, 0x2702, 0x2701, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
  60 //         0xfffd, 0x2706, 0x2709, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
  61 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
  62 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0x2707, 0x270d,
  63 //         0xfffd, 0x270c, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
  64 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
  65 //         0xfffd, 0x2708, 0xfffd, 0xfffd, 0x2744, 0xfffd, 0x271e, 0xfffd,
  66 //         0x2720, 0x2721, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
  67 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
  68 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
  69 //         0xfffd, 0x2751, 0x2752, 0xfffd, 0xfffd, 0x2756, 0xfffd, 0xfffd,
  70 //         0xfffd, 0xfffd, 0xfffd, 0x2740, 0x273f, 0x275d, 0x275e, 0xfffd,
  71 //         0xfffd, 0x2780, 0x2781, 0x2782, 0x2783, 0x2784, 0x2785, 0x2786,
  72 //         0x2787, 0x2788, 0x2789, 0xfffd, 0x278a, 0x278b, 0x278c, 0x278d,
  73 //         0x278e, 0x278f, 0x2790, 0x2791, 0x2792, 0x2793, 0xfffd, 0xfffd,
  74 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
  75 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0x274d, 0xfffd,
  76 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0x2736, 0x2734, 0xfffd, 0x2735,
  77 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0x272a, 0x2730, 0xfffd,
  78 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
  79 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0x27a5, 0xfffd, 0x27a6, 0xfffd,
  80 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
  81 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
  82 //         0x27a2, 0xfffd, 0xfffd, 0xfffd, 0x27b3, 0xfffd, 0xfffd, 0xfffd,
  83 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
  84 //         0x27a1, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
  85 //         0x27a9, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
  86 //         0xfffd, 0xfffd, 0xfffd, 0x2717, 0x2713, 0xfffd, 0xfffd, 0xfffd,
  87 //    };
  88 
  89 //     static char Symbols_b2c[] = {
  90 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
  91 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
  92 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
  93 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
  94 //         0xfffd, 0xfffd, 0x2200, 0xfffd, 0x2203, 0xfffd, 0xfffd, 0x220d,
  95 //         0xfffd, 0xfffd, 0x2217, 0xfffd, 0xfffd, 0x2212, 0xfffd, 0xfffd,
  96 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
  97 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
  98 //         0x2245, 0x0391, 0x0392, 0x03a7, 0x0394, 0x0395, 0x03a6, 0x0393,
  99 //         0x0397, 0x0399, 0x03d1, 0x039a, 0x039b, 0x039c, 0x039d, 0x039f,
 100 //         0x03a0, 0x0398, 0x03a1, 0x03a3, 0x03a4, 0x03a5, 0x03c2, 0x03a9,
 101 //         0x039e, 0x03a8, 0x0396, 0xfffd, 0x2234, 0xfffd, 0x22a5, 0xfffd,
 102 //         0xfffd, 0x03b1, 0x03b2, 0x03c7, 0x03b4, 0x03b5, 0x03c6, 0x03b3,
 103 //         0x03b7, 0x03b9, 0x03d5, 0x03ba, 0x03bb, 0x03bc, 0x03bd, 0x03bf,
 104 //         0x03c0, 0x03b8, 0x03c1, 0x03c3, 0x03c4, 0x03c5, 0x03d6, 0x03c9,
 105 //         0x03be, 0x03c8, 0x03b6, 0xfffd, 0xfffd, 0xfffd, 0x223c, 0xfffd,
 106 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
 107 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
 108 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
 109 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
 110 //         0xfffd, 0x03d2, 0xfffd, 0x2264, 0x2215, 0x221e, 0xfffd, 0xfffd,
 111 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
 112 //         0x2218, 0xfffd, 0xfffd, 0x2265, 0xfffd, 0x221d, 0xfffd, 0x2219,
 113 //         0xfffd, 0x2260, 0x2261, 0x2248, 0x22ef, 0x2223, 0xfffd, 0xfffd,
 114 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0x2297, 0x2295, 0x2205, 0x2229,
 115 //         0x222a, 0x2283, 0x2287, 0x2284, 0x2282, 0x2286, 0x2208, 0x2209,
 116 //         0xfffd, 0x2207, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0x221a, 0x22c5,
 117 //         0xfffd, 0x2227, 0x2228, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
 118 //         0x22c4, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0x2211, 0xfffd, 0xfffd,
 119 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
 120 //         0xfffd, 0xfffd, 0x222b, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
 121 //         0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd,
 122 //     };
 123 
 124     static final short ShiftJISEncoding = 2;
 125     static final short GBKEncoding      = 3;
 126     static final short Big5Encoding     = 4;
 127     static final short WansungEncoding  = 5;
 128     static final short JohabEncoding    = 6;
 129     static final short MSUnicodeSurrogateEncoding = 10;
 130 
 131     static final char noSuchChar = (char)0xfffd;
 132     static final int SHORTMASK = 0x0000ffff;
 133     static final int INTMASK   = 0x7fffffff;
 134 
 135     static final char[][] converterMaps = new char[7][];
 136 
 137     /*
 138      * Unicode->other encoding translation array. A pre-computed look up
 139      * which can be shared across all fonts using that encoding.
 140      * Using this saves running character coverters repeatedly.
 141      */
 142     char[] xlat;
 143     UVS uvs = null;
 144 
 145     static CMap initialize(TrueTypeFont font) {
 146 
 147         CMap cmap = null;
 148 
 149         int offset, platformID, encodingID=-1;
 150 
 151         int three0=0, three1=0, three2=0, three3=0, three4=0, three5=0,
 152             three6=0, three10=0;
 153         int zero5 = 0; // for Unicode Variation Sequences
 154         boolean threeStar = false;
 155 
 156         ByteBuffer cmapBuffer = font.getTableBuffer(TrueTypeFont.cmapTag);
 157         int cmapTableOffset = font.getTableSize(TrueTypeFont.cmapTag);
 158         short numberSubTables = cmapBuffer.getShort(2);
 159 
 160         /* locate the offsets of all 3,*  (ie Microsoft platform) encodings */
 161         for (int i=0; i<numberSubTables; i++) {
 162             cmapBuffer.position(i * 8 + 4);
 163             platformID = cmapBuffer.getShort();
 164             if (platformID == 3) {
 165                 threeStar = true;
 166                 encodingID = cmapBuffer.getShort();
 167                 offset     = cmapBuffer.getInt();
 168                 switch (encodingID) {
 169                 case 0:  three0  = offset; break; // MS Symbol encoding
 170                 case 1:  three1  = offset; break; // MS Unicode cmap
 171                 case 2:  three2  = offset; break; // ShiftJIS cmap.
 172                 case 3:  three3  = offset; break; // GBK cmap
 173                 case 4:  three4  = offset; break; // Big 5 cmap
 174                 case 5:  three5  = offset; break; // Wansung
 175                 case 6:  three6  = offset; break; // Johab
 176                 case 10: three10 = offset; break; // MS Unicode surrogates
 177                 }
 178             } else if (platformID == 0) {
 179                 encodingID = cmapBuffer.getShort();
 180                 offset     = cmapBuffer.getInt();
 181                 if (encodingID == 5) {
 182                     zero5 = offset;
 183                 }
 184             }
 185         }
 186 
 187         /* This defines the preference order for cmap subtables */
 188         if (threeStar) {
 189             if (three10 != 0) {
 190                 cmap = createCMap(cmapBuffer, three10, null);
 191             }
 192             else if  (three0 != 0) {
 193                 /* The special case treatment of these fonts leads to
 194                  * anomalies where a user can view "wingdings" and "wingdings2"
 195                  * and the latter shows all its code points in the unicode
 196                  * private use area at 0xF000->0XF0FF and the former shows
 197                  * a scattered subset of its glyphs that are known mappings to
 198                  * unicode code points.
 199                  * The primary purpose of these mappings was to facilitate
 200                  * display of symbol chars etc in composite fonts, however
 201                  * this is not needed as all these code points are covered
 202                  * by some other platform symbol font.
 203                  * Commenting this out reduces the role of these two files
 204                  * (assuming that they continue to be used in font.properties)
 205                  * to just one of contributing to the overall composite
 206                  * font metrics, and also AWT can still access the fonts.
 207                  * Clients which explicitly accessed these fonts as names
 208                  * "Symbol" and "Wingdings" (ie as physical fonts) and
 209                  * expected to see a scattering of these characters will
 210                  * see them now as missing. How much of a problem is this?
 211                  * Perhaps we could still support this mapping just for
 212                  * "Symbol.ttf" but I suspect some users would prefer it
 213                  * to be mapped in to the Latin range as that is how
 214                  * the "symbol" font is used in native apps.
 215                  */
 216 //              String name = font.platName.toLowerCase(Locale.ENGLISH);
 217 //              if (name.endsWith("symbol.ttf")) {
 218 //                  cmap = createSymbolCMap(cmapBuffer, three0, Symbols_b2c);
 219 //              } else if (name.endsWith("wingding.ttf")) {
 220 //                  cmap = createSymbolCMap(cmapBuffer, three0, WingDings_b2c);
 221 //              } else {
 222                     cmap = createCMap(cmapBuffer, three0, null);
 223 //              }
 224             }
 225             else if (three1 != 0) {
 226                 cmap = createCMap(cmapBuffer, three1, null);
 227             }
 228             else if (three2 != 0) {
 229                 cmap = createCMap(cmapBuffer, three2,
 230                                   getConverterMap(ShiftJISEncoding));
 231             }
 232             else if (three3 != 0) {
 233                 cmap = createCMap(cmapBuffer, three3,
 234                                   getConverterMap(GBKEncoding));
 235             }
 236             else if (three4 != 0) {
 237                 /* GB2312 TrueType fonts on Solaris have wrong encoding ID for
 238                  * cmap table, these fonts have EncodingID 4 which is Big5
 239                  * encoding according the TrueType spec, but actually the
 240                  * fonts are using gb2312 encoding, have to use this
 241                  * workaround to make Solaris zh_CN locale work.  -sherman
 242                  */
 243                 if (FontUtilities.isSolaris && font.platName != null &&
 244                     (font.platName.startsWith(
 245                      "/usr/openwin/lib/locale/zh_CN.EUC/X11/fonts/TrueType") ||
 246                      font.platName.startsWith(
 247                      "/usr/openwin/lib/locale/zh_CN/X11/fonts/TrueType") ||
 248                      font.platName.startsWith(
 249                      "/usr/openwin/lib/locale/zh/X11/fonts/TrueType"))) {
 250                     cmap = createCMap(cmapBuffer, three4,
 251                                        getConverterMap(GBKEncoding));
 252                 }
 253                 else {
 254                     cmap = createCMap(cmapBuffer, three4,
 255                                       getConverterMap(Big5Encoding));
 256                 }
 257             }
 258             else if (three5 != 0) {
 259                 cmap = createCMap(cmapBuffer, three5,
 260                                   getConverterMap(WansungEncoding));
 261             }
 262             else if (three6 != 0) {
 263                 cmap = createCMap(cmapBuffer, three6,
 264                                   getConverterMap(JohabEncoding));
 265             }
 266         } else {
 267             /* No 3,* subtable was found. Just use whatever is the first
 268              * table listed. Not very useful but maybe better than
 269              * rejecting the font entirely?
 270              */
 271             cmap = createCMap(cmapBuffer, cmapBuffer.getInt(8), null);
 272         }
 273         // For Unicode Variation Sequences
 274         if (cmap != null && zero5 != 0) {
 275             cmap.createUVS(cmapBuffer, zero5);
 276         }
 277         return cmap;
 278     }
 279 
 280     /* speed up the converting by setting the range for double
 281      * byte characters;
 282      */
 283     static char[] getConverter(short encodingID) {
 284         int dBegin = 0x8000;
 285         int dEnd   = 0xffff;
 286         String encoding;
 287 
 288         switch (encodingID) {
 289         case ShiftJISEncoding:
 290             dBegin = 0x8140;
 291             dEnd   = 0xfcfc;
 292             encoding = "SJIS";
 293             break;
 294         case GBKEncoding:
 295             dBegin = 0x8140;
 296             dEnd   = 0xfea0;
 297             encoding = "GBK";
 298             break;
 299         case Big5Encoding:
 300             dBegin = 0xa140;
 301             dEnd   = 0xfefe;
 302             encoding = "Big5";
 303             break;
 304         case WansungEncoding:
 305             dBegin = 0xa1a1;
 306             dEnd   = 0xfede;
 307             encoding = "EUC_KR";
 308             break;
 309         case JohabEncoding:
 310             dBegin = 0x8141;
 311             dEnd   = 0xfdfe;
 312             encoding = "Johab";
 313             break;
 314         default:
 315             return null;
 316         }
 317 
 318         try {
 319             char[] convertedChars = new char[65536];
 320             for (int i=0; i<65536; i++) {
 321                 convertedChars[i] = noSuchChar;
 322             }
 323 
 324             byte[] inputBytes = new byte[(dEnd-dBegin+1)*2];
 325             char[] outputChars = new char[(dEnd-dBegin+1)];
 326 
 327             int j = 0;
 328             int firstByte;
 329             if (encodingID == ShiftJISEncoding) {
 330                 for (int i = dBegin; i <= dEnd; i++) {
 331                     firstByte = (i >> 8 & 0xff);
 332                     if (firstByte >= 0xa1 && firstByte <= 0xdf) {
 333                         //sjis halfwidth katakana
 334                         inputBytes[j++] = (byte)0xff;
 335                         inputBytes[j++] = (byte)0xff;
 336                     } else {
 337                         inputBytes[j++] = (byte)firstByte;
 338                         inputBytes[j++] = (byte)(i & 0xff);
 339                     }
 340                 }
 341             } else {
 342                 for (int i = dBegin; i <= dEnd; i++) {
 343                     inputBytes[j++] = (byte)(i>>8 & 0xff);
 344                     inputBytes[j++] = (byte)(i & 0xff);
 345                 }
 346             }
 347 
 348             Charset.forName(encoding).newDecoder()
 349             .onMalformedInput(CodingErrorAction.REPLACE)
 350             .onUnmappableCharacter(CodingErrorAction.REPLACE)
 351             .replaceWith("\u0000")
 352             .decode(ByteBuffer.wrap(inputBytes, 0, inputBytes.length),
 353                     CharBuffer.wrap(outputChars, 0, outputChars.length),
 354                     true);
 355 
 356             // ensure single byte ascii
 357             for (int i = 0x20; i <= 0x7e; i++) {
 358                 convertedChars[i] = (char)i;
 359             }
 360 
 361             //sjis halfwidth katakana
 362             if (encodingID == ShiftJISEncoding) {
 363                 for (int i = 0xa1; i <= 0xdf; i++) {
 364                     convertedChars[i] = (char)(i - 0xa1 + 0xff61);
 365                 }
 366             }
 367 
 368             /* It would save heap space (approx 60Kbytes for each of these
 369              * converters) if stored only valid ranges (ie returned
 370              * outputChars directly. But this is tricky since want to
 371              * include the ASCII range too.
 372              */
 373 //          System.err.println("oc.len="+outputChars.length);
 374 //          System.err.println("cc.len="+convertedChars.length);
 375 //          System.err.println("dbegin="+dBegin);
 376             System.arraycopy(outputChars, 0, convertedChars, dBegin,
 377                              outputChars.length);
 378 
 379             //return convertedChars;
 380             /* invert this map as now want it to map from Unicode
 381              * to other encoding.
 382              */
 383             char [] invertedChars = new char[65536];
 384             for (int i=0;i<65536;i++) {
 385                 if (convertedChars[i] != noSuchChar) {
 386                     invertedChars[convertedChars[i]] = (char)i;
 387                 }
 388             }
 389             return invertedChars;
 390 
 391         } catch (Exception e) {
 392             e.printStackTrace();
 393         }
 394         return null;
 395     }
 396 
 397     /*
 398      * The returned array maps to unicode from some other 2 byte encoding
 399      * eg for a 2byte index which represents a SJIS char, the indexed
 400      * value is the corresponding unicode char.
 401      */
 402     static char[] getConverterMap(short encodingID) {
 403         if (converterMaps[encodingID] == null) {
 404            converterMaps[encodingID] = getConverter(encodingID);
 405         }
 406         return converterMaps[encodingID];
 407     }
 408 
 409 
 410     static CMap createCMap(ByteBuffer buffer, int offset, char[] xlat) {
 411         /* First do a sanity check that this cmap subtable is contained
 412          * within the cmap table.
 413          */
 414         int subtableFormat = buffer.getChar(offset);
 415         long subtableLength;
 416         if (subtableFormat < 8) {
 417             subtableLength = buffer.getChar(offset+2);
 418         } else {
 419             subtableLength = buffer.getInt(offset+4) & INTMASK;
 420         }
 421         if (offset+subtableLength > buffer.capacity()) {
 422             if (FontUtilities.isLogging()) {
 423                 FontUtilities.getLogger().warning("Cmap subtable overflows buffer.");
 424             }
 425         }
 426         switch (subtableFormat) {
 427         case 0:  return new CMapFormat0(buffer, offset);
 428         case 2:  return new CMapFormat2(buffer, offset, xlat);
 429         case 4:  return new CMapFormat4(buffer, offset, xlat);
 430         case 6:  return new CMapFormat6(buffer, offset, xlat);
 431         case 8:  return new CMapFormat8(buffer, offset, xlat);
 432         case 10: return new CMapFormat10(buffer, offset, xlat);
 433         case 12: return new CMapFormat12(buffer, offset, xlat);
 434         default: throw new RuntimeException("Cmap format unimplemented: " +
 435                                             (int)buffer.getChar(offset));
 436         }
 437     }
 438 
 439     private void createUVS(ByteBuffer buffer, int offset) {
 440         int subtableFormat = buffer.getChar(offset);
 441         if (subtableFormat == 14) {
 442             long subtableLength = buffer.getInt(offset + 2) & INTMASK;
 443             if (offset + subtableLength > buffer.capacity()) {
 444                 if (FontUtilities.isLogging()) {
 445                     FontUtilities.getLogger()
 446                             .warning("Cmap UVS subtable overflows buffer.");
 447                 }
 448             }
 449             try {
 450                 this.uvs = new UVS(buffer, offset);
 451             } catch (Throwable t) {
 452                 t.printStackTrace();
 453             }
 454         }
 455         return;
 456     }
 457 
 458 /*
 459     final char charVal(byte[] cmap, int index) {
 460         return (char)(((0xff & cmap[index]) << 8)+(0xff & cmap[index+1]));
 461     }
 462 
 463     final short shortVal(byte[] cmap, int index) {
 464         return (short)(((0xff & cmap[index]) << 8)+(0xff & cmap[index+1]));
 465     }
 466 */
 467     abstract char getGlyph(int charCode);
 468 
 469     /* Format 4 Header is
 470      * ushort format (off=0)
 471      * ushort length (off=2)
 472      * ushort language (off=4)
 473      * ushort segCountX2 (off=6)
 474      * ushort searchRange (off=8)
 475      * ushort entrySelector (off=10)
 476      * ushort rangeShift (off=12)
 477      * ushort endCount[segCount] (off=14)
 478      * ushort reservedPad
 479      * ushort startCount[segCount]
 480      * short idDelta[segCount]
 481      * idRangeOFfset[segCount]
 482      * ushort glyphIdArray[]
 483      */
 484     static class CMapFormat4 extends CMap {
 485         int segCount;
 486         int entrySelector;
 487         int rangeShift;
 488         char[] endCount;
 489         char[] startCount;
 490         short[] idDelta;
 491         char[] idRangeOffset;
 492         char[] glyphIds;
 493 
 494         CMapFormat4(ByteBuffer bbuffer, int offset, char[] xlat) {
 495 
 496             this.xlat = xlat;
 497 
 498             bbuffer.position(offset);
 499             CharBuffer buffer = bbuffer.asCharBuffer();
 500             buffer.get(); // skip, we already know format=4
 501             int subtableLength = buffer.get();
 502             /* Try to recover from some bad fonts which specify a subtable
 503              * length that would overflow the byte buffer holding the whole
 504              * cmap table. If this isn't a recoverable situation an exception
 505              * may be thrown which is caught higher up the call stack.
 506              * Whilst this may seem lenient, in practice, unless the "bad"
 507              * subtable we are using is the last one in the cmap table we
 508              * would have no way of knowing about this problem anyway.
 509              */
 510             if (offset+subtableLength > bbuffer.capacity()) {
 511                 subtableLength = bbuffer.capacity() - offset;
 512             }
 513             buffer.get(); // skip language
 514             segCount = buffer.get()/2;
 515             int searchRange = buffer.get();
 516             entrySelector = buffer.get();
 517             rangeShift    = buffer.get()/2;
 518             startCount = new char[segCount];
 519             endCount = new char[segCount];
 520             idDelta = new short[segCount];
 521             idRangeOffset = new char[segCount];
 522 
 523             for (int i=0; i<segCount; i++) {
 524                 endCount[i] = buffer.get();
 525             }
 526             buffer.get(); // 2 bytes for reserved pad
 527             for (int i=0; i<segCount; i++) {
 528                 startCount[i] = buffer.get();
 529             }
 530 
 531             for (int i=0; i<segCount; i++) {
 532                 idDelta[i] = (short)buffer.get();
 533             }
 534 
 535             for (int i=0; i<segCount; i++) {
 536                 char ctmp = buffer.get();
 537                 idRangeOffset[i] = (char)((ctmp>>1)&0xffff);
 538             }
 539             /* Can calculate the number of glyph IDs by subtracting
 540              * "pos" from the length of the cmap
 541              */
 542             int pos = (segCount*8+16)/2;
 543             buffer.position(pos);
 544             int numGlyphIds = (subtableLength/2 - pos);
 545             glyphIds = new char[numGlyphIds];
 546             for (int i=0;i<numGlyphIds;i++) {
 547                 glyphIds[i] = buffer.get();
 548             }
 549 /*
 550             System.err.println("segcount="+segCount);
 551             System.err.println("entrySelector="+entrySelector);
 552             System.err.println("rangeShift="+rangeShift);
 553             for (int j=0;j<segCount;j++) {
 554               System.err.println("j="+j+ " sc="+(int)(startCount[j]&0xffff)+
 555                                  " ec="+(int)(endCount[j]&0xffff)+
 556                                  " delta="+idDelta[j] +
 557                                  " ro="+(int)idRangeOffset[j]);
 558             }
 559 
 560             //System.err.println("numglyphs="+glyphIds.length);
 561             for (int i=0;i<numGlyphIds;i++) {
 562                   System.err.println("gid["+i+"]="+(int)glyphIds[i]);
 563             }
 564 */
 565         }
 566 
 567         char getGlyph(int charCode) {
 568 
 569             final int origCharCode = charCode;
 570             int index = 0;
 571             char glyphCode = 0;
 572 
 573             int controlGlyph = getControlCodeGlyph(charCode, true);
 574             if (controlGlyph >= 0) {
 575                 return (char)controlGlyph;
 576             }
 577 
 578             /* presence of translation array indicates that this
 579              * cmap is in some other (non-unicode encoding).
 580              * In order to look-up a char->glyph mapping we need to
 581              * translate the unicode code point to the encoding of
 582              * the cmap.
 583              * REMIND: VALID CHARCODES??
 584              */
 585             if (xlat != null) {
 586                 charCode = xlat[charCode];
 587             }
 588 
 589             /*
 590              * Citation from the TrueType (and OpenType) spec:
 591              *   The segments are sorted in order of increasing endCode
 592              *   values, and the segment values are specified in four parallel
 593              *   arrays. You search for the first endCode that is greater than
 594              *   or equal to the character code you want to map. If the
 595              *   corresponding startCode is less than or equal to the
 596              *   character code, then you use the corresponding idDelta and
 597              *   idRangeOffset to map the character code to a glyph index
 598              *   (otherwise, the missingGlyph is returned).
 599              */
 600 
 601             /*
 602              * CMAP format4 defines several fields for optimized search of
 603              * the segment list (entrySelector, searchRange, rangeShift).
 604              * However, benefits are neglible and some fonts have incorrect
 605              * data - so we use straightforward binary search (see bug 6247425)
 606              */
 607             int left = 0, right = startCount.length;
 608             index = startCount.length >> 1;
 609             while (left < right) {
 610                 if (endCount[index] < charCode) {
 611                     left = index + 1;
 612                 } else {
 613                     right = index;
 614                 }
 615                 index = (left + right) >> 1;
 616             }
 617 
 618             if (charCode >= startCount[index] && charCode <= endCount[index]) {
 619                 int rangeOffset = idRangeOffset[index];
 620 
 621                 if (rangeOffset == 0) {
 622                     glyphCode = (char)(charCode + idDelta[index]);
 623                 } else {
 624                     /* Calculate an index into the glyphIds array */
 625 
 626 /*
 627                     System.err.println("rangeoffset="+rangeOffset+
 628                                        " charCode=" + charCode +
 629                                        " scnt["+index+"]="+(int)startCount[index] +
 630                                        " segCnt="+segCount);
 631 */
 632 
 633                     int glyphIDIndex = rangeOffset - segCount + index
 634                                          + (charCode - startCount[index]);
 635                     glyphCode = glyphIds[glyphIDIndex];
 636                     if (glyphCode != 0) {
 637                         glyphCode = (char)(glyphCode + idDelta[index]);
 638                     }
 639                 }
 640             }
 641             if (glyphCode == 0) {
 642               glyphCode = getFormatCharGlyph(origCharCode);
 643             }
 644             return glyphCode;
 645         }
 646     }
 647 
 648     // Format 0: Byte Encoding table
 649     static class CMapFormat0 extends CMap {
 650         byte [] cmap;
 651 
 652         CMapFormat0(ByteBuffer buffer, int offset) {
 653 
 654             /* skip 6 bytes of format, length, and version */
 655             int len = buffer.getChar(offset+2);
 656             cmap = new byte[len-6];
 657             buffer.position(offset+6);
 658             buffer.get(cmap);
 659         }
 660 
 661         char getGlyph(int charCode) {
 662             if (charCode < 256) {
 663                 if (charCode < 0x0010) {
 664                     switch (charCode) {
 665                     case 0x0009:
 666                     case 0x000a:
 667                     case 0x000d: return CharToGlyphMapper.INVISIBLE_GLYPH_ID;
 668                     }
 669                 }
 670                 return (char)(0xff & cmap[charCode]);
 671             } else {
 672                 return 0;
 673             }
 674         }
 675     }
 676 
 677 //     static CMap createSymbolCMap(ByteBuffer buffer, int offset, char[] syms) {
 678 
 679 //      CMap cmap = createCMap(buffer, offset, null);
 680 //      if (cmap == null) {
 681 //          return null;
 682 //      } else {
 683 //          return new CMapFormatSymbol(cmap, syms);
 684 //      }
 685 //     }
 686 
 687 //     static class CMapFormatSymbol extends CMap {
 688 
 689 //      CMap cmap;
 690 //      static final int NUM_BUCKETS = 128;
 691 //      Bucket[] buckets = new Bucket[NUM_BUCKETS];
 692 
 693 //      class Bucket {
 694 //          char unicode;
 695 //          char glyph;
 696 //          Bucket next;
 697 
 698 //          Bucket(char u, char g) {
 699 //              unicode = u;
 700 //              glyph = g;
 701 //          }
 702 //      }
 703 
 704 //      CMapFormatSymbol(CMap cmap, char[] syms) {
 705 
 706 //          this.cmap = cmap;
 707 
 708 //          for (int i=0;i<syms.length;i++) {
 709 //              char unicode = syms[i];
 710 //              if (unicode != noSuchChar) {
 711 //                  char glyph = cmap.getGlyph(i + 0xf000);
 712 //                  int hash = unicode % NUM_BUCKETS;
 713 //                  Bucket bucket = new Bucket(unicode, glyph);
 714 //                  if (buckets[hash] == null) {
 715 //                      buckets[hash] = bucket;
 716 //                  } else {
 717 //                      Bucket b = buckets[hash];
 718 //                      while (b.next != null) {
 719 //                          b = b.next;
 720 //                      }
 721 //                      b.next = bucket;
 722 //                  }
 723 //              }
 724 //          }
 725 //      }
 726 
 727 //      char getGlyph(int unicode) {
 728 //          if (unicode >= 0x1000) {
 729 //              return 0;
 730 //          }
 731 //          else if (unicode >=0xf000 && unicode < 0xf100) {
 732 //              return cmap.getGlyph(unicode);
 733 //          } else {
 734 //              Bucket b = buckets[unicode % NUM_BUCKETS];
 735 //              while (b != null) {
 736 //                  if (b.unicode == unicode) {
 737 //                      return b.glyph;
 738 //                  } else {
 739 //                      b = b.next;
 740 //                  }
 741 //              }
 742 //              return 0;
 743 //          }
 744 //      }
 745 //     }
 746 
 747     // Format 2: High-byte mapping through table
 748     static class CMapFormat2 extends CMap {
 749 
 750         char[] subHeaderKey = new char[256];
 751          /* Store subheaders in individual arrays
 752           * A SubHeader entry theortically looks like {
 753           *   char firstCode;
 754           *   char entryCount;
 755           *   short idDelta;
 756           *   char idRangeOffset;
 757           * }
 758           */
 759         char[] firstCodeArray;
 760         char[] entryCountArray;
 761         short[] idDeltaArray;
 762         char[] idRangeOffSetArray;
 763 
 764         char[] glyphIndexArray;
 765 
 766         CMapFormat2(ByteBuffer buffer, int offset, char[] xlat) {
 767 
 768             this.xlat = xlat;
 769 
 770             int tableLen = buffer.getChar(offset+2);
 771             buffer.position(offset+6);
 772             CharBuffer cBuffer = buffer.asCharBuffer();
 773             char maxSubHeader = 0;
 774             for (int i=0;i<256;i++) {
 775                 subHeaderKey[i] = cBuffer.get();
 776                 if (subHeaderKey[i] > maxSubHeader) {
 777                     maxSubHeader = subHeaderKey[i];
 778                 }
 779             }
 780             /* The value of the subHeaderKey is 8 * the subHeader index,
 781              * so the number of subHeaders can be obtained by dividing
 782              * this value bv 8 and adding 1.
 783              */
 784             int numSubHeaders = (maxSubHeader >> 3) +1;
 785             firstCodeArray = new char[numSubHeaders];
 786             entryCountArray = new char[numSubHeaders];
 787             idDeltaArray  = new short[numSubHeaders];
 788             idRangeOffSetArray  = new char[numSubHeaders];
 789             for (int i=0; i<numSubHeaders; i++) {
 790                 firstCodeArray[i] = cBuffer.get();
 791                 entryCountArray[i] = cBuffer.get();
 792                 idDeltaArray[i] = (short)cBuffer.get();
 793                 idRangeOffSetArray[i] = cBuffer.get();
 794 //              System.out.println("sh["+i+"]:fc="+(int)firstCodeArray[i]+
 795 //                                 " ec="+(int)entryCountArray[i]+
 796 //                                 " delta="+(int)idDeltaArray[i]+
 797 //                                 " offset="+(int)idRangeOffSetArray[i]);
 798             }
 799 
 800             int glyphIndexArrSize = (tableLen-518-numSubHeaders*8)/2;
 801             glyphIndexArray = new char[glyphIndexArrSize];
 802             for (int i=0; i<glyphIndexArrSize;i++) {
 803                 glyphIndexArray[i] = cBuffer.get();
 804             }
 805         }
 806 
 807         char getGlyph(int charCode) {
 808             final int origCharCode = charCode;
 809             int controlGlyph = getControlCodeGlyph(charCode, true);
 810             if (controlGlyph >= 0) {
 811                 return (char)controlGlyph;
 812             }
 813 
 814             if (xlat != null) {
 815                 charCode = xlat[charCode];
 816             }
 817 
 818             char highByte = (char)(charCode >> 8);
 819             char lowByte = (char)(charCode & 0xff);
 820             int key = subHeaderKey[highByte]>>3; // index into subHeaders
 821             char mapMe;
 822 
 823             if (key != 0) {
 824                 mapMe = lowByte;
 825             } else {
 826                 mapMe = highByte;
 827                 if (mapMe == 0) {
 828                     mapMe = lowByte;
 829                 }
 830             }
 831 
 832 //          System.err.println("charCode="+Integer.toHexString(charCode)+
 833 //                             " key="+key+ " mapMe="+Integer.toHexString(mapMe));
 834             char firstCode = firstCodeArray[key];
 835             if (mapMe < firstCode) {
 836                 return 0;
 837             } else {
 838                 mapMe -= firstCode;
 839             }
 840 
 841             if (mapMe < entryCountArray[key]) {
 842                 /* "address" arithmetic is needed to calculate the offset
 843                  * into glyphIndexArray. "idRangeOffSetArray[key]" specifies
 844                  * the number of bytes from that location in the table where
 845                  * the subarray of glyphIndexes starting at "firstCode" begins.
 846                  * Each entry in the subHeader table is 8 bytes, and the
 847                  * idRangeOffSetArray field is at offset 6 in the entry.
 848                  * The glyphIndexArray immediately follows the subHeaders.
 849                  * So if there are "N" entries then the number of bytes to the
 850                  * start of glyphIndexArray is (N-key)*8-6.
 851                  * Subtract this from the idRangeOffSetArray value to get
 852                  * the number of bytes into glyphIndexArray and divide by 2 to
 853                  * get the (char) array index.
 854                  */
 855                 int glyphArrayOffset = ((idRangeOffSetArray.length-key)*8)-6;
 856                 int glyphSubArrayStart =
 857                         (idRangeOffSetArray[key] - glyphArrayOffset)/2;
 858                 char glyphCode = glyphIndexArray[glyphSubArrayStart+mapMe];
 859                 if (glyphCode != 0) {
 860                     glyphCode += idDeltaArray[key]; //idDelta
 861                     return glyphCode;
 862                 }
 863             }
 864             return getFormatCharGlyph(origCharCode);
 865         }
 866     }
 867 
 868     // Format 6: Trimmed table mapping
 869     static class CMapFormat6 extends CMap {
 870 
 871         char firstCode;
 872         char entryCount;
 873         char[] glyphIdArray;
 874 
 875         CMapFormat6(ByteBuffer bbuffer, int offset, char[] xlat) {
 876 
 877              bbuffer.position(offset+6);
 878              CharBuffer buffer = bbuffer.asCharBuffer();
 879              firstCode = buffer.get();
 880              entryCount = buffer.get();
 881              glyphIdArray = new char[entryCount];
 882              for (int i=0; i< entryCount; i++) {
 883                  glyphIdArray[i] = buffer.get();
 884              }
 885          }
 886 
 887          char getGlyph(int charCode) {
 888             final int origCharCode = charCode;
 889             int controlGlyph = getControlCodeGlyph(charCode, true);
 890             if (controlGlyph >= 0) {
 891                 return (char)controlGlyph;
 892             }
 893 
 894              if (xlat != null) {
 895                  charCode = xlat[charCode];
 896              }
 897 
 898              charCode -= firstCode;
 899              if (charCode < 0 || charCode >= entryCount) {
 900                   return getFormatCharGlyph(origCharCode);
 901              } else {
 902                   return glyphIdArray[charCode];
 903              }
 904          }
 905     }
 906 
 907     // Format 8: mixed 16-bit and 32-bit coverage
 908     // Seems unlikely this code will ever get tested as we look for
 909     // MS platform Cmaps and MS states (in the Opentype spec on their website)
 910     // that MS doesn't support this format
 911     static class CMapFormat8 extends CMap {
 912          byte[] is32 = new byte[8192];
 913          int nGroups;
 914          int[] startCharCode;
 915          int[] endCharCode;
 916          int[] startGlyphID;
 917 
 918          CMapFormat8(ByteBuffer bbuffer, int offset, char[] xlat) {
 919 
 920              bbuffer.position(12);
 921              bbuffer.get(is32);
 922              nGroups = bbuffer.getInt() & INTMASK;
 923              // A map group record is three uint32's making for 12 bytes total
 924              if (bbuffer.remaining() < (12 * (long)nGroups)) {
 925                  throw new RuntimeException("Format 8 table exceeded");
 926              }
 927              startCharCode = new int[nGroups];
 928              endCharCode   = new int[nGroups];
 929              startGlyphID  = new int[nGroups];
 930          }
 931 
 932         char getGlyph(int charCode) {
 933             if (xlat != null) {
 934                 throw new RuntimeException("xlat array for cmap fmt=8");
 935             }
 936             return 0;
 937         }
 938 
 939     }
 940 
 941 
 942     // Format 4-byte 10: Trimmed table mapping
 943     // Seems unlikely this code will ever get tested as we look for
 944     // MS platform Cmaps and MS states (in the Opentype spec on their website)
 945     // that MS doesn't support this format
 946     static class CMapFormat10 extends CMap {
 947 
 948          long firstCode;
 949          int entryCount;
 950          char[] glyphIdArray;
 951 
 952          CMapFormat10(ByteBuffer bbuffer, int offset, char[] xlat) {
 953 
 954              bbuffer.position(offset+12);
 955              firstCode = bbuffer.getInt() & INTMASK;
 956              entryCount = bbuffer.getInt() & INTMASK;
 957              // each glyph is a uint16, so 2 bytes per value.
 958              if (bbuffer.remaining() < (2 * (long)entryCount)) {
 959                  throw new RuntimeException("Format 10 table exceeded");
 960              }
 961              CharBuffer buffer = bbuffer.asCharBuffer();
 962              glyphIdArray = new char[entryCount];
 963              for (int i=0; i< entryCount; i++) {
 964                  glyphIdArray[i] = buffer.get();
 965              }
 966          }
 967 
 968          char getGlyph(int charCode) {
 969 
 970              if (xlat != null) {
 971                  throw new RuntimeException("xlat array for cmap fmt=10");
 972              }
 973 
 974              int code = (int)(charCode - firstCode);
 975              if (code < 0 || code >= entryCount) {
 976                  return 0;
 977              } else {
 978                  return glyphIdArray[code];
 979              }
 980          }
 981     }
 982 
 983     // Format 12: Segmented coverage for UCS-4 (fonts supporting
 984     // surrogate pairs)
 985     static class CMapFormat12 extends CMap {
 986 
 987         int numGroups;
 988         int highBit =0;
 989         int power;
 990         int extra;
 991         long[] startCharCode;
 992         long[] endCharCode;
 993         int[] startGlyphID;
 994 
 995         CMapFormat12(ByteBuffer buffer, int offset, char[] xlat) {
 996             if (xlat != null) {
 997                 throw new RuntimeException("xlat array for cmap fmt=12");
 998             }
 999 
1000             buffer.position(offset+12);
1001             numGroups = buffer.getInt() & INTMASK;
1002             // A map group record is three uint32's making for 12 bytes total
1003             if (buffer.remaining() < (12 * (long)numGroups)) {
1004                 throw new RuntimeException("Format 12 table exceeded");
1005             }
1006             startCharCode = new long[numGroups];
1007             endCharCode = new long[numGroups];
1008             startGlyphID = new int[numGroups];
1009             buffer = buffer.slice();
1010             IntBuffer ibuffer = buffer.asIntBuffer();
1011             for (int i=0; i<numGroups; i++) {
1012                 startCharCode[i] = ibuffer.get() & INTMASK;
1013                 endCharCode[i] = ibuffer.get() & INTMASK;
1014                 startGlyphID[i] = ibuffer.get() & INTMASK;
1015             }
1016 
1017             /* Finds the high bit by binary searching through the bits */
1018             int value = numGroups;
1019 
1020             if (value >= 1 << 16) {
1021                 value >>= 16;
1022                 highBit += 16;
1023             }
1024 
1025             if (value >= 1 << 8) {
1026                 value >>= 8;
1027                 highBit += 8;
1028             }
1029 
1030             if (value >= 1 << 4) {
1031                 value >>= 4;
1032                 highBit += 4;
1033             }
1034 
1035             if (value >= 1 << 2) {
1036                 value >>= 2;
1037                 highBit += 2;
1038             }
1039 
1040             if (value >= 1 << 1) {
1041                 value >>= 1;
1042                 highBit += 1;
1043             }
1044 
1045             power = 1 << highBit;
1046             extra = numGroups - power;
1047         }
1048 
1049         char getGlyph(int charCode) {
1050             final int origCharCode = charCode;
1051             int controlGlyph = getControlCodeGlyph(charCode, false);
1052             if (controlGlyph >= 0) {
1053                 return (char)controlGlyph;
1054             }
1055             int probe = power;
1056             int range = 0;
1057 
1058             if (startCharCode[extra] <= charCode) {
1059                 range = extra;
1060             }
1061 
1062             while (probe > 1) {
1063                 probe >>= 1;
1064 
1065                 if (startCharCode[range+probe] <= charCode) {
1066                     range += probe;
1067                 }
1068             }
1069 
1070             if (startCharCode[range] <= charCode &&
1071                   endCharCode[range] >= charCode) {
1072                 return (char)
1073                     (startGlyphID[range] + (charCode - startCharCode[range]));
1074             }
1075 
1076             return getFormatCharGlyph(origCharCode);
1077         }
1078 
1079     }
1080 
1081     /* Used to substitute for bad Cmaps. */
1082     static class NullCMapClass extends CMap {
1083 
1084         char getGlyph(int charCode) {
1085             return 0;
1086         }
1087     }
1088 
1089     public static final NullCMapClass theNullCmap = new NullCMapClass();
1090 
1091     final int getControlCodeGlyph(int charCode, boolean noSurrogates) {
1092         if (charCode < 0x0010) {
1093             switch (charCode) {
1094             case 0x0009:
1095             case 0x000a:
1096             case 0x000d: return CharToGlyphMapper.INVISIBLE_GLYPH_ID;
1097             }
1098          } else if (noSurrogates && charCode >= 0xFFFF) {
1099             return 0;
1100         }
1101         return -1;
1102     }
1103 
1104     final char getFormatCharGlyph(int charCode) {
1105         if (charCode >= 0x200c) {
1106             if ((charCode <= 0x200f) ||
1107                 (charCode >= 0x2028 && charCode <= 0x202e) ||
1108                 (charCode >= 0x206a && charCode <= 0x206f)) {
1109                 return (char)CharToGlyphMapper.INVISIBLE_GLYPH_ID;
1110             }
1111         }
1112         return 0;
1113     }
1114 
1115     static class UVS {
1116         int numSelectors;
1117         int[] selector;
1118 
1119         //for Non-Default UVS Table
1120         int[] numUVSMapping;
1121         int[][] unicodeValue;
1122         char[][] glyphID;
1123 
1124         UVS(ByteBuffer buffer, int offset) {
1125             buffer.position(offset+6);
1126             numSelectors = buffer.getInt() & INTMASK;
1127             // A variation selector record is one 3 byte int + two int32's
1128             // making for 11 bytes per record.
1129             if (buffer.remaining() < (11 * (long)numSelectors)) {
1130                 throw new RuntimeException("Variations exceed buffer");
1131             }
1132             selector = new int[numSelectors];
1133             numUVSMapping = new int[numSelectors];
1134             unicodeValue = new int[numSelectors][];
1135             glyphID = new char[numSelectors][];
1136 
1137             for (int i = 0; i < numSelectors; i++) {
1138                 buffer.position(offset + 10 + i * 11);
1139                 selector[i] = (buffer.get() & 0xff) << 16; //UINT24
1140                 selector[i] += (buffer.get() & 0xff) << 8;
1141                 selector[i] += buffer.get() & 0xff;
1142 
1143                 //skip Default UVS Table
1144 
1145                 //for Non-Default UVS Table
1146                 int tableOffset = buffer.getInt(offset + 10 + i * 11 + 7);
1147                 if (tableOffset == 0) {
1148                     numUVSMapping[i] = 0;
1149                 } else if (tableOffset > 0) {
1150                     buffer.position(offset+tableOffset);
1151                     numUVSMapping[i] = buffer.getInt() & INTMASK;
1152                     // a UVS mapping record is one 3 byte int + uint16
1153                     // making for 5 bytes per record.
1154                     if (buffer.remaining() < (5 * (long)numUVSMapping[i])) {
1155                         throw new RuntimeException("Variations exceed buffer");
1156                     }
1157                     unicodeValue[i] = new int[numUVSMapping[i]];
1158                     glyphID[i] = new char[numUVSMapping[i]];
1159 
1160                     for (int j = 0; j < numUVSMapping[i]; j++) {
1161                         int temp = (buffer.get() & 0xff) << 16; //UINT24
1162                         temp += (buffer.get() & 0xff) << 8;
1163                         temp += buffer.get() & 0xff;
1164                         unicodeValue[i][j] = temp;
1165                         glyphID[i][j] = buffer.getChar();
1166                     }
1167                 }
1168             }
1169         }
1170 
1171         static final int VS_NOGLYPH = 0;
1172         private int getGlyph(int charCode, int variationSelector) {
1173             int targetSelector = -1;
1174             for (int i = 0; i < numSelectors; i++) {
1175                 if (selector[i] == variationSelector) {
1176                     targetSelector = i;
1177                     break;
1178                 }
1179             }
1180             if (targetSelector == -1) {
1181                 return VS_NOGLYPH;
1182             }
1183             if (numUVSMapping[targetSelector] > 0) {
1184                 int index = java.util.Arrays.binarySearch(
1185                                 unicodeValue[targetSelector], charCode);
1186                 if (index >= 0) {
1187                     return glyphID[targetSelector][index];
1188                 }
1189             }
1190             return VS_NOGLYPH;
1191         }
1192     }
1193 
1194     char getVariationGlyph(int charCode, int variationSelector) {
1195         char glyph = 0;
1196         if (uvs == null) {
1197             glyph = getGlyph(charCode);
1198         } else {
1199             int result = uvs.getGlyph(charCode, variationSelector);
1200             if (result > 0) {
1201                 glyph = (char)(result & 0xFFFF);
1202             } else {
1203                 glyph = getGlyph(charCode);
1204             }
1205         }
1206         return glyph;
1207     }
1208 }