1 /* 2 * Copyright (c) 1999, 2020, 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 /* 27 * (C) Copyright Taligent, Inc. 1996, 1997 - All Rights Reserved 28 * (C) Copyright IBM Corp. 1996-1998 - All Rights Reserved 29 * 30 * The original version of this source code and documentation is copyrighted 31 * and owned by Taligent, Inc., a wholly-owned subsidiary of IBM. These 32 * materials are provided under terms of a License Agreement between Taligent 33 * and Sun. This technology is protected by multiple US and International 34 * patents. This notice and attribution to Taligent may not be removed. 35 * Taligent is a registered trademark of Taligent, Inc. 36 * 37 */ 38 39 package java.text; 40 41 import java.util.Vector; 42 import sun.text.UCompactIntArray; 43 import sun.text.IntHashtable; 44 import sun.text.ComposedCharIter; 45 import jdk.internal.icu.impl.NormalizerImpl; 46 47 /** 48 * This class contains all the code to parse a RuleBasedCollator pattern 49 * and build a RBCollationTables object from it. A particular instance 50 * of tis class exists only during the actual build process-- once an 51 * RBCollationTables object has been built, the RBTableBuilder object 52 * goes away. This object carries all of the state which is only needed 53 * during the build process, plus a "shadow" copy of all of the state 54 * that will go into the tables object itself. This object communicates 55 * with RBCollationTables through a separate class, RBCollationTables.BuildAPI, 56 * this is an inner class of RBCollationTables and provides a separate 57 * private API for communication with RBTableBuilder. 58 * This class isn't just an inner class of RBCollationTables itself because 59 * of its large size. For source-code readability, it seemed better for the 60 * builder to have its own source file. 61 */ 62 final class RBTableBuilder { 63 64 public RBTableBuilder(RBCollationTables.BuildAPI tables) { 65 this.tables = tables; 66 } 67 68 /** 69 * Create a table-based collation object with the given rules. 70 * This is the main function that actually builds the tables and 71 * stores them back in the RBCollationTables object. It is called 72 * ONLY by the RBCollationTables constructor. 73 * @see RuleBasedCollator#RuleBasedCollator 74 * @throws ParseException If the rules format is incorrect. 75 */ 76 77 public void build(String pattern, int decmp) throws ParseException { 78 String expChars; 79 String groupChars; 80 if (pattern.isEmpty()) 81 throw new ParseException("Build rules empty.", 0); 82 83 // This array maps Unicode characters to their collation ordering 84 mapping = new UCompactIntArray(RBCollationTables.UNMAPPED); 85 // Normalize the build rules. Find occurances of all decomposed characters 86 // and normalize the rules before feeding into the builder. By "normalize", 87 // we mean that all precomposed Unicode characters must be converted into 88 // a base character and one or more combining characters (such as accents). 89 // When there are multiple combining characters attached to a base character, 90 // the combining characters must be in their canonical order 91 // 92 // sherman/Note: 93 //(1)decmp will be NO_DECOMPOSITION only in ko locale to prevent decompose 94 //hangual syllables to jamos, so we can actually just call decompose with 95 //normalizer's IGNORE_HANGUL option turned on 96 // 97 //(2)just call the "special version" in NormalizerImpl directly 98 //pattern = Normalizer.decompose(pattern, false, Normalizer.IGNORE_HANGUL, true); 99 // 100 //Normalizer.Mode mode = CollatorUtilities.toNormalizerMode(decmp); 101 //pattern = Normalizer.normalize(pattern, mode, 0, true); 102 103 pattern = NormalizerImpl.canonicalDecomposeWithSingleQuotation(pattern); 104 105 // Build the merged collation entries 106 // Since rules can be specified in any order in the string 107 // (e.g. "c , C < d , D < e , E .... C < CH") 108 // this splits all of the rules in the string out into separate 109 // objects and then sorts them. In the above example, it merges the 110 // "C < CH" rule in just before the "C < D" rule. 111 // 112 113 mPattern = new MergeCollation(pattern); 114 115 int order = 0; 116 117 // Now walk though each entry and add it to my own tables 118 for (int i = 0; i < mPattern.getCount(); ++i) { 119 PatternEntry entry = mPattern.getItemAt(i); 120 if (entry != null) { 121 groupChars = entry.getChars(); 122 if (groupChars.length() > 1) { 123 switch(groupChars.charAt(groupChars.length()-1)) { 124 case '@': 125 frenchSec = true; 126 groupChars = groupChars.substring(0, groupChars.length()-1); 127 break; 128 case '!': 129 seAsianSwapping = true; 130 groupChars = groupChars.substring(0, groupChars.length()-1); 131 break; 132 } 133 } 134 135 order = increment(entry.getStrength(), order); 136 expChars = entry.getExtension(); 137 138 if (!expChars.isEmpty()) { 139 addExpandOrder(groupChars, expChars, order); 140 } else if (groupChars.length() > 1) { 141 char ch = groupChars.charAt(0); 142 if (Character.isHighSurrogate(ch) && groupChars.length() == 2) { 143 addOrder(Character.toCodePoint(ch, groupChars.charAt(1)), order); 144 } else { 145 addContractOrder(groupChars, order); 146 } 147 } else { 148 char ch = groupChars.charAt(0); 149 addOrder(ch, order); 150 } 151 } 152 } 153 addComposedChars(); 154 155 commit(); 156 mapping.compact(); 157 /* 158 System.out.println("mappingSize=" + mapping.getKSize()); 159 for (int j = 0; j < 0xffff; j++) { 160 int value = mapping.elementAt(j); 161 if (value != RBCollationTables.UNMAPPED) 162 System.out.println("index=" + Integer.toString(j, 16) 163 + ", value=" + Integer.toString(value, 16)); 164 } 165 */ 166 tables.fillInTables(frenchSec, seAsianSwapping, mapping, contractTable, expandTable, 167 contractFlags, maxSecOrder, maxTerOrder); 168 } 169 170 /** Add expanding entries for pre-composed unicode characters so that this 171 * collator can be used reasonably well with decomposition turned off. 172 */ 173 private void addComposedChars() throws ParseException { 174 // Iterate through all of the pre-composed characters in Unicode 175 ComposedCharIter iter = new ComposedCharIter(); 176 int c; 177 while ((c = iter.next()) != ComposedCharIter.DONE) { 178 if (getCharOrder(c) == RBCollationTables.UNMAPPED) { 179 // 180 // We don't already have an ordering for this pre-composed character. 181 // 182 // First, see if the decomposed string is already in our 183 // tables as a single contracting-string ordering. 184 // If so, just map the precomposed character to that order. 185 // 186 // TODO: What we should really be doing here is trying to find the 187 // longest initial substring of the decomposition that is present 188 // in the tables as a contracting character sequence, and find its 189 // ordering. Then do this recursively with the remaining chars 190 // so that we build a list of orderings, and add that list to 191 // the expansion table. 192 // That would be more correct but also significantly slower, so 193 // I'm not totally sure it's worth doing. 194 // 195 String s = iter.decomposition(); 196 197 //sherman/Note: if this is 1 character decomposed string, the 198 //only thing need to do is to check if this decomposed character 199 //has an entry in our order table, this order is not necessary 200 //to be a contraction order, if it does have one, add an entry 201 //for the precomposed character by using the same order, the 202 //previous impl unnecessarily adds a single character expansion 203 //entry. 204 if (s.length() == 1) { 205 int order = getCharOrder(s.charAt(0)); 206 if (order != RBCollationTables.UNMAPPED) { 207 addOrder(c, order); 208 } 209 continue; 210 } else if (s.length() == 2) { 211 char ch0 = s.charAt(0); 212 if (Character.isHighSurrogate(ch0)) { 213 int order = getCharOrder(s.codePointAt(0)); 214 if (order != RBCollationTables.UNMAPPED) { 215 addOrder(c, order); 216 } 217 continue; 218 } 219 } 220 int contractOrder = getContractOrder(s); 221 if (contractOrder != RBCollationTables.UNMAPPED) { 222 addOrder(c, contractOrder); 223 } else { 224 // 225 // We don't have a contracting ordering for the entire string 226 // that results from the decomposition, but if we have orders 227 // for each individual character, we can add an expanding 228 // table entry for the pre-composed character 229 // 230 boolean allThere = true; 231 for (int i = 0; i < s.length(); i++) { 232 if (getCharOrder(s.charAt(i)) == RBCollationTables.UNMAPPED) { 233 allThere = false; 234 break; 235 } 236 } 237 if (allThere) { 238 addExpandOrder(c, s, RBCollationTables.UNMAPPED); 239 } 240 } 241 } 242 } 243 } 244 245 /** 246 * Look up for unmapped values in the expanded character table. 247 * 248 * When the expanding character tables are built by addExpandOrder, 249 * it doesn't know what the final ordering of each character 250 * in the expansion will be. Instead, it just puts the raw character 251 * code into the table, adding CHARINDEX as a flag. Now that we've 252 * finished building the mapping table, we can go back and look up 253 * that character to see what its real collation order is and 254 * stick that into the expansion table. That lets us avoid doing 255 * a two-stage lookup later. 256 */ 257 private final void commit() 258 { 259 if (expandTable != null) { 260 for (int i = 0; i < expandTable.size(); i++) { 261 int[] valueList = expandTable.elementAt(i); 262 for (int j = 0; j < valueList.length; j++) { 263 int order = valueList[j]; 264 if (order < RBCollationTables.EXPANDCHARINDEX && order > CHARINDEX) { 265 // found a expanding character that isn't filled in yet 266 int ch = order - CHARINDEX; 267 268 // Get the real values for the non-filled entry 269 int realValue = getCharOrder(ch); 270 271 if (realValue == RBCollationTables.UNMAPPED) { 272 // The real value is still unmapped, maybe it's ignorable 273 valueList[j] = IGNORABLEMASK & ch; 274 } else { 275 // just fill in the value 276 valueList[j] = realValue; 277 } 278 } 279 } 280 } 281 } 282 } 283 /** 284 * Increment of the last order based on the comparison level. 285 */ 286 private final int increment(int aStrength, int lastValue) 287 { 288 switch(aStrength) 289 { 290 case Collator.PRIMARY: 291 // increment priamry order and mask off secondary and tertiary difference 292 lastValue += PRIMARYORDERINCREMENT; 293 lastValue &= RBCollationTables.PRIMARYORDERMASK; 294 isOverIgnore = true; 295 break; 296 case Collator.SECONDARY: 297 // increment secondary order and mask off tertiary difference 298 lastValue += SECONDARYORDERINCREMENT; 299 lastValue &= RBCollationTables.SECONDARYDIFFERENCEONLY; 300 // record max # of ignorable chars with secondary difference 301 if (!isOverIgnore) 302 maxSecOrder++; 303 break; 304 case Collator.TERTIARY: 305 // increment tertiary order 306 lastValue += TERTIARYORDERINCREMENT; 307 // record max # of ignorable chars with tertiary difference 308 if (!isOverIgnore) 309 maxTerOrder++; 310 break; 311 } 312 return lastValue; 313 } 314 315 /** 316 * Adds a character and its designated order into the collation table. 317 */ 318 private final void addOrder(int ch, int anOrder) 319 { 320 // See if the char already has an order in the mapping table 321 int order = mapping.elementAt(ch); 322 323 if (order >= RBCollationTables.CONTRACTCHARINDEX) { 324 // There's already an entry for this character that points to a contracting 325 // character table. Instead of adding the character directly to the mapping 326 // table, we must add it to the contract table instead. 327 int length = 1; 328 if (Character.isSupplementaryCodePoint(ch)) { 329 length = Character.toChars(ch, keyBuf, 0); 330 } else { 331 keyBuf[0] = (char)ch; 332 } 333 addContractOrder(new String(keyBuf, 0, length), anOrder); 334 } else { 335 // add the entry to the mapping table, 336 // the same later entry replaces the previous one 337 mapping.setElementAt(ch, anOrder); 338 } 339 } 340 341 private final void addContractOrder(String groupChars, int anOrder) { 342 addContractOrder(groupChars, anOrder, true); 343 } 344 345 /** 346 * Adds the contracting string into the collation table. 347 */ 348 private final void addContractOrder(String groupChars, int anOrder, 349 boolean fwd) 350 { 351 if (contractTable == null) { 352 contractTable = new Vector<>(INITIALTABLESIZE); 353 } 354 355 //initial character 356 int ch = groupChars.codePointAt(0); 357 /* 358 char ch0 = groupChars.charAt(0); 359 int ch = Character.isHighSurrogate(ch0)? 360 Character.toCodePoint(ch0, groupChars.charAt(1)):ch0; 361 */ 362 // See if the initial character of the string already has a contract table. 363 int entry = mapping.elementAt(ch); 364 Vector<EntryPair> entryTable = getContractValuesImpl(entry - RBCollationTables.CONTRACTCHARINDEX); 365 366 if (entryTable == null) { 367 // We need to create a new table of contract entries for this base char 368 int tableIndex = RBCollationTables.CONTRACTCHARINDEX + contractTable.size(); 369 entryTable = new Vector<>(INITIALTABLESIZE); 370 contractTable.addElement(entryTable); 371 372 // Add the initial character's current ordering first. then 373 // update its mapping to point to this contract table 374 entryTable.addElement(new EntryPair(groupChars.substring(0,Character.charCount(ch)), entry)); 375 mapping.setElementAt(ch, tableIndex); 376 } 377 378 // Now add (or replace) this string in the table 379 int index = RBCollationTables.getEntry(entryTable, groupChars, fwd); 380 if (index != RBCollationTables.UNMAPPED) { 381 EntryPair pair = entryTable.elementAt(index); 382 pair.value = anOrder; 383 } else { 384 EntryPair pair = entryTable.lastElement(); 385 386 // NOTE: This little bit of logic is here to speed CollationElementIterator 387 // .nextContractChar(). This code ensures that the longest sequence in 388 // this list is always the _last_ one in the list. This keeps 389 // nextContractChar() from having to search the entire list for the longest 390 // sequence. 391 if (groupChars.length() > pair.entryName.length()) { 392 entryTable.addElement(new EntryPair(groupChars, anOrder, fwd)); 393 } else { 394 entryTable.insertElementAt(new EntryPair(groupChars, anOrder, 395 fwd), entryTable.size() - 1); 396 } 397 } 398 399 // If this was a forward mapping for a contracting string, also add a 400 // reverse mapping for it, so that CollationElementIterator.previous 401 // can work right 402 if (fwd && groupChars.length() > 1) { 403 addContractFlags(groupChars); 404 addContractOrder(new StringBuffer(groupChars).reverse().toString(), 405 anOrder, false); 406 } 407 } 408 409 /** 410 * If the given string has been specified as a contracting string 411 * in this collation table, return its ordering. 412 * Otherwise return UNMAPPED. 413 */ 414 private int getContractOrder(String groupChars) 415 { 416 int result = RBCollationTables.UNMAPPED; 417 if (contractTable != null) { 418 int ch = groupChars.codePointAt(0); 419 /* 420 char ch0 = groupChars.charAt(0); 421 int ch = Character.isHighSurrogate(ch0)? 422 Character.toCodePoint(ch0, groupChars.charAt(1)):ch0; 423 */ 424 Vector<EntryPair> entryTable = getContractValues(ch); 425 if (entryTable != null) { 426 int index = RBCollationTables.getEntry(entryTable, groupChars, true); 427 if (index != RBCollationTables.UNMAPPED) { 428 EntryPair pair = entryTable.elementAt(index); 429 result = pair.value; 430 } 431 } 432 } 433 return result; 434 } 435 436 private final int getCharOrder(int ch) { 437 int order = mapping.elementAt(ch); 438 439 if (order >= RBCollationTables.CONTRACTCHARINDEX) { 440 Vector<EntryPair> groupList = getContractValuesImpl(order - RBCollationTables.CONTRACTCHARINDEX); 441 EntryPair pair = groupList.firstElement(); 442 order = pair.value; 443 } 444 return order; 445 } 446 447 /** 448 * Get the entry of hash table of the contracting string in the collation 449 * table. 450 * @param ch the starting character of the contracting string 451 */ 452 private Vector<EntryPair> getContractValues(int ch) 453 { 454 int index = mapping.elementAt(ch); 455 return getContractValuesImpl(index - RBCollationTables.CONTRACTCHARINDEX); 456 } 457 458 private Vector<EntryPair> getContractValuesImpl(int index) 459 { 460 if (index >= 0) 461 { 462 return contractTable.elementAt(index); 463 } 464 else // not found 465 { 466 return null; 467 } 468 } 469 470 /** 471 * Adds the expanding string into the collation table. 472 */ 473 private final void addExpandOrder(String contractChars, 474 String expandChars, 475 int anOrder) throws ParseException 476 { 477 // Create an expansion table entry 478 int tableIndex = addExpansion(anOrder, expandChars); 479 480 // And add its index into the main mapping table 481 if (contractChars.length() > 1) { 482 char ch = contractChars.charAt(0); 483 if (Character.isHighSurrogate(ch) && contractChars.length() == 2) { 484 char ch2 = contractChars.charAt(1); 485 if (Character.isLowSurrogate(ch2)) { 486 //only add into table when it is a legal surrogate 487 addOrder(Character.toCodePoint(ch, ch2), tableIndex); 488 } 489 } else { 490 addContractOrder(contractChars, tableIndex); 491 } 492 } else { 493 addOrder(contractChars.charAt(0), tableIndex); 494 } 495 } 496 497 private final void addExpandOrder(int ch, String expandChars, int anOrder) 498 throws ParseException 499 { 500 int tableIndex = addExpansion(anOrder, expandChars); 501 addOrder(ch, tableIndex); 502 } 503 504 /** 505 * Create a new entry in the expansion table that contains the orderings 506 * for the given characers. If anOrder is valid, it is added to the 507 * beginning of the expanded list of orders. 508 */ 509 private int addExpansion(int anOrder, String expandChars) { 510 if (expandTable == null) { 511 expandTable = new Vector<>(INITIALTABLESIZE); 512 } 513 514 // If anOrder is valid, we want to add it at the beginning of the list 515 int offset = (anOrder == RBCollationTables.UNMAPPED) ? 0 : 1; 516 517 int[] valueList = new int[expandChars.length() + offset]; 518 if (offset == 1) { 519 valueList[0] = anOrder; 520 } 521 522 int j = offset; 523 for (int i = 0; i < expandChars.length(); i++) { 524 char ch0 = expandChars.charAt(i); 525 char ch1; 526 int ch; 527 if (Character.isHighSurrogate(ch0)) { 528 if (++i == expandChars.length() || 529 !Character.isLowSurrogate(ch1=expandChars.charAt(i))) { 530 //ether we are missing the low surrogate or the next char 531 //is not a legal low surrogate, so stop loop 532 break; 533 } 534 ch = Character.toCodePoint(ch0, ch1); 535 536 } else { 537 ch = ch0; 538 } 539 540 int mapValue = getCharOrder(ch); 541 542 if (mapValue != RBCollationTables.UNMAPPED) { 543 valueList[j++] = mapValue; 544 } else { 545 // can't find it in the table, will be filled in by commit(). 546 valueList[j++] = CHARINDEX + ch; 547 } 548 } 549 if (j < valueList.length) { 550 //we had at least one supplementary character, the size of valueList 551 //is bigger than it really needs... 552 int[] tmpBuf = new int[j]; 553 while (--j >= 0) { 554 tmpBuf[j] = valueList[j]; 555 } 556 valueList = tmpBuf; 557 } 558 // Add the expanding char list into the expansion table. 559 int tableIndex = RBCollationTables.EXPANDCHARINDEX + expandTable.size(); 560 expandTable.addElement(valueList); 561 562 return tableIndex; 563 } 564 565 private void addContractFlags(String chars) { 566 char c0; 567 int c; 568 int len = chars.length(); 569 for (int i = 0; i < len; i++) { 570 c0 = chars.charAt(i); 571 c = Character.isHighSurrogate(c0) 572 ?Character.toCodePoint(c0, chars.charAt(++i)) 573 :c0; 574 contractFlags.put(c, 1); 575 } 576 } 577 578 // ============================================================== 579 // constants 580 // ============================================================== 581 static final int CHARINDEX = 0x70000000; // need look up in .commit() 582 583 private static final int IGNORABLEMASK = 0x0000ffff; 584 private static final int PRIMARYORDERINCREMENT = 0x00010000; 585 private static final int SECONDARYORDERINCREMENT = 0x00000100; 586 private static final int TERTIARYORDERINCREMENT = 0x00000001; 587 private static final int INITIALTABLESIZE = 20; 588 private static final int MAXKEYSIZE = 5; 589 590 // ============================================================== 591 // instance variables 592 // ============================================================== 593 594 // variables used by the build process 595 private RBCollationTables.BuildAPI tables = null; 596 private MergeCollation mPattern = null; 597 private boolean isOverIgnore = false; 598 private char[] keyBuf = new char[MAXKEYSIZE]; 599 private IntHashtable contractFlags = new IntHashtable(100); 600 601 // "shadow" copies of the instance variables in RBCollationTables 602 // (the values in these variables are copied back into RBCollationTables 603 // at the end of the build process) 604 private boolean frenchSec = false; 605 private boolean seAsianSwapping = false; 606 607 private UCompactIntArray mapping = null; 608 private Vector<Vector<EntryPair>> contractTable = null; 609 private Vector<int[]> expandTable = null; 610 611 private short maxSecOrder = 0; 612 private short maxTerOrder = 0; 613 }