RuleBasedCollator.java revision 12745:f068a4ffddd2
1/* 2 * Copyright (c) 1997, 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/* 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 39package java.text; 40 41import java.text.Normalizer; 42import java.util.Vector; 43import java.util.Locale; 44 45/** 46 * The <code>RuleBasedCollator</code> class is a concrete subclass of 47 * <code>Collator</code> that provides a simple, data-driven, table 48 * collator. With this class you can create a customized table-based 49 * <code>Collator</code>. <code>RuleBasedCollator</code> maps 50 * characters to sort keys. 51 * 52 * <p> 53 * <code>RuleBasedCollator</code> has the following restrictions 54 * for efficiency (other subclasses may be used for more complex languages) : 55 * <ol> 56 * <li>If a special collation rule controlled by a <modifier> is 57 specified it applies to the whole collator object. 58 * <li>All non-mentioned characters are at the end of the 59 * collation order. 60 * </ol> 61 * 62 * <p> 63 * The collation table is composed of a list of collation rules, where each 64 * rule is of one of three forms: 65 * <pre> 66 * <modifier> 67 * <relation> <text-argument> 68 * <reset> <text-argument> 69 * </pre> 70 * The definitions of the rule elements is as follows: 71 * <UL> 72 * <LI><strong>Text-Argument</strong>: A text-argument is any sequence of 73 * characters, excluding special characters (that is, common 74 * whitespace characters [0009-000D, 0020] and rule syntax characters 75 * [0021-002F, 003A-0040, 005B-0060, 007B-007E]). If those 76 * characters are desired, you can put them in single quotes 77 * (e.g. ampersand => '&'). Note that unquoted white space characters 78 * are ignored; e.g. <code>b c</code> is treated as <code>bc</code>. 79 * <LI><strong>Modifier</strong>: There are currently two modifiers that 80 * turn on special collation rules. 81 * <UL> 82 * <LI>'@' : Turns on backwards sorting of accents (secondary 83 * differences), as in French. 84 * <LI>'!' : Turns on Thai/Lao vowel-consonant swapping. If this 85 * rule is in force when a Thai vowel of the range 86 * \U0E40-\U0E44 precedes a Thai consonant of the range 87 * \U0E01-\U0E2E OR a Lao vowel of the range \U0EC0-\U0EC4 88 * precedes a Lao consonant of the range \U0E81-\U0EAE then 89 * the vowel is placed after the consonant for collation 90 * purposes. 91 * </UL> 92 * <p>'@' : Indicates that accents are sorted backwards, as in French. 93 * <LI><strong>Relation</strong>: The relations are the following: 94 * <UL> 95 * <LI>'<' : Greater, as a letter difference (primary) 96 * <LI>';' : Greater, as an accent difference (secondary) 97 * <LI>',' : Greater, as a case difference (tertiary) 98 * <LI>'=' : Equal 99 * </UL> 100 * <LI><strong>Reset</strong>: There is a single reset 101 * which is used primarily for contractions and expansions, but which 102 * can also be used to add a modification at the end of a set of rules. 103 * <p>'&' : Indicates that the next rule follows the position to where 104 * the reset text-argument would be sorted. 105 * </UL> 106 * 107 * <p> 108 * This sounds more complicated than it is in practice. For example, the 109 * following are equivalent ways of expressing the same thing: 110 * <blockquote> 111 * <pre> 112 * a < b < c 113 * a < b & b < c 114 * a < c & a < b 115 * </pre> 116 * </blockquote> 117 * Notice that the order is important, as the subsequent item goes immediately 118 * after the text-argument. The following are not equivalent: 119 * <blockquote> 120 * <pre> 121 * a < b & a < c 122 * a < c & a < b 123 * </pre> 124 * </blockquote> 125 * Either the text-argument must already be present in the sequence, or some 126 * initial substring of the text-argument must be present. (e.g. "a < b & ae < 127 * e" is valid since "a" is present in the sequence before "ae" is reset). In 128 * this latter case, "ae" is not entered and treated as a single character; 129 * instead, "e" is sorted as if it were expanded to two characters: "a" 130 * followed by an "e". This difference appears in natural languages: in 131 * traditional Spanish "ch" is treated as though it contracts to a single 132 * character (expressed as "c < ch < d"), while in traditional German 133 * a-umlaut is treated as though it expanded to two characters 134 * (expressed as "a,A < b,B ... &ae;\u00e3&AE;\u00c3"). 135 * [\u00e3 and \u00c3 are, of course, the escape sequences for a-umlaut.] 136 * <p> 137 * <strong>Ignorable Characters</strong> 138 * <p> 139 * For ignorable characters, the first rule must start with a relation (the 140 * examples we have used above are really fragments; "a < b" really should be 141 * "< a < b"). If, however, the first relation is not "<", then all the all 142 * text-arguments up to the first "<" are ignorable. For example, ", - < a < b" 143 * makes "-" an ignorable character, as we saw earlier in the word 144 * "black-birds". In the samples for different languages, you see that most 145 * accents are ignorable. 146 * 147 * <p><strong>Normalization and Accents</strong> 148 * <p> 149 * <code>RuleBasedCollator</code> automatically processes its rule table to 150 * include both pre-composed and combining-character versions of 151 * accented characters. Even if the provided rule string contains only 152 * base characters and separate combining accent characters, the pre-composed 153 * accented characters matching all canonical combinations of characters from 154 * the rule string will be entered in the table. 155 * <p> 156 * This allows you to use a RuleBasedCollator to compare accented strings 157 * even when the collator is set to NO_DECOMPOSITION. There are two caveats, 158 * however. First, if the strings to be collated contain combining 159 * sequences that may not be in canonical order, you should set the collator to 160 * CANONICAL_DECOMPOSITION or FULL_DECOMPOSITION to enable sorting of 161 * combining sequences. Second, if the strings contain characters with 162 * compatibility decompositions (such as full-width and half-width forms), 163 * you must use FULL_DECOMPOSITION, since the rule tables only include 164 * canonical mappings. 165 * 166 * <p><strong>Errors</strong> 167 * <p> 168 * The following are errors: 169 * <UL> 170 * <LI>A text-argument contains unquoted punctuation symbols 171 * (e.g. "a < b-c < d"). 172 * <LI>A relation or reset character not followed by a text-argument 173 * (e.g. "a < ,b"). 174 * <LI>A reset where the text-argument (or an initial substring of the 175 * text-argument) is not already in the sequence. 176 * (e.g. "a < b & e < f") 177 * </UL> 178 * If you produce one of these errors, a <code>RuleBasedCollator</code> throws 179 * a <code>ParseException</code>. 180 * 181 * <p><strong>Examples</strong> 182 * <p>Simple: "< a < b < c < d" 183 * <p>Norwegian: "< a, A < b, B < c, C < d, D < e, E < f, F 184 * < g, G < h, H < i, I < j, J < k, K < l, L 185 * < m, M < n, N < o, O < p, P < q, Q < r, R 186 * < s, S < t, T < u, U < v, V < w, W < x, X 187 * < y, Y < z, Z 188 * < \u00E6, \u00C6 189 * < \u00F8, \u00D8 190 * < \u00E5 = a\u030A, \u00C5 = A\u030A; 191 * aa, AA" 192 * 193 * <p> 194 * To create a <code>RuleBasedCollator</code> object with specialized 195 * rules tailored to your needs, you construct the <code>RuleBasedCollator</code> 196 * with the rules contained in a <code>String</code> object. For example: 197 * <blockquote> 198 * <pre> 199 * String simple = "< a< b< c< d"; 200 * RuleBasedCollator mySimple = new RuleBasedCollator(simple); 201 * </pre> 202 * </blockquote> 203 * Or: 204 * <blockquote> 205 * <pre> 206 * String Norwegian = "< a, A < b, B < c, C < d, D < e, E < f, F < g, G < h, H < i, I" + 207 * "< j, J < k, K < l, L < m, M < n, N < o, O < p, P < q, Q < r, R" + 208 * "< s, S < t, T < u, U < v, V < w, W < x, X < y, Y < z, Z" + 209 * "< \u00E6, \u00C6" + // Latin letter ae & AE 210 * "< \u00F8, \u00D8" + // Latin letter o & O with stroke 211 * "< \u00E5 = a\u030A," + // Latin letter a with ring above 212 * " \u00C5 = A\u030A;" + // Latin letter A with ring above 213 * " aa, AA"; 214 * RuleBasedCollator myNorwegian = new RuleBasedCollator(Norwegian); 215 * </pre> 216 * </blockquote> 217 * 218 * <p> 219 * A new collation rules string can be created by concatenating rules 220 * strings. For example, the rules returned by {@link #getRules()} could 221 * be concatenated to combine multiple <code>RuleBasedCollator</code>s. 222 * 223 * <p> 224 * The following example demonstrates how to change the order of 225 * non-spacing accents, 226 * <blockquote> 227 * <pre> 228 * // old rule 229 * String oldRules = "=\u0301;\u0300;\u0302;\u0308" // main accents 230 * + ";\u0327;\u0303;\u0304;\u0305" // main accents 231 * + ";\u0306;\u0307;\u0309;\u030A" // main accents 232 * + ";\u030B;\u030C;\u030D;\u030E" // main accents 233 * + ";\u030F;\u0310;\u0311;\u0312" // main accents 234 * + "< a , A ; ae, AE ; \u00e6 , \u00c6" 235 * + "< b , B < c, C < e, E & C < d, D"; 236 * // change the order of accent characters 237 * String addOn = "& \u0300 ; \u0308 ; \u0302"; 238 * RuleBasedCollator myCollator = new RuleBasedCollator(oldRules + addOn); 239 * </pre> 240 * </blockquote> 241 * 242 * @see Collator 243 * @see CollationElementIterator 244 * @author Helena Shih, Laura Werner, Richard Gillam 245 */ 246public class RuleBasedCollator extends Collator{ 247 // IMPLEMENTATION NOTES: The implementation of the collation algorithm is 248 // divided across three classes: RuleBasedCollator, RBCollationTables, and 249 // CollationElementIterator. RuleBasedCollator contains the collator's 250 // transient state and includes the code that uses the other classes to 251 // implement comparison and sort-key building. RuleBasedCollator also 252 // contains the logic to handle French secondary accent sorting. 253 // A RuleBasedCollator has two CollationElementIterators. State doesn't 254 // need to be preserved in these objects between calls to compare() or 255 // getCollationKey(), but the objects persist anyway to avoid wasting extra 256 // creation time. compare() and getCollationKey() are synchronized to ensure 257 // thread safety with this scheme. The CollationElementIterator is responsible 258 // for generating collation elements from strings and returning one element at 259 // a time (sometimes there's a one-to-many or many-to-one mapping between 260 // characters and collation elements-- this class handles that). 261 // CollationElementIterator depends on RBCollationTables, which contains the 262 // collator's static state. RBCollationTables contains the actual data 263 // tables specifying the collation order of characters for a particular locale 264 // or use. It also contains the base logic that CollationElementIterator 265 // uses to map from characters to collation elements. A single RBCollationTables 266 // object is shared among all RuleBasedCollators for the same locale, and 267 // thus by all the CollationElementIterators they create. 268 269 /** 270 * RuleBasedCollator constructor. This takes the table rules and builds 271 * a collation table out of them. Please see RuleBasedCollator class 272 * description for more details on the collation rule syntax. 273 * @see java.util.Locale 274 * @param rules the collation rules to build the collation table from. 275 * @exception ParseException A format exception 276 * will be thrown if the build process of the rules fails. For 277 * example, build rule "a < ? < d" will cause the constructor to 278 * throw the ParseException because the '?' is not quoted. 279 */ 280 public RuleBasedCollator(String rules) throws ParseException { 281 this(rules, Collator.CANONICAL_DECOMPOSITION); 282 } 283 284 /** 285 * RuleBasedCollator constructor. This takes the table rules and builds 286 * a collation table out of them. Please see RuleBasedCollator class 287 * description for more details on the collation rule syntax. 288 * @see java.util.Locale 289 * @param rules the collation rules to build the collation table from. 290 * @param decomp the decomposition strength used to build the 291 * collation table and to perform comparisons. 292 * @exception ParseException A format exception 293 * will be thrown if the build process of the rules fails. For 294 * example, build rule "a < ? < d" will cause the constructor to 295 * throw the ParseException because the '?' is not quoted. 296 */ 297 RuleBasedCollator(String rules, int decomp) throws ParseException { 298 setStrength(Collator.TERTIARY); 299 setDecomposition(decomp); 300 tables = new RBCollationTables(rules, decomp); 301 } 302 303 /** 304 * "Copy constructor." Used in clone() for performance. 305 */ 306 private RuleBasedCollator(RuleBasedCollator that) { 307 setStrength(that.getStrength()); 308 setDecomposition(that.getDecomposition()); 309 tables = that.tables; 310 } 311 312 /** 313 * Gets the table-based rules for the collation object. 314 * @return returns the collation rules that the table collation object 315 * was created from. 316 */ 317 public String getRules() 318 { 319 return tables.getRules(); 320 } 321 322 /** 323 * Returns a CollationElementIterator for the given String. 324 * 325 * @param source the string to be collated 326 * @return a {@code CollationElementIterator} object 327 * @see java.text.CollationElementIterator 328 */ 329 public CollationElementIterator getCollationElementIterator(String source) { 330 return new CollationElementIterator( source, this ); 331 } 332 333 /** 334 * Returns a CollationElementIterator for the given CharacterIterator. 335 * 336 * @param source the character iterator to be collated 337 * @return a {@code CollationElementIterator} object 338 * @see java.text.CollationElementIterator 339 * @since 1.2 340 */ 341 public CollationElementIterator getCollationElementIterator( 342 CharacterIterator source) { 343 return new CollationElementIterator( source, this ); 344 } 345 346 /** 347 * Compares the character data stored in two different strings based on the 348 * collation rules. Returns information about whether a string is less 349 * than, greater than or equal to another string in a language. 350 * This can be overriden in a subclass. 351 * 352 * @exception NullPointerException if <code>source</code> or <code>target</code> is null. 353 */ 354 public synchronized int compare(String source, String target) 355 { 356 if (source == null || target == null) { 357 throw new NullPointerException(); 358 } 359 360 // The basic algorithm here is that we use CollationElementIterators 361 // to step through both the source and target strings. We compare each 362 // collation element in the source string against the corresponding one 363 // in the target, checking for differences. 364 // 365 // If a difference is found, we set <result> to LESS or GREATER to 366 // indicate whether the source string is less or greater than the target. 367 // 368 // However, it's not that simple. If we find a tertiary difference 369 // (e.g. 'A' vs. 'a') near the beginning of a string, it can be 370 // overridden by a primary difference (e.g. "A" vs. "B") later in 371 // the string. For example, "AA" < "aB", even though 'A' > 'a'. 372 // 373 // To keep track of this, we use strengthResult to keep track of the 374 // strength of the most significant difference that has been found 375 // so far. When we find a difference whose strength is greater than 376 // strengthResult, it overrides the last difference (if any) that 377 // was found. 378 379 int result = Collator.EQUAL; 380 381 if (sourceCursor == null) { 382 sourceCursor = getCollationElementIterator(source); 383 } else { 384 sourceCursor.setText(source); 385 } 386 if (targetCursor == null) { 387 targetCursor = getCollationElementIterator(target); 388 } else { 389 targetCursor.setText(target); 390 } 391 392 int sOrder = 0, tOrder = 0; 393 394 boolean initialCheckSecTer = getStrength() >= Collator.SECONDARY; 395 boolean checkSecTer = initialCheckSecTer; 396 boolean checkTertiary = getStrength() >= Collator.TERTIARY; 397 398 boolean gets = true, gett = true; 399 400 while(true) { 401 // Get the next collation element in each of the strings, unless 402 // we've been requested to skip it. 403 if (gets) sOrder = sourceCursor.next(); else gets = true; 404 if (gett) tOrder = targetCursor.next(); else gett = true; 405 406 // If we've hit the end of one of the strings, jump out of the loop 407 if ((sOrder == CollationElementIterator.NULLORDER)|| 408 (tOrder == CollationElementIterator.NULLORDER)) 409 break; 410 411 int pSOrder = CollationElementIterator.primaryOrder(sOrder); 412 int pTOrder = CollationElementIterator.primaryOrder(tOrder); 413 414 // If there's no difference at this position, we can skip it 415 if (sOrder == tOrder) { 416 if (tables.isFrenchSec() && pSOrder != 0) { 417 if (!checkSecTer) { 418 // in french, a secondary difference more to the right is stronger, 419 // so accents have to be checked with each base element 420 checkSecTer = initialCheckSecTer; 421 // but tertiary differences are less important than the first 422 // secondary difference, so checking tertiary remains disabled 423 checkTertiary = false; 424 } 425 } 426 continue; 427 } 428 429 // Compare primary differences first. 430 if ( pSOrder != pTOrder ) 431 { 432 if (sOrder == 0) { 433 // The entire source element is ignorable. 434 // Skip to the next source element, but don't fetch another target element. 435 gett = false; 436 continue; 437 } 438 if (tOrder == 0) { 439 gets = false; 440 continue; 441 } 442 443 // The source and target elements aren't ignorable, but it's still possible 444 // for the primary component of one of the elements to be ignorable.... 445 446 if (pSOrder == 0) // primary order in source is ignorable 447 { 448 // The source's primary is ignorable, but the target's isn't. We treat ignorables 449 // as a secondary difference, so remember that we found one. 450 if (checkSecTer) { 451 result = Collator.GREATER; // (strength is SECONDARY) 452 checkSecTer = false; 453 } 454 // Skip to the next source element, but don't fetch another target element. 455 gett = false; 456 } 457 else if (pTOrder == 0) 458 { 459 // record differences - see the comment above. 460 if (checkSecTer) { 461 result = Collator.LESS; // (strength is SECONDARY) 462 checkSecTer = false; 463 } 464 // Skip to the next source element, but don't fetch another target element. 465 gets = false; 466 } else { 467 // Neither of the orders is ignorable, and we already know that the primary 468 // orders are different because of the (pSOrder != pTOrder) test above. 469 // Record the difference and stop the comparison. 470 if (pSOrder < pTOrder) { 471 return Collator.LESS; // (strength is PRIMARY) 472 } else { 473 return Collator.GREATER; // (strength is PRIMARY) 474 } 475 } 476 } else { // else of if ( pSOrder != pTOrder ) 477 // primary order is the same, but complete order is different. So there 478 // are no base elements at this point, only ignorables (Since the strings are 479 // normalized) 480 481 if (checkSecTer) { 482 // a secondary or tertiary difference may still matter 483 short secSOrder = CollationElementIterator.secondaryOrder(sOrder); 484 short secTOrder = CollationElementIterator.secondaryOrder(tOrder); 485 if (secSOrder != secTOrder) { 486 // there is a secondary difference 487 result = (secSOrder < secTOrder) ? Collator.LESS : Collator.GREATER; 488 // (strength is SECONDARY) 489 checkSecTer = false; 490 // (even in french, only the first secondary difference within 491 // a base character matters) 492 } else { 493 if (checkTertiary) { 494 // a tertiary difference may still matter 495 short terSOrder = CollationElementIterator.tertiaryOrder(sOrder); 496 short terTOrder = CollationElementIterator.tertiaryOrder(tOrder); 497 if (terSOrder != terTOrder) { 498 // there is a tertiary difference 499 result = (terSOrder < terTOrder) ? Collator.LESS : Collator.GREATER; 500 // (strength is TERTIARY) 501 checkTertiary = false; 502 } 503 } 504 } 505 } // if (checkSecTer) 506 507 } // if ( pSOrder != pTOrder ) 508 } // while() 509 510 if (sOrder != CollationElementIterator.NULLORDER) { 511 // (tOrder must be CollationElementIterator::NULLORDER, 512 // since this point is only reached when sOrder or tOrder is NULLORDER.) 513 // The source string has more elements, but the target string hasn't. 514 do { 515 if (CollationElementIterator.primaryOrder(sOrder) != 0) { 516 // We found an additional non-ignorable base character in the source string. 517 // This is a primary difference, so the source is greater 518 return Collator.GREATER; // (strength is PRIMARY) 519 } 520 else if (CollationElementIterator.secondaryOrder(sOrder) != 0) { 521 // Additional secondary elements mean the source string is greater 522 if (checkSecTer) { 523 result = Collator.GREATER; // (strength is SECONDARY) 524 checkSecTer = false; 525 } 526 } 527 } while ((sOrder = sourceCursor.next()) != CollationElementIterator.NULLORDER); 528 } 529 else if (tOrder != CollationElementIterator.NULLORDER) { 530 // The target string has more elements, but the source string hasn't. 531 do { 532 if (CollationElementIterator.primaryOrder(tOrder) != 0) 533 // We found an additional non-ignorable base character in the target string. 534 // This is a primary difference, so the source is less 535 return Collator.LESS; // (strength is PRIMARY) 536 else if (CollationElementIterator.secondaryOrder(tOrder) != 0) { 537 // Additional secondary elements in the target mean the source string is less 538 if (checkSecTer) { 539 result = Collator.LESS; // (strength is SECONDARY) 540 checkSecTer = false; 541 } 542 } 543 } while ((tOrder = targetCursor.next()) != CollationElementIterator.NULLORDER); 544 } 545 546 // For IDENTICAL comparisons, we use a bitwise character comparison 547 // as a tiebreaker if all else is equal 548 if (result == 0 && getStrength() == IDENTICAL) { 549 int mode = getDecomposition(); 550 Normalizer.Form form; 551 if (mode == CANONICAL_DECOMPOSITION) { 552 form = Normalizer.Form.NFD; 553 } else if (mode == FULL_DECOMPOSITION) { 554 form = Normalizer.Form.NFKD; 555 } else { 556 return source.compareTo(target); 557 } 558 559 String sourceDecomposition = Normalizer.normalize(source, form); 560 String targetDecomposition = Normalizer.normalize(target, form); 561 return sourceDecomposition.compareTo(targetDecomposition); 562 } 563 return result; 564 } 565 566 /** 567 * Transforms the string into a series of characters that can be compared 568 * with CollationKey.compareTo. This overrides java.text.Collator.getCollationKey. 569 * It can be overriden in a subclass. 570 */ 571 public synchronized CollationKey getCollationKey(String source) 572 { 573 // 574 // The basic algorithm here is to find all of the collation elements for each 575 // character in the source string, convert them to a char representation, 576 // and put them into the collation key. But it's trickier than that. 577 // Each collation element in a string has three components: primary (A vs B), 578 // secondary (A vs A-acute), and tertiary (A' vs a); and a primary difference 579 // at the end of a string takes precedence over a secondary or tertiary 580 // difference earlier in the string. 581 // 582 // To account for this, we put all of the primary orders at the beginning of the 583 // string, followed by the secondary and tertiary orders, separated by nulls. 584 // 585 // Here's a hypothetical example, with the collation element represented as 586 // a three-digit number, one digit for primary, one for secondary, etc. 587 // 588 // String: A a B \u00e9 <--(e-acute) 589 // Collation Elements: 101 100 201 510 590 // 591 // Collation Key: 1125<null>0001<null>1010 592 // 593 // To make things even trickier, secondary differences (accent marks) are compared 594 // starting at the *end* of the string in languages with French secondary ordering. 595 // But when comparing the accent marks on a single base character, they are compared 596 // from the beginning. To handle this, we reverse all of the accents that belong 597 // to each base character, then we reverse the entire string of secondary orderings 598 // at the end. Taking the same example above, a French collator might return 599 // this instead: 600 // 601 // Collation Key: 1125<null>1000<null>1010 602 // 603 if (source == null) 604 return null; 605 606 if (primResult == null) { 607 primResult = new StringBuffer(); 608 secResult = new StringBuffer(); 609 terResult = new StringBuffer(); 610 } else { 611 primResult.setLength(0); 612 secResult.setLength(0); 613 terResult.setLength(0); 614 } 615 int order = 0; 616 boolean compareSec = (getStrength() >= Collator.SECONDARY); 617 boolean compareTer = (getStrength() >= Collator.TERTIARY); 618 int secOrder = CollationElementIterator.NULLORDER; 619 int terOrder = CollationElementIterator.NULLORDER; 620 int preSecIgnore = 0; 621 622 if (sourceCursor == null) { 623 sourceCursor = getCollationElementIterator(source); 624 } else { 625 sourceCursor.setText(source); 626 } 627 628 // walk through each character 629 while ((order = sourceCursor.next()) != 630 CollationElementIterator.NULLORDER) 631 { 632 secOrder = CollationElementIterator.secondaryOrder(order); 633 terOrder = CollationElementIterator.tertiaryOrder(order); 634 if (!CollationElementIterator.isIgnorable(order)) 635 { 636 primResult.append((char) (CollationElementIterator.primaryOrder(order) 637 + COLLATIONKEYOFFSET)); 638 639 if (compareSec) { 640 // 641 // accumulate all of the ignorable/secondary characters attached 642 // to a given base character 643 // 644 if (tables.isFrenchSec() && preSecIgnore < secResult.length()) { 645 // 646 // We're doing reversed secondary ordering and we've hit a base 647 // (non-ignorable) character. Reverse any secondary orderings 648 // that applied to the last base character. (see block comment above.) 649 // 650 RBCollationTables.reverse(secResult, preSecIgnore, secResult.length()); 651 } 652 // Remember where we are in the secondary orderings - this is how far 653 // back to go if we need to reverse them later. 654 secResult.append((char)(secOrder+ COLLATIONKEYOFFSET)); 655 preSecIgnore = secResult.length(); 656 } 657 if (compareTer) { 658 terResult.append((char)(terOrder+ COLLATIONKEYOFFSET)); 659 } 660 } 661 else 662 { 663 if (compareSec && secOrder != 0) 664 secResult.append((char) 665 (secOrder + tables.getMaxSecOrder() + COLLATIONKEYOFFSET)); 666 if (compareTer && terOrder != 0) 667 terResult.append((char) 668 (terOrder + tables.getMaxTerOrder() + COLLATIONKEYOFFSET)); 669 } 670 } 671 if (tables.isFrenchSec()) 672 { 673 if (preSecIgnore < secResult.length()) { 674 // If we've accumulated any secondary characters after the last base character, 675 // reverse them. 676 RBCollationTables.reverse(secResult, preSecIgnore, secResult.length()); 677 } 678 // And now reverse the entire secResult to get French secondary ordering. 679 RBCollationTables.reverse(secResult, 0, secResult.length()); 680 } 681 primResult.append((char)0); 682 secResult.append((char)0); 683 secResult.append(terResult.toString()); 684 primResult.append(secResult.toString()); 685 686 if (getStrength() == IDENTICAL) { 687 primResult.append((char)0); 688 int mode = getDecomposition(); 689 if (mode == CANONICAL_DECOMPOSITION) { 690 primResult.append(Normalizer.normalize(source, Normalizer.Form.NFD)); 691 } else if (mode == FULL_DECOMPOSITION) { 692 primResult.append(Normalizer.normalize(source, Normalizer.Form.NFKD)); 693 } else { 694 primResult.append(source); 695 } 696 } 697 return new RuleBasedCollationKey(source, primResult.toString()); 698 } 699 700 /** 701 * Standard override; no change in semantics. 702 */ 703 public Object clone() { 704 // if we know we're not actually a subclass of RuleBasedCollator 705 // (this class really should have been made final), bypass 706 // Object.clone() and use our "copy constructor". This is faster. 707 if (getClass() == RuleBasedCollator.class) { 708 return new RuleBasedCollator(this); 709 } 710 else { 711 RuleBasedCollator result = (RuleBasedCollator) super.clone(); 712 result.primResult = null; 713 result.secResult = null; 714 result.terResult = null; 715 result.sourceCursor = null; 716 result.targetCursor = null; 717 return result; 718 } 719 } 720 721 /** 722 * Compares the equality of two collation objects. 723 * @param obj the table-based collation object to be compared with this. 724 * @return true if the current table-based collation object is the same 725 * as the table-based collation object obj; false otherwise. 726 */ 727 public boolean equals(Object obj) { 728 if (obj == null) return false; 729 if (!super.equals(obj)) return false; // super does class check 730 RuleBasedCollator other = (RuleBasedCollator) obj; 731 // all other non-transient information is also contained in rules. 732 return (getRules().equals(other.getRules())); 733 } 734 735 /** 736 * Generates the hash code for the table-based collation object 737 */ 738 public int hashCode() { 739 return getRules().hashCode(); 740 } 741 742 /** 743 * Allows CollationElementIterator access to the tables object 744 */ 745 RBCollationTables getTables() { 746 return tables; 747 } 748 749 // ============================================================== 750 // private 751 // ============================================================== 752 753 static final int CHARINDEX = 0x70000000; // need look up in .commit() 754 static final int EXPANDCHARINDEX = 0x7E000000; // Expand index follows 755 static final int CONTRACTCHARINDEX = 0x7F000000; // contract indexes follow 756 static final int UNMAPPED = 0xFFFFFFFF; 757 758 private static final int COLLATIONKEYOFFSET = 1; 759 760 private RBCollationTables tables = null; 761 762 // Internal objects that are cached across calls so that they don't have to 763 // be created/destroyed on every call to compare() and getCollationKey() 764 private StringBuffer primResult = null; 765 private StringBuffer secResult = null; 766 private StringBuffer terResult = null; 767 private CollationElementIterator sourceCursor = null; 768 private CollationElementIterator targetCursor = null; 769} 770