1 /* 2 ****************************************************************************** 3 * Copyright (C) 1996-2012, International Business Machines Corporation and 4 * others. All Rights Reserved. 5 ****************************************************************************** 6 */ 7 8 /** 9 * File tblcoll.cpp 10 * 11 * Created by: Helena Shih 12 * 13 * Modification History: 14 * 15 * Date Name Description 16 * 2/5/97 aliu Added streamIn and streamOut methods. Added 17 * constructor which reads RuleBasedCollator object from 18 * a binary file. Added writeToFile method which streams 19 * RuleBasedCollator out to a binary file. The streamIn 20 * and streamOut methods use istream and ostream objects 21 * in binary mode. 22 * 2/11/97 aliu Moved declarations out of for loop initializer. 23 * Added Mac compatibility #ifdef for ios::nocreate. 24 * 2/12/97 aliu Modified to use TableCollationData sub-object to 25 * hold invariant data. 26 * 2/13/97 aliu Moved several methods into this class from Collation. 27 * Added a private RuleBasedCollator(Locale&) constructor, 28 * to be used by Collator::getInstance(). General 29 * clean up. Made use of UErrorCode variables consistent. 30 * 2/20/97 helena Added clone, operator==, operator!=, operator=, and copy 31 * constructor and getDynamicClassID. 32 * 3/5/97 aliu Changed compaction cycle to improve performance. We 33 * use the maximum allowable value which is kBlockCount. 34 * Modified getRules() to load rules dynamically. Changed 35 * constructFromFile() call to accomodate this (added 36 * parameter to specify whether binary loading is to 37 * take place). 38 * 05/06/97 helena Added memory allocation error check. 39 * 6/20/97 helena Java class name change. 40 * 6/23/97 helena Adding comments to make code more readable. 41 * 09/03/97 helena Added createCollationKeyValues(). 42 * 06/26/98 erm Changes for CollationKeys using byte arrays. 43 * 08/10/98 erm Synched with 1.2 version of RuleBasedCollator.java 44 * 04/23/99 stephen Removed EDecompositionMode, merged with 45 * Normalizer::EMode 46 * 06/14/99 stephen Removed kResourceBundleSuffix 47 * 06/22/99 stephen Fixed logic in constructFromFile() since .ctx 48 * files are no longer used. 49 * 11/02/99 helena Collator performance enhancements. Special case 50 * for NO_OP situations. 51 * 11/17/99 srl More performance enhancements. Inlined some internal functions. 52 * 12/15/99 aliu Update to support Thai collation. Move NormalizerIterator 53 * to implementation file. 54 * 01/29/01 synwee Modified into a C++ wrapper calling C APIs (ucol.h) 55 */ 56 57 #include "unicode/utypes.h" 58 59 #if !UCONFIG_NO_COLLATION 60 61 #include "unicode/tblcoll.h" 62 #include "unicode/coleitr.h" 63 #include "unicode/ures.h" 64 #include "unicode/uset.h" 65 #include "ucol_imp.h" 66 #include "uresimp.h" 67 #include "uhash.h" 68 #include "cmemory.h" 69 #include "cstring.h" 70 #include "putilimp.h" 71 #include "ustr_imp.h" 72 73 /* public RuleBasedCollator constructor ---------------------------------- */ 74 75 U_NAMESPACE_BEGIN 76 77 /** 78 * Copy constructor, aliasing, not write-through 79 */ 80 RuleBasedCollator::RuleBasedCollator(const RuleBasedCollator& that) 81 : Collator(that) 82 , dataIsOwned(FALSE) 83 , isWriteThroughAlias(FALSE) 84 , ucollator(NULL) 85 { 86 RuleBasedCollator::operator=(that); 87 } 88 89 RuleBasedCollator::RuleBasedCollator(const UnicodeString& rules, 90 UErrorCode& status) : 91 dataIsOwned(FALSE) 92 { 93 construct(rules, 94 UCOL_DEFAULT_STRENGTH, 95 UCOL_DEFAULT, 96 status); 97 } 98 99 RuleBasedCollator::RuleBasedCollator(const UnicodeString& rules, 100 ECollationStrength collationStrength, 101 UErrorCode& status) : dataIsOwned(FALSE) 102 { 103 construct(rules, 104 (UColAttributeValue)collationStrength, 105 UCOL_DEFAULT, 106 status); 107 } 108 109 RuleBasedCollator::RuleBasedCollator(const UnicodeString& rules, 110 UColAttributeValue decompositionMode, 111 UErrorCode& status) : 112 dataIsOwned(FALSE) 113 { 114 construct(rules, 115 UCOL_DEFAULT_STRENGTH, 116 decompositionMode, 117 status); 118 } 119 120 RuleBasedCollator::RuleBasedCollator(const UnicodeString& rules, 121 ECollationStrength collationStrength, 122 UColAttributeValue decompositionMode, 123 UErrorCode& status) : dataIsOwned(FALSE) 124 { 125 construct(rules, 126 (UColAttributeValue)collationStrength, 127 decompositionMode, 128 status); 129 } 130 RuleBasedCollator::RuleBasedCollator(const uint8_t *bin, int32_t length, 131 const RuleBasedCollator *base, 132 UErrorCode &status) : 133 dataIsOwned(TRUE), 134 isWriteThroughAlias(FALSE) 135 { 136 ucollator = ucol_openBinary(bin, length, base->ucollator, &status); 137 } 138 139 void 140 RuleBasedCollator::setRuleStringFromCollator() 141 { 142 int32_t length; 143 const UChar *r = ucol_getRules(ucollator, &length); 144 145 if (r && length > 0) { 146 // alias the rules string 147 urulestring.setTo(TRUE, r, length); 148 } 149 else { 150 urulestring.truncate(0); // Clear string. 151 } 152 } 153 154 // not aliasing, not write-through 155 void 156 RuleBasedCollator::construct(const UnicodeString& rules, 157 UColAttributeValue collationStrength, 158 UColAttributeValue decompositionMode, 159 UErrorCode& status) 160 { 161 ucollator = ucol_openRules(rules.getBuffer(), rules.length(), 162 decompositionMode, collationStrength, 163 NULL, &status); 164 165 dataIsOwned = TRUE; // since we own a collator now, we need to get rid of it 166 isWriteThroughAlias = FALSE; 167 168 if(ucollator == NULL) { 169 if(U_SUCCESS(status)) { 170 status = U_MEMORY_ALLOCATION_ERROR; 171 } 172 return; // Failure 173 } 174 175 setRuleStringFromCollator(); 176 } 177 178 /* RuleBasedCollator public destructor ----------------------------------- */ 179 180 RuleBasedCollator::~RuleBasedCollator() 181 { 182 if (dataIsOwned) 183 { 184 ucol_close(ucollator); 185 } 186 ucollator = 0; 187 } 188 189 /* RuleBaseCollator public methods --------------------------------------- */ 190 191 UBool RuleBasedCollator::operator==(const Collator& that) const 192 { 193 /* only checks for address equals here */ 194 if (this == &that) { 195 return TRUE; 196 } 197 if (!Collator::operator==(that)) { 198 return FALSE; /* not the same class */ 199 } 200 201 RuleBasedCollator& thatAlias = (RuleBasedCollator&)that; 202 203 return ucol_equals(this->ucollator, thatAlias.ucollator); 204 } 205 206 // aliasing, not write-through 207 RuleBasedCollator& RuleBasedCollator::operator=(const RuleBasedCollator& that) 208 { 209 if (this == &that) { return *this; } 210 211 UErrorCode intStatus = U_ZERO_ERROR; 212 int32_t buffersize = U_COL_SAFECLONE_BUFFERSIZE; 213 UCollator *ucol = ucol_safeClone(that.ucollator, NULL, &buffersize, &intStatus); 214 if (U_FAILURE(intStatus)) { return *this; } 215 216 if (dataIsOwned) { 217 ucol_close(ucollator); 218 } 219 ucollator = ucol; 220 dataIsOwned = TRUE; 221 isWriteThroughAlias = FALSE; 222 setRuleStringFromCollator(); 223 return *this; 224 } 225 226 // aliasing, not write-through 227 Collator* RuleBasedCollator::clone() const 228 { 229 RuleBasedCollator* coll = new RuleBasedCollator(*this); 230 // There is a small chance that the internal ucol_safeClone() call fails. 231 if (coll != NULL && coll->ucollator == NULL) { 232 delete coll; 233 return NULL; 234 } 235 return coll; 236 } 237 238 239 CollationElementIterator* RuleBasedCollator::createCollationElementIterator 240 (const UnicodeString& source) const 241 { 242 UErrorCode status = U_ZERO_ERROR; 243 CollationElementIterator *result = new CollationElementIterator(source, this, 244 status); 245 if (U_FAILURE(status)) { 246 delete result; 247 return NULL; 248 } 249 250 return result; 251 } 252 253 /** 254 * Create a CollationElementIterator object that will iterate over the 255 * elements in a string, using the collation rules defined in this 256 * RuleBasedCollator 257 */ 258 CollationElementIterator* RuleBasedCollator::createCollationElementIterator 259 (const CharacterIterator& source) const 260 { 261 UErrorCode status = U_ZERO_ERROR; 262 CollationElementIterator *result = new CollationElementIterator(source, this, 263 status); 264 265 if (U_FAILURE(status)) { 266 delete result; 267 return NULL; 268 } 269 270 return result; 271 } 272 273 /** 274 * Return a string representation of this collator's rules. The string can 275 * later be passed to the constructor that takes a UnicodeString argument, 276 * which will construct a collator that's functionally identical to this one. 277 * You can also allow users to edit the string in order to change the collation 278 * data, or you can print it out for inspection, or whatever. 279 */ 280 const UnicodeString& RuleBasedCollator::getRules() const 281 { 282 return urulestring; 283 } 284 285 void RuleBasedCollator::getRules(UColRuleOption delta, UnicodeString &buffer) 286 { 287 int32_t rulesize = ucol_getRulesEx(ucollator, delta, NULL, -1); 288 289 if (rulesize > 0) { 290 UChar *rules = (UChar*) uprv_malloc( sizeof(UChar) * (rulesize) ); 291 if(rules != NULL) { 292 ucol_getRulesEx(ucollator, delta, rules, rulesize); 293 buffer.setTo(rules, rulesize); 294 uprv_free(rules); 295 } else { // couldn't allocate 296 buffer.remove(); 297 } 298 } 299 else { 300 buffer.remove(); 301 } 302 } 303 304 UnicodeSet * 305 RuleBasedCollator::getTailoredSet(UErrorCode &status) const 306 { 307 if(U_FAILURE(status)) { 308 return NULL; 309 } 310 return (UnicodeSet *)ucol_getTailoredSet(this->ucollator, &status); 311 } 312 313 314 void RuleBasedCollator::getVersion(UVersionInfo versionInfo) const 315 { 316 if (versionInfo!=NULL){ 317 ucol_getVersion(ucollator, versionInfo); 318 } 319 } 320 321 /** 322 * Compare two strings using this collator 323 */ 324 UCollationResult RuleBasedCollator::compare( 325 const UnicodeString& source, 326 const UnicodeString& target, 327 int32_t length, 328 UErrorCode &status) const 329 { 330 return compare(source.getBuffer(), uprv_min(length,source.length()), target.getBuffer(), uprv_min(length,target.length()), status); 331 } 332 333 UCollationResult RuleBasedCollator::compare(const UChar* source, 334 int32_t sourceLength, 335 const UChar* target, 336 int32_t targetLength, 337 UErrorCode &status) const 338 { 339 if(U_SUCCESS(status)) { 340 return ucol_strcoll(ucollator, source, sourceLength, target, targetLength); 341 } else { 342 return UCOL_EQUAL; 343 } 344 } 345 346 UCollationResult RuleBasedCollator::compare( 347 const UnicodeString& source, 348 const UnicodeString& target, 349 UErrorCode &status) const 350 { 351 if(U_SUCCESS(status)) { 352 return ucol_strcoll(ucollator, source.getBuffer(), source.length(), 353 target.getBuffer(), target.length()); 354 } else { 355 return UCOL_EQUAL; 356 } 357 } 358 359 UCollationResult RuleBasedCollator::compare(UCharIterator &sIter, 360 UCharIterator &tIter, 361 UErrorCode &status) const { 362 if(U_SUCCESS(status)) { 363 return ucol_strcollIter(ucollator, &sIter, &tIter, &status); 364 } else { 365 return UCOL_EQUAL; 366 } 367 } 368 369 /** 370 * Retrieve a collation key for the specified string. The key can be compared 371 * with other collation keys using a bitwise comparison (e.g. memcmp) to find 372 * the ordering of their respective source strings. This is handy when doing a 373 * sort, where each sort key must be compared many times. 374 * 375 * The basic algorithm here is to find all of the collation elements for each 376 * character in the source string, convert them to an ASCII representation, and 377 * put them into the collation key. But it's trickier than that. Each 378 * collation element in a string has three components: primary ('A' vs 'B'), 379 * secondary ('u' vs '\u00FC'), and tertiary ('A' vs 'a'), and a primary difference 380 * at the end of a string takes precedence over a secondary or tertiary 381 * difference earlier in the string. 382 * 383 * To account for this, we put all of the primary orders at the beginning of 384 * the string, followed by the secondary and tertiary orders. Each set of 385 * orders is terminated by nulls so that a key for a string which is a initial 386 * substring of another key will compare less without any special case. 387 * 388 * Here's a hypothetical example, with the collation element represented as a 389 * three-digit number, one digit for primary, one for secondary, etc. 390 * 391 * String: A a B \u00C9 392 * Collation Elements: 101 100 201 511 393 * Collation Key: 1125<null>0001<null>1011<null> 394 * 395 * To make things even trickier, secondary differences (accent marks) are 396 * compared starting at the *end* of the string in languages with French 397 * secondary ordering. But when comparing the accent marks on a single base 398 * character, they are compared from the beginning. To handle this, we reverse 399 * all of the accents that belong to each base character, then we reverse the 400 * entire string of secondary orderings at the end. 401 */ 402 CollationKey& RuleBasedCollator::getCollationKey( 403 const UnicodeString& source, 404 CollationKey& sortkey, 405 UErrorCode& status) const 406 { 407 return getCollationKey(source.getBuffer(), source.length(), sortkey, status); 408 } 409 410 CollationKey& RuleBasedCollator::getCollationKey(const UChar* source, 411 int32_t sourceLen, 412 CollationKey& sortkey, 413 UErrorCode& status) const 414 { 415 if (U_FAILURE(status)) { 416 return sortkey.setToBogus(); 417 } 418 if (sourceLen < -1 || (source == NULL && sourceLen != 0)) { 419 status = U_ILLEGAL_ARGUMENT_ERROR; 420 return sortkey.setToBogus(); 421 } 422 423 if (sourceLen < 0) { 424 sourceLen = u_strlen(source); 425 } 426 if (sourceLen == 0) { 427 return sortkey.reset(); 428 } 429 430 int32_t resultLen = ucol_getCollationKey(ucollator, source, sourceLen, sortkey, status); 431 432 if (U_SUCCESS(status)) { 433 sortkey.setLength(resultLen); 434 } else { 435 sortkey.setToBogus(); 436 } 437 return sortkey; 438 } 439 440 /** 441 * Return the maximum length of any expansion sequences that end with the 442 * specified comparison order. 443 * @param order a collation order returned by previous or next. 444 * @return the maximum length of any expansion seuences ending with the 445 * specified order or 1 if collation order does not occur at the end of any 446 * expansion sequence. 447 * @see CollationElementIterator#getMaxExpansion 448 */ 449 int32_t RuleBasedCollator::getMaxExpansion(int32_t order) const 450 { 451 uint8_t result; 452 UCOL_GETMAXEXPANSION(ucollator, (uint32_t)order, result); 453 return result; 454 } 455 456 uint8_t* RuleBasedCollator::cloneRuleData(int32_t &length, 457 UErrorCode &status) 458 { 459 return ucol_cloneRuleData(ucollator, &length, &status); 460 } 461 462 463 int32_t RuleBasedCollator::cloneBinary(uint8_t *buffer, int32_t capacity, UErrorCode &status) 464 { 465 return ucol_cloneBinary(ucollator, buffer, capacity, &status); 466 } 467 468 void RuleBasedCollator::setAttribute(UColAttribute attr, 469 UColAttributeValue value, 470 UErrorCode &status) 471 { 472 if (U_FAILURE(status)) 473 return; 474 checkOwned(); 475 ucol_setAttribute(ucollator, attr, value, &status); 476 } 477 478 UColAttributeValue RuleBasedCollator::getAttribute(UColAttribute attr, 479 UErrorCode &status) const 480 { 481 if (U_FAILURE(status)) 482 return UCOL_DEFAULT; 483 return ucol_getAttribute(ucollator, attr, &status); 484 } 485 486 uint32_t RuleBasedCollator::setVariableTop(const UChar *varTop, int32_t len, UErrorCode &status) { 487 checkOwned(); 488 return ucol_setVariableTop(ucollator, varTop, len, &status); 489 } 490 491 uint32_t RuleBasedCollator::setVariableTop(const UnicodeString &varTop, UErrorCode &status) { 492 checkOwned(); 493 return ucol_setVariableTop(ucollator, varTop.getBuffer(), varTop.length(), &status); 494 } 495 496 void RuleBasedCollator::setVariableTop(uint32_t varTop, UErrorCode &status) { 497 checkOwned(); 498 ucol_restoreVariableTop(ucollator, varTop, &status); 499 } 500 501 uint32_t RuleBasedCollator::getVariableTop(UErrorCode &status) const { 502 return ucol_getVariableTop(ucollator, &status); 503 } 504 505 int32_t RuleBasedCollator::getSortKey(const UnicodeString& source, 506 uint8_t *result, int32_t resultLength) 507 const 508 { 509 return ucol_getSortKey(ucollator, source.getBuffer(), source.length(), result, resultLength); 510 } 511 512 int32_t RuleBasedCollator::getSortKey(const UChar *source, 513 int32_t sourceLength, uint8_t *result, 514 int32_t resultLength) const 515 { 516 return ucol_getSortKey(ucollator, source, sourceLength, result, resultLength); 517 } 518 519 int32_t RuleBasedCollator::getReorderCodes(int32_t *dest, 520 int32_t destCapacity, 521 UErrorCode& status) const 522 { 523 return ucol_getReorderCodes(ucollator, dest, destCapacity, &status); 524 } 525 526 void RuleBasedCollator::setReorderCodes(const int32_t *reorderCodes, 527 int32_t reorderCodesLength, 528 UErrorCode& status) 529 { 530 checkOwned(); 531 ucol_setReorderCodes(ucollator, reorderCodes, reorderCodesLength, &status); 532 } 533 534 int32_t RuleBasedCollator::getEquivalentReorderCodes(int32_t reorderCode, 535 int32_t* dest, 536 int32_t destCapacity, 537 UErrorCode& status) 538 { 539 return ucol_getEquivalentReorderCodes(reorderCode, dest, destCapacity, &status); 540 } 541 542 /** 543 * Create a hash code for this collation. Just hash the main rule table -- that 544 * should be good enough for almost any use. 545 */ 546 int32_t RuleBasedCollator::hashCode() const 547 { 548 int32_t length; 549 const UChar *rules = ucol_getRules(ucollator, &length); 550 return ustr_hashUCharsN(rules, length); 551 } 552 553 /** 554 * return the locale of this collator 555 */ 556 Locale RuleBasedCollator::getLocale(ULocDataLocaleType type, UErrorCode &status) const { 557 const char *result = ucol_getLocaleByType(ucollator, type, &status); 558 if(result == NULL) { 559 Locale res(""); 560 res.setToBogus(); 561 return res; 562 } else { 563 return Locale(result); 564 } 565 } 566 567 void 568 RuleBasedCollator::setLocales(const Locale& requestedLocale, const Locale& validLocale, const Locale& actualLocale) { 569 checkOwned(); 570 char* rloc = uprv_strdup(requestedLocale.getName()); 571 if (rloc) { 572 char* vloc = uprv_strdup(validLocale.getName()); 573 if (vloc) { 574 char* aloc = uprv_strdup(actualLocale.getName()); 575 if (aloc) { 576 ucol_setReqValidLocales(ucollator, rloc, vloc, aloc); 577 return; 578 } 579 uprv_free(vloc); 580 } 581 uprv_free(rloc); 582 } 583 } 584 585 // RuleBaseCollatorNew private constructor ---------------------------------- 586 587 RuleBasedCollator::RuleBasedCollator() 588 : dataIsOwned(FALSE), isWriteThroughAlias(FALSE), ucollator(NULL) 589 { 590 } 591 592 RuleBasedCollator::RuleBasedCollator(const Locale& desiredLocale, 593 UErrorCode& status) 594 : dataIsOwned(FALSE), isWriteThroughAlias(FALSE), ucollator(NULL) 595 { 596 if (U_FAILURE(status)) 597 return; 598 599 /* 600 Try to load, in order: 601 1. The desired locale's collation. 602 2. A fallback of the desired locale. 603 3. The default locale's collation. 604 4. A fallback of the default locale. 605 5. The default collation rules, which contains en_US collation rules. 606 607 To reiterate, we try: 608 Specific: 609 language+country+variant 610 language+country 611 language 612 Default: 613 language+country+variant 614 language+country 615 language 616 Root: (aka DEFAULTRULES) 617 steps 1-5 are handled by resource bundle fallback mechanism. 618 however, in a very unprobable situation that no resource bundle 619 data exists, step 5 is repeated with hardcoded default rules. 620 */ 621 622 setUCollator(desiredLocale, status); 623 624 if (U_FAILURE(status)) 625 { 626 status = U_ZERO_ERROR; 627 628 setUCollator(kRootLocaleName, status); 629 if (status == U_ZERO_ERROR) { 630 status = U_USING_DEFAULT_WARNING; 631 } 632 } 633 634 if (U_SUCCESS(status)) 635 { 636 setRuleStringFromCollator(); 637 } 638 } 639 640 void 641 RuleBasedCollator::setUCollator(const char *locale, 642 UErrorCode &status) 643 { 644 if (U_FAILURE(status)) { 645 return; 646 } 647 if (ucollator && dataIsOwned) 648 ucol_close(ucollator); 649 ucollator = ucol_open_internal(locale, &status); 650 dataIsOwned = TRUE; 651 isWriteThroughAlias = FALSE; 652 } 653 654 655 void 656 RuleBasedCollator::checkOwned() { 657 if (!(dataIsOwned || isWriteThroughAlias)) { 658 UErrorCode status = U_ZERO_ERROR; 659 ucollator = ucol_safeClone(ucollator, NULL, NULL, &status); 660 setRuleStringFromCollator(); 661 dataIsOwned = TRUE; 662 isWriteThroughAlias = FALSE; 663 } 664 } 665 666 667 int32_t RuleBasedCollator::internalGetShortDefinitionString(const char *locale, 668 char *buffer, 669 int32_t capacity, 670 UErrorCode &status) const { 671 /* simply delegate */ 672 return ucol_getShortDefinitionString(ucollator, locale, buffer, capacity, &status); 673 } 674 675 676 UOBJECT_DEFINE_RTTI_IMPLEMENTATION(RuleBasedCollator) 677 678 U_NAMESPACE_END 679 680 #endif /* #if !UCONFIG_NO_COLLATION */ 681