1// © 2016 and later: Unicode, Inc. and others. 2// License & terms of use: http://www.unicode.org/copyright.html 3/* 4******************************************************************************* 5* 6* Copyright (C) 1999-2014, International Business Machines 7* Corporation and others. All Rights Reserved. 8* 9******************************************************************************* 10* file name: uniset_props.cpp 11* encoding: UTF-8 12* tab size: 8 (not used) 13* indentation:4 14* 15* created on: 2004aug25 16* created by: Markus W. Scherer 17* 18* Character property dependent functions moved here from uniset.cpp 19*/ 20 21#include "unicode/utypes.h" 22#include "unicode/uniset.h" 23#include "unicode/parsepos.h" 24#include "unicode/uchar.h" 25#include "unicode/uscript.h" 26#include "unicode/symtable.h" 27#include "unicode/uset.h" 28#include "unicode/locid.h" 29#include "unicode/brkiter.h" 30#include "uset_imp.h" 31#include "ruleiter.h" 32#include "cmemory.h" 33#include "ucln_cmn.h" 34#include "util.h" 35#include "uvector.h" 36#include "uprops.h" 37#include "propname.h" 38#include "normalizer2impl.h" 39#include "ucase.h" 40#include "ubidi_props.h" 41#include "uinvchar.h" 42#include "uprops.h" 43#include "charstr.h" 44#include "cstring.h" 45#include "mutex.h" 46#include "umutex.h" 47#include "uassert.h" 48#include "hash.h" 49 50U_NAMESPACE_USE 51 52// initial storage. Must be >= 0 53// *** same as in uniset.cpp ! *** 54#define START_EXTRA 16 55 56// Define UChar constants using hex for EBCDIC compatibility 57// Used #define to reduce private static exports and memory access time. 58#define SET_OPEN ((UChar)0x005B) /*[*/ 59#define SET_CLOSE ((UChar)0x005D) /*]*/ 60#define HYPHEN ((UChar)0x002D) /*-*/ 61#define COMPLEMENT ((UChar)0x005E) /*^*/ 62#define COLON ((UChar)0x003A) /*:*/ 63#define BACKSLASH ((UChar)0x005C) /*\*/ 64#define INTERSECTION ((UChar)0x0026) /*&*/ 65#define UPPER_U ((UChar)0x0055) /*U*/ 66#define LOWER_U ((UChar)0x0075) /*u*/ 67#define OPEN_BRACE ((UChar)123) /*{*/ 68#define CLOSE_BRACE ((UChar)125) /*}*/ 69#define UPPER_P ((UChar)0x0050) /*P*/ 70#define LOWER_P ((UChar)0x0070) /*p*/ 71#define UPPER_N ((UChar)78) /*N*/ 72#define EQUALS ((UChar)0x003D) /*=*/ 73 74//static const UChar POSIX_OPEN[] = { SET_OPEN,COLON,0 }; // "[:" 75static const UChar POSIX_CLOSE[] = { COLON,SET_CLOSE,0 }; // ":]" 76//static const UChar PERL_OPEN[] = { BACKSLASH,LOWER_P,0 }; // "\\p" 77//static const UChar PERL_CLOSE[] = { CLOSE_BRACE,0 }; // "}" 78//static const UChar NAME_OPEN[] = { BACKSLASH,UPPER_N,0 }; // "\\N" 79static const UChar HYPHEN_RIGHT_BRACE[] = {HYPHEN,SET_CLOSE,0}; /*-]*/ 80 81// Special property set IDs 82static const char ANY[] = "ANY"; // [\u0000-\U0010FFFF] 83static const char ASCII[] = "ASCII"; // [\u0000-\u007F] 84static const char ASSIGNED[] = "Assigned"; // [:^Cn:] 85 86// Unicode name property alias 87#define NAME_PROP "na" 88#define NAME_PROP_LENGTH 2 89 90/** 91 * Delimiter string used in patterns to close a category reference: 92 * ":]". Example: "[:Lu:]". 93 */ 94//static const UChar CATEGORY_CLOSE[] = {COLON, SET_CLOSE, 0x0000}; /* ":]" */ 95 96// Cached sets ------------------------------------------------------------- *** 97 98U_CDECL_BEGIN 99static UBool U_CALLCONV uset_cleanup(); 100 101struct Inclusion { 102 UnicodeSet *fSet; 103 UInitOnce fInitOnce; 104}; 105static Inclusion gInclusions[UPROPS_SRC_COUNT]; // cached getInclusions() 106 107static UnicodeSet *uni32Singleton; 108static icu::UInitOnce uni32InitOnce = U_INITONCE_INITIALIZER; 109 110//---------------------------------------------------------------- 111// Inclusions list 112//---------------------------------------------------------------- 113 114// USetAdder implementation 115// Does not use uset.h to reduce code dependencies 116static void U_CALLCONV 117_set_add(USet *set, UChar32 c) { 118 ((UnicodeSet *)set)->add(c); 119} 120 121static void U_CALLCONV 122_set_addRange(USet *set, UChar32 start, UChar32 end) { 123 ((UnicodeSet *)set)->add(start, end); 124} 125 126static void U_CALLCONV 127_set_addString(USet *set, const UChar *str, int32_t length) { 128 ((UnicodeSet *)set)->add(UnicodeString((UBool)(length<0), str, length)); 129} 130 131/** 132 * Cleanup function for UnicodeSet 133 */ 134static UBool U_CALLCONV uset_cleanup(void) { 135 for(int32_t i = UPROPS_SRC_NONE; i < UPROPS_SRC_COUNT; ++i) { 136 Inclusion &in = gInclusions[i]; 137 delete in.fSet; 138 in.fSet = NULL; 139 in.fInitOnce.reset(); 140 } 141 142 delete uni32Singleton; 143 uni32Singleton = NULL; 144 uni32InitOnce.reset(); 145 return TRUE; 146} 147 148U_CDECL_END 149 150U_NAMESPACE_BEGIN 151 152/* 153Reduce excessive reallocation, and make it easier to detect initialization problems. 154Usually you don't see smaller sets than this for Unicode 5.0. 155*/ 156#define DEFAULT_INCLUSION_CAPACITY 3072 157 158void U_CALLCONV UnicodeSet_initInclusion(int32_t src, UErrorCode &status) { 159 // This function is invoked only via umtx_initOnce(). 160 // This function is a friend of class UnicodeSet. 161 162 U_ASSERT(src >=0 && src<UPROPS_SRC_COUNT); 163 UnicodeSet * &incl = gInclusions[src].fSet; 164 U_ASSERT(incl == NULL); 165 166 incl = new UnicodeSet(); 167 if (incl == NULL) { 168 status = U_MEMORY_ALLOCATION_ERROR; 169 return; 170 } 171 USetAdder sa = { 172 (USet *)incl, 173 _set_add, 174 _set_addRange, 175 _set_addString, 176 NULL, // don't need remove() 177 NULL // don't need removeRange() 178 }; 179 180 incl->ensureCapacity(DEFAULT_INCLUSION_CAPACITY, status); 181 switch(src) { 182 case UPROPS_SRC_CHAR: 183 uchar_addPropertyStarts(&sa, &status); 184 break; 185 case UPROPS_SRC_PROPSVEC: 186 upropsvec_addPropertyStarts(&sa, &status); 187 break; 188 case UPROPS_SRC_CHAR_AND_PROPSVEC: 189 uchar_addPropertyStarts(&sa, &status); 190 upropsvec_addPropertyStarts(&sa, &status); 191 break; 192#if !UCONFIG_NO_NORMALIZATION 193 case UPROPS_SRC_CASE_AND_NORM: { 194 const Normalizer2Impl *impl=Normalizer2Factory::getNFCImpl(status); 195 if(U_SUCCESS(status)) { 196 impl->addPropertyStarts(&sa, status); 197 } 198 ucase_addPropertyStarts(&sa, &status); 199 break; 200 } 201 case UPROPS_SRC_NFC: { 202 const Normalizer2Impl *impl=Normalizer2Factory::getNFCImpl(status); 203 if(U_SUCCESS(status)) { 204 impl->addPropertyStarts(&sa, status); 205 } 206 break; 207 } 208 case UPROPS_SRC_NFKC: { 209 const Normalizer2Impl *impl=Normalizer2Factory::getNFKCImpl(status); 210 if(U_SUCCESS(status)) { 211 impl->addPropertyStarts(&sa, status); 212 } 213 break; 214 } 215 case UPROPS_SRC_NFKC_CF: { 216 const Normalizer2Impl *impl=Normalizer2Factory::getNFKC_CFImpl(status); 217 if(U_SUCCESS(status)) { 218 impl->addPropertyStarts(&sa, status); 219 } 220 break; 221 } 222 case UPROPS_SRC_NFC_CANON_ITER: { 223 const Normalizer2Impl *impl=Normalizer2Factory::getNFCImpl(status); 224 if(U_SUCCESS(status)) { 225 impl->addCanonIterPropertyStarts(&sa, status); 226 } 227 break; 228 } 229#endif 230 case UPROPS_SRC_CASE: 231 ucase_addPropertyStarts(&sa, &status); 232 break; 233 case UPROPS_SRC_BIDI: 234 ubidi_addPropertyStarts(ubidi_getSingleton(), &sa, &status); 235 break; 236 default: 237 status = U_INTERNAL_PROGRAM_ERROR; 238 break; 239 } 240 241 if (U_FAILURE(status)) { 242 delete incl; 243 incl = NULL; 244 return; 245 } 246 // Compact for caching 247 incl->compact(); 248 ucln_common_registerCleanup(UCLN_COMMON_USET, uset_cleanup); 249} 250 251 252 253const UnicodeSet* UnicodeSet::getInclusions(int32_t src, UErrorCode &status) { 254 U_ASSERT(src >=0 && src<UPROPS_SRC_COUNT); 255 Inclusion &i = gInclusions[src]; 256 umtx_initOnce(i.fInitOnce, &UnicodeSet_initInclusion, src, status); 257 return i.fSet; 258} 259 260 261// Cache some sets for other services -------------------------------------- *** 262void U_CALLCONV createUni32Set(UErrorCode &errorCode) { 263 U_ASSERT(uni32Singleton == NULL); 264 uni32Singleton = new UnicodeSet(UNICODE_STRING_SIMPLE("[:age=3.2:]"), errorCode); 265 if(uni32Singleton==NULL) { 266 errorCode=U_MEMORY_ALLOCATION_ERROR; 267 } else { 268 uni32Singleton->freeze(); 269 } 270 ucln_common_registerCleanup(UCLN_COMMON_USET, uset_cleanup); 271} 272 273 274U_CFUNC UnicodeSet * 275uniset_getUnicode32Instance(UErrorCode &errorCode) { 276 umtx_initOnce(uni32InitOnce, &createUni32Set, errorCode); 277 return uni32Singleton; 278} 279 280// helper functions for matching of pattern syntax pieces ------------------ *** 281// these functions are parallel to the PERL_OPEN etc. strings above 282 283// using these functions is not only faster than UnicodeString::compare() and 284// caseCompare(), but they also make UnicodeSet work for simple patterns when 285// no Unicode properties data is available - when caseCompare() fails 286 287static inline UBool 288isPerlOpen(const UnicodeString &pattern, int32_t pos) { 289 UChar c; 290 return pattern.charAt(pos)==BACKSLASH && ((c=pattern.charAt(pos+1))==LOWER_P || c==UPPER_P); 291} 292 293/*static inline UBool 294isPerlClose(const UnicodeString &pattern, int32_t pos) { 295 return pattern.charAt(pos)==CLOSE_BRACE; 296}*/ 297 298static inline UBool 299isNameOpen(const UnicodeString &pattern, int32_t pos) { 300 return pattern.charAt(pos)==BACKSLASH && pattern.charAt(pos+1)==UPPER_N; 301} 302 303static inline UBool 304isPOSIXOpen(const UnicodeString &pattern, int32_t pos) { 305 return pattern.charAt(pos)==SET_OPEN && pattern.charAt(pos+1)==COLON; 306} 307 308/*static inline UBool 309isPOSIXClose(const UnicodeString &pattern, int32_t pos) { 310 return pattern.charAt(pos)==COLON && pattern.charAt(pos+1)==SET_CLOSE; 311}*/ 312 313// TODO memory debugging provided inside uniset.cpp 314// could be made available here but probably obsolete with use of modern 315// memory leak checker tools 316#define _dbgct(me) 317 318//---------------------------------------------------------------- 319// Constructors &c 320//---------------------------------------------------------------- 321 322/** 323 * Constructs a set from the given pattern, optionally ignoring 324 * white space. See the class description for the syntax of the 325 * pattern language. 326 * @param pattern a string specifying what characters are in the set 327 */ 328UnicodeSet::UnicodeSet(const UnicodeString& pattern, 329 UErrorCode& status) : 330 len(0), capacity(START_EXTRA), list(0), bmpSet(0), buffer(0), 331 bufferCapacity(0), patLen(0), pat(NULL), strings(NULL), stringSpan(NULL), 332 fFlags(0) 333{ 334 if(U_SUCCESS(status)){ 335 list = (UChar32*) uprv_malloc(sizeof(UChar32) * capacity); 336 /* test for NULL */ 337 if(list == NULL) { 338 status = U_MEMORY_ALLOCATION_ERROR; 339 }else{ 340 allocateStrings(status); 341 applyPattern(pattern, status); 342 } 343 } 344 _dbgct(this); 345} 346 347//---------------------------------------------------------------- 348// Public API 349//---------------------------------------------------------------- 350 351UnicodeSet& UnicodeSet::applyPattern(const UnicodeString& pattern, 352 UErrorCode& status) { 353 // Equivalent to 354 // return applyPattern(pattern, USET_IGNORE_SPACE, NULL, status); 355 // but without dependency on closeOver(). 356 ParsePosition pos(0); 357 applyPatternIgnoreSpace(pattern, pos, NULL, status); 358 if (U_FAILURE(status)) return *this; 359 360 int32_t i = pos.getIndex(); 361 // Skip over trailing whitespace 362 ICU_Utility::skipWhitespace(pattern, i, TRUE); 363 if (i != pattern.length()) { 364 status = U_ILLEGAL_ARGUMENT_ERROR; 365 } 366 return *this; 367} 368 369void 370UnicodeSet::applyPatternIgnoreSpace(const UnicodeString& pattern, 371 ParsePosition& pos, 372 const SymbolTable* symbols, 373 UErrorCode& status) { 374 if (U_FAILURE(status)) { 375 return; 376 } 377 if (isFrozen()) { 378 status = U_NO_WRITE_PERMISSION; 379 return; 380 } 381 // Need to build the pattern in a temporary string because 382 // _applyPattern calls add() etc., which set pat to empty. 383 UnicodeString rebuiltPat; 384 RuleCharacterIterator chars(pattern, symbols, pos); 385 applyPattern(chars, symbols, rebuiltPat, USET_IGNORE_SPACE, NULL, status); 386 if (U_FAILURE(status)) return; 387 if (chars.inVariable()) { 388 // syntaxError(chars, "Extra chars in variable value"); 389 status = U_MALFORMED_SET; 390 return; 391 } 392 setPattern(rebuiltPat); 393} 394 395/** 396 * Return true if the given position, in the given pattern, appears 397 * to be the start of a UnicodeSet pattern. 398 */ 399UBool UnicodeSet::resemblesPattern(const UnicodeString& pattern, int32_t pos) { 400 return ((pos+1) < pattern.length() && 401 pattern.charAt(pos) == (UChar)91/*[*/) || 402 resemblesPropertyPattern(pattern, pos); 403} 404 405//---------------------------------------------------------------- 406// Implementation: Pattern parsing 407//---------------------------------------------------------------- 408 409/** 410 * A small all-inline class to manage a UnicodeSet pointer. Add 411 * operator->() etc. as needed. 412 */ 413class UnicodeSetPointer { 414 UnicodeSet* p; 415public: 416 inline UnicodeSetPointer() : p(0) {} 417 inline ~UnicodeSetPointer() { delete p; } 418 inline UnicodeSet* pointer() { return p; } 419 inline UBool allocate() { 420 if (p == 0) { 421 p = new UnicodeSet(); 422 } 423 return p != 0; 424 } 425}; 426 427/** 428 * Parse the pattern from the given RuleCharacterIterator. The 429 * iterator is advanced over the parsed pattern. 430 * @param chars iterator over the pattern characters. Upon return 431 * it will be advanced to the first character after the parsed 432 * pattern, or the end of the iteration if all characters are 433 * parsed. 434 * @param symbols symbol table to use to parse and dereference 435 * variables, or null if none. 436 * @param rebuiltPat the pattern that was parsed, rebuilt or 437 * copied from the input pattern, as appropriate. 438 * @param options a bit mask of zero or more of the following: 439 * IGNORE_SPACE, CASE. 440 */ 441void UnicodeSet::applyPattern(RuleCharacterIterator& chars, 442 const SymbolTable* symbols, 443 UnicodeString& rebuiltPat, 444 uint32_t options, 445 UnicodeSet& (UnicodeSet::*caseClosure)(int32_t attribute), 446 UErrorCode& ec) { 447 if (U_FAILURE(ec)) return; 448 449 // Syntax characters: [ ] ^ - & { } 450 451 // Recognized special forms for chars, sets: c-c s-s s&s 452 453 int32_t opts = RuleCharacterIterator::PARSE_VARIABLES | 454 RuleCharacterIterator::PARSE_ESCAPES; 455 if ((options & USET_IGNORE_SPACE) != 0) { 456 opts |= RuleCharacterIterator::SKIP_WHITESPACE; 457 } 458 459 UnicodeString patLocal, buf; 460 UBool usePat = FALSE; 461 UnicodeSetPointer scratch; 462 RuleCharacterIterator::Pos backup; 463 464 // mode: 0=before [, 1=between [...], 2=after ] 465 // lastItem: 0=none, 1=char, 2=set 466 int8_t lastItem = 0, mode = 0; 467 UChar32 lastChar = 0; 468 UChar op = 0; 469 470 UBool invert = FALSE; 471 472 clear(); 473 474 while (mode != 2 && !chars.atEnd()) { 475 U_ASSERT((lastItem == 0 && op == 0) || 476 (lastItem == 1 && (op == 0 || op == HYPHEN /*'-'*/)) || 477 (lastItem == 2 && (op == 0 || op == HYPHEN /*'-'*/ || 478 op == INTERSECTION /*'&'*/))); 479 480 UChar32 c = 0; 481 UBool literal = FALSE; 482 UnicodeSet* nested = 0; // alias - do not delete 483 484 // -------- Check for property pattern 485 486 // setMode: 0=none, 1=unicodeset, 2=propertypat, 3=preparsed 487 int8_t setMode = 0; 488 if (resemblesPropertyPattern(chars, opts)) { 489 setMode = 2; 490 } 491 492 // -------- Parse '[' of opening delimiter OR nested set. 493 // If there is a nested set, use `setMode' to define how 494 // the set should be parsed. If the '[' is part of the 495 // opening delimiter for this pattern, parse special 496 // strings "[", "[^", "[-", and "[^-". Check for stand-in 497 // characters representing a nested set in the symbol 498 // table. 499 500 else { 501 // Prepare to backup if necessary 502 chars.getPos(backup); 503 c = chars.next(opts, literal, ec); 504 if (U_FAILURE(ec)) return; 505 506 if (c == 0x5B /*'['*/ && !literal) { 507 if (mode == 1) { 508 chars.setPos(backup); // backup 509 setMode = 1; 510 } else { 511 // Handle opening '[' delimiter 512 mode = 1; 513 patLocal.append((UChar) 0x5B /*'['*/); 514 chars.getPos(backup); // prepare to backup 515 c = chars.next(opts, literal, ec); 516 if (U_FAILURE(ec)) return; 517 if (c == 0x5E /*'^'*/ && !literal) { 518 invert = TRUE; 519 patLocal.append((UChar) 0x5E /*'^'*/); 520 chars.getPos(backup); // prepare to backup 521 c = chars.next(opts, literal, ec); 522 if (U_FAILURE(ec)) return; 523 } 524 // Fall through to handle special leading '-'; 525 // otherwise restart loop for nested [], \p{}, etc. 526 if (c == HYPHEN /*'-'*/) { 527 literal = TRUE; 528 // Fall through to handle literal '-' below 529 } else { 530 chars.setPos(backup); // backup 531 continue; 532 } 533 } 534 } else if (symbols != 0) { 535 const UnicodeFunctor *m = symbols->lookupMatcher(c); 536 if (m != 0) { 537 const UnicodeSet *ms = dynamic_cast<const UnicodeSet *>(m); 538 if (ms == NULL) { 539 ec = U_MALFORMED_SET; 540 return; 541 } 542 // casting away const, but `nested' won't be modified 543 // (important not to modify stored set) 544 nested = const_cast<UnicodeSet*>(ms); 545 setMode = 3; 546 } 547 } 548 } 549 550 // -------- Handle a nested set. This either is inline in 551 // the pattern or represented by a stand-in that has 552 // previously been parsed and was looked up in the symbol 553 // table. 554 555 if (setMode != 0) { 556 if (lastItem == 1) { 557 if (op != 0) { 558 // syntaxError(chars, "Char expected after operator"); 559 ec = U_MALFORMED_SET; 560 return; 561 } 562 add(lastChar, lastChar); 563 _appendToPat(patLocal, lastChar, FALSE); 564 lastItem = 0; 565 op = 0; 566 } 567 568 if (op == HYPHEN /*'-'*/ || op == INTERSECTION /*'&'*/) { 569 patLocal.append(op); 570 } 571 572 if (nested == 0) { 573 // lazy allocation 574 if (!scratch.allocate()) { 575 ec = U_MEMORY_ALLOCATION_ERROR; 576 return; 577 } 578 nested = scratch.pointer(); 579 } 580 switch (setMode) { 581 case 1: 582 nested->applyPattern(chars, symbols, patLocal, options, caseClosure, ec); 583 break; 584 case 2: 585 chars.skipIgnored(opts); 586 nested->applyPropertyPattern(chars, patLocal, ec); 587 if (U_FAILURE(ec)) return; 588 break; 589 case 3: // `nested' already parsed 590 nested->_toPattern(patLocal, FALSE); 591 break; 592 } 593 594 usePat = TRUE; 595 596 if (mode == 0) { 597 // Entire pattern is a category; leave parse loop 598 *this = *nested; 599 mode = 2; 600 break; 601 } 602 603 switch (op) { 604 case HYPHEN: /*'-'*/ 605 removeAll(*nested); 606 break; 607 case INTERSECTION: /*'&'*/ 608 retainAll(*nested); 609 break; 610 case 0: 611 addAll(*nested); 612 break; 613 } 614 615 op = 0; 616 lastItem = 2; 617 618 continue; 619 } 620 621 if (mode == 0) { 622 // syntaxError(chars, "Missing '['"); 623 ec = U_MALFORMED_SET; 624 return; 625 } 626 627 // -------- Parse special (syntax) characters. If the 628 // current character is not special, or if it is escaped, 629 // then fall through and handle it below. 630 631 if (!literal) { 632 switch (c) { 633 case 0x5D /*']'*/: 634 if (lastItem == 1) { 635 add(lastChar, lastChar); 636 _appendToPat(patLocal, lastChar, FALSE); 637 } 638 // Treat final trailing '-' as a literal 639 if (op == HYPHEN /*'-'*/) { 640 add(op, op); 641 patLocal.append(op); 642 } else if (op == INTERSECTION /*'&'*/) { 643 // syntaxError(chars, "Trailing '&'"); 644 ec = U_MALFORMED_SET; 645 return; 646 } 647 patLocal.append((UChar) 0x5D /*']'*/); 648 mode = 2; 649 continue; 650 case HYPHEN /*'-'*/: 651 if (op == 0) { 652 if (lastItem != 0) { 653 op = (UChar) c; 654 continue; 655 } else { 656 // Treat final trailing '-' as a literal 657 add(c, c); 658 c = chars.next(opts, literal, ec); 659 if (U_FAILURE(ec)) return; 660 if (c == 0x5D /*']'*/ && !literal) { 661 patLocal.append(HYPHEN_RIGHT_BRACE, 2); 662 mode = 2; 663 continue; 664 } 665 } 666 } 667 // syntaxError(chars, "'-' not after char or set"); 668 ec = U_MALFORMED_SET; 669 return; 670 case INTERSECTION /*'&'*/: 671 if (lastItem == 2 && op == 0) { 672 op = (UChar) c; 673 continue; 674 } 675 // syntaxError(chars, "'&' not after set"); 676 ec = U_MALFORMED_SET; 677 return; 678 case 0x5E /*'^'*/: 679 // syntaxError(chars, "'^' not after '['"); 680 ec = U_MALFORMED_SET; 681 return; 682 case 0x7B /*'{'*/: 683 if (op != 0) { 684 // syntaxError(chars, "Missing operand after operator"); 685 ec = U_MALFORMED_SET; 686 return; 687 } 688 if (lastItem == 1) { 689 add(lastChar, lastChar); 690 _appendToPat(patLocal, lastChar, FALSE); 691 } 692 lastItem = 0; 693 buf.truncate(0); 694 { 695 UBool ok = FALSE; 696 while (!chars.atEnd()) { 697 c = chars.next(opts, literal, ec); 698 if (U_FAILURE(ec)) return; 699 if (c == 0x7D /*'}'*/ && !literal) { 700 ok = TRUE; 701 break; 702 } 703 buf.append(c); 704 } 705 if (buf.length() < 1 || !ok) { 706 // syntaxError(chars, "Invalid multicharacter string"); 707 ec = U_MALFORMED_SET; 708 return; 709 } 710 } 711 // We have new string. Add it to set and continue; 712 // we don't need to drop through to the further 713 // processing 714 add(buf); 715 patLocal.append((UChar) 0x7B /*'{'*/); 716 _appendToPat(patLocal, buf, FALSE); 717 patLocal.append((UChar) 0x7D /*'}'*/); 718 continue; 719 case SymbolTable::SYMBOL_REF: 720 // symbols nosymbols 721 // [a-$] error error (ambiguous) 722 // [a$] anchor anchor 723 // [a-$x] var "x"* literal '$' 724 // [a-$.] error literal '$' 725 // *We won't get here in the case of var "x" 726 { 727 chars.getPos(backup); 728 c = chars.next(opts, literal, ec); 729 if (U_FAILURE(ec)) return; 730 UBool anchor = (c == 0x5D /*']'*/ && !literal); 731 if (symbols == 0 && !anchor) { 732 c = SymbolTable::SYMBOL_REF; 733 chars.setPos(backup); 734 break; // literal '$' 735 } 736 if (anchor && op == 0) { 737 if (lastItem == 1) { 738 add(lastChar, lastChar); 739 _appendToPat(patLocal, lastChar, FALSE); 740 } 741 add(U_ETHER); 742 usePat = TRUE; 743 patLocal.append((UChar) SymbolTable::SYMBOL_REF); 744 patLocal.append((UChar) 0x5D /*']'*/); 745 mode = 2; 746 continue; 747 } 748 // syntaxError(chars, "Unquoted '$'"); 749 ec = U_MALFORMED_SET; 750 return; 751 } 752 default: 753 break; 754 } 755 } 756 757 // -------- Parse literal characters. This includes both 758 // escaped chars ("\u4E01") and non-syntax characters 759 // ("a"). 760 761 switch (lastItem) { 762 case 0: 763 lastItem = 1; 764 lastChar = c; 765 break; 766 case 1: 767 if (op == HYPHEN /*'-'*/) { 768 if (lastChar >= c) { 769 // Don't allow redundant (a-a) or empty (b-a) ranges; 770 // these are most likely typos. 771 // syntaxError(chars, "Invalid range"); 772 ec = U_MALFORMED_SET; 773 return; 774 } 775 add(lastChar, c); 776 _appendToPat(patLocal, lastChar, FALSE); 777 patLocal.append(op); 778 _appendToPat(patLocal, c, FALSE); 779 lastItem = 0; 780 op = 0; 781 } else { 782 add(lastChar, lastChar); 783 _appendToPat(patLocal, lastChar, FALSE); 784 lastChar = c; 785 } 786 break; 787 case 2: 788 if (op != 0) { 789 // syntaxError(chars, "Set expected after operator"); 790 ec = U_MALFORMED_SET; 791 return; 792 } 793 lastChar = c; 794 lastItem = 1; 795 break; 796 } 797 } 798 799 if (mode != 2) { 800 // syntaxError(chars, "Missing ']'"); 801 ec = U_MALFORMED_SET; 802 return; 803 } 804 805 chars.skipIgnored(opts); 806 807 /** 808 * Handle global flags (invert, case insensitivity). If this 809 * pattern should be compiled case-insensitive, then we need 810 * to close over case BEFORE COMPLEMENTING. This makes 811 * patterns like /[^abc]/i work. 812 */ 813 if ((options & USET_CASE_INSENSITIVE) != 0) { 814 (this->*caseClosure)(USET_CASE_INSENSITIVE); 815 } 816 else if ((options & USET_ADD_CASE_MAPPINGS) != 0) { 817 (this->*caseClosure)(USET_ADD_CASE_MAPPINGS); 818 } 819 if (invert) { 820 complement(); 821 } 822 823 // Use the rebuilt pattern (patLocal) only if necessary. Prefer the 824 // generated pattern. 825 if (usePat) { 826 rebuiltPat.append(patLocal); 827 } else { 828 _generatePattern(rebuiltPat, FALSE); 829 } 830 if (isBogus() && U_SUCCESS(ec)) { 831 // We likely ran out of memory. AHHH! 832 ec = U_MEMORY_ALLOCATION_ERROR; 833 } 834} 835 836//---------------------------------------------------------------- 837// Property set implementation 838//---------------------------------------------------------------- 839 840static UBool numericValueFilter(UChar32 ch, void* context) { 841 return u_getNumericValue(ch) == *(double*)context; 842} 843 844static UBool generalCategoryMaskFilter(UChar32 ch, void* context) { 845 int32_t value = *(int32_t*)context; 846 return (U_GET_GC_MASK((UChar32) ch) & value) != 0; 847} 848 849static UBool versionFilter(UChar32 ch, void* context) { 850 static const UVersionInfo none = { 0, 0, 0, 0 }; 851 UVersionInfo v; 852 u_charAge(ch, v); 853 UVersionInfo* version = (UVersionInfo*)context; 854 return uprv_memcmp(&v, &none, sizeof(v)) > 0 && uprv_memcmp(&v, version, sizeof(v)) <= 0; 855} 856 857typedef struct { 858 UProperty prop; 859 int32_t value; 860} IntPropertyContext; 861 862static UBool intPropertyFilter(UChar32 ch, void* context) { 863 IntPropertyContext* c = (IntPropertyContext*)context; 864 return u_getIntPropertyValue((UChar32) ch, c->prop) == c->value; 865} 866 867static UBool scriptExtensionsFilter(UChar32 ch, void* context) { 868 return uscript_hasScript(ch, *(UScriptCode*)context); 869} 870 871/** 872 * Generic filter-based scanning code for UCD property UnicodeSets. 873 */ 874void UnicodeSet::applyFilter(UnicodeSet::Filter filter, 875 void* context, 876 int32_t src, 877 UErrorCode &status) { 878 if (U_FAILURE(status)) return; 879 880 // Logically, walk through all Unicode characters, noting the start 881 // and end of each range for which filter.contain(c) is 882 // true. Add each range to a set. 883 // 884 // To improve performance, use an inclusions set which 885 // encodes information about character ranges that are known 886 // to have identical properties. 887 // getInclusions(src) contains exactly the first characters of 888 // same-value ranges for the given properties "source". 889 const UnicodeSet* inclusions = getInclusions(src, status); 890 if (U_FAILURE(status)) { 891 return; 892 } 893 894 clear(); 895 896 UChar32 startHasProperty = -1; 897 int32_t limitRange = inclusions->getRangeCount(); 898 899 for (int j=0; j<limitRange; ++j) { 900 // get current range 901 UChar32 start = inclusions->getRangeStart(j); 902 UChar32 end = inclusions->getRangeEnd(j); 903 904 // for all the code points in the range, process 905 for (UChar32 ch = start; ch <= end; ++ch) { 906 // only add to this UnicodeSet on inflection points -- 907 // where the hasProperty value changes to false 908 if ((*filter)(ch, context)) { 909 if (startHasProperty < 0) { 910 startHasProperty = ch; 911 } 912 } else if (startHasProperty >= 0) { 913 add(startHasProperty, ch-1); 914 startHasProperty = -1; 915 } 916 } 917 } 918 if (startHasProperty >= 0) { 919 add((UChar32)startHasProperty, (UChar32)0x10FFFF); 920 } 921 if (isBogus() && U_SUCCESS(status)) { 922 // We likely ran out of memory. AHHH! 923 status = U_MEMORY_ALLOCATION_ERROR; 924 } 925} 926 927static UBool mungeCharName(char* dst, const char* src, int32_t dstCapacity) { 928 /* Note: we use ' ' in compiler code page */ 929 int32_t j = 0; 930 char ch; 931 --dstCapacity; /* make room for term. zero */ 932 while ((ch = *src++) != 0) { 933 if (ch == ' ' && (j==0 || (j>0 && dst[j-1]==' '))) { 934 continue; 935 } 936 if (j >= dstCapacity) return FALSE; 937 dst[j++] = ch; 938 } 939 if (j > 0 && dst[j-1] == ' ') --j; 940 dst[j] = 0; 941 return TRUE; 942} 943 944//---------------------------------------------------------------- 945// Property set API 946//---------------------------------------------------------------- 947 948#define FAIL(ec) {ec=U_ILLEGAL_ARGUMENT_ERROR; return *this;} 949 950UnicodeSet& 951UnicodeSet::applyIntPropertyValue(UProperty prop, int32_t value, UErrorCode& ec) { 952 if (U_FAILURE(ec) || isFrozen()) return *this; 953 954 if (prop == UCHAR_GENERAL_CATEGORY_MASK) { 955 applyFilter(generalCategoryMaskFilter, &value, UPROPS_SRC_CHAR, ec); 956 } else if (prop == UCHAR_SCRIPT_EXTENSIONS) { 957 UScriptCode script = (UScriptCode)value; 958 applyFilter(scriptExtensionsFilter, &script, UPROPS_SRC_PROPSVEC, ec); 959 } else { 960 IntPropertyContext c = {prop, value}; 961 applyFilter(intPropertyFilter, &c, uprops_getSource(prop), ec); 962 } 963 return *this; 964} 965 966UnicodeSet& 967UnicodeSet::applyPropertyAlias(const UnicodeString& prop, 968 const UnicodeString& value, 969 UErrorCode& ec) { 970 if (U_FAILURE(ec) || isFrozen()) return *this; 971 972 // prop and value used to be converted to char * using the default 973 // converter instead of the invariant conversion. 974 // This should not be necessary because all Unicode property and value 975 // names use only invariant characters. 976 // If there are any variant characters, then we won't find them anyway. 977 // Checking first avoids assertion failures in the conversion. 978 if( !uprv_isInvariantUString(prop.getBuffer(), prop.length()) || 979 !uprv_isInvariantUString(value.getBuffer(), value.length()) 980 ) { 981 FAIL(ec); 982 } 983 CharString pname, vname; 984 pname.appendInvariantChars(prop, ec); 985 vname.appendInvariantChars(value, ec); 986 if (U_FAILURE(ec)) return *this; 987 988 UProperty p; 989 int32_t v; 990 UBool invert = FALSE; 991 992 if (value.length() > 0) { 993 p = u_getPropertyEnum(pname.data()); 994 if (p == UCHAR_INVALID_CODE) FAIL(ec); 995 996 // Treat gc as gcm 997 if (p == UCHAR_GENERAL_CATEGORY) { 998 p = UCHAR_GENERAL_CATEGORY_MASK; 999 } 1000 1001 if ((p >= UCHAR_BINARY_START && p < UCHAR_BINARY_LIMIT) || 1002 (p >= UCHAR_INT_START && p < UCHAR_INT_LIMIT) || 1003 (p >= UCHAR_MASK_START && p < UCHAR_MASK_LIMIT)) { 1004 v = u_getPropertyValueEnum(p, vname.data()); 1005 if (v == UCHAR_INVALID_CODE) { 1006 // Handle numeric CCC 1007 if (p == UCHAR_CANONICAL_COMBINING_CLASS || 1008 p == UCHAR_TRAIL_CANONICAL_COMBINING_CLASS || 1009 p == UCHAR_LEAD_CANONICAL_COMBINING_CLASS) { 1010 char* end; 1011 double value = uprv_strtod(vname.data(), &end); 1012 // Anything between 0 and 255 is valid even if unused. 1013 // Cast double->int only after range check. 1014 // We catch NaN here because comparing it with both 0 and 255 will be false 1015 // (as are all comparisons with NaN). 1016 if (*end != 0 || !(0 <= value && value <= 255) || 1017 (v = (int32_t)value) != value) { 1018 // non-integral value or outside 0..255, or trailing junk 1019 FAIL(ec); 1020 } 1021 } else { 1022 FAIL(ec); 1023 } 1024 } 1025 } 1026 1027 else { 1028 1029 switch (p) { 1030 case UCHAR_NUMERIC_VALUE: 1031 { 1032 char* end; 1033 double value = uprv_strtod(vname.data(), &end); 1034 if (*end != 0) { 1035 FAIL(ec); 1036 } 1037 applyFilter(numericValueFilter, &value, UPROPS_SRC_CHAR, ec); 1038 return *this; 1039 } 1040 case UCHAR_NAME: 1041 { 1042 // Must munge name, since u_charFromName() does not do 1043 // 'loose' matching. 1044 char buf[128]; // it suffices that this be > uprv_getMaxCharNameLength 1045 if (!mungeCharName(buf, vname.data(), sizeof(buf))) FAIL(ec); 1046 UChar32 ch = u_charFromName(U_EXTENDED_CHAR_NAME, buf, &ec); 1047 if (U_SUCCESS(ec)) { 1048 clear(); 1049 add(ch); 1050 return *this; 1051 } else { 1052 FAIL(ec); 1053 } 1054 } 1055 case UCHAR_UNICODE_1_NAME: 1056 // ICU 49 deprecates the Unicode_1_Name property APIs. 1057 FAIL(ec); 1058 case UCHAR_AGE: 1059 { 1060 // Must munge name, since u_versionFromString() does not do 1061 // 'loose' matching. 1062 char buf[128]; 1063 if (!mungeCharName(buf, vname.data(), sizeof(buf))) FAIL(ec); 1064 UVersionInfo version; 1065 u_versionFromString(version, buf); 1066 applyFilter(versionFilter, &version, UPROPS_SRC_PROPSVEC, ec); 1067 return *this; 1068 } 1069 case UCHAR_SCRIPT_EXTENSIONS: 1070 v = u_getPropertyValueEnum(UCHAR_SCRIPT, vname.data()); 1071 if (v == UCHAR_INVALID_CODE) { 1072 FAIL(ec); 1073 } 1074 // fall through to calling applyIntPropertyValue() 1075 break; 1076 default: 1077 // p is a non-binary, non-enumerated property that we 1078 // don't support (yet). 1079 FAIL(ec); 1080 } 1081 } 1082 } 1083 1084 else { 1085 // value is empty. Interpret as General Category, Script, or 1086 // Binary property. 1087 p = UCHAR_GENERAL_CATEGORY_MASK; 1088 v = u_getPropertyValueEnum(p, pname.data()); 1089 if (v == UCHAR_INVALID_CODE) { 1090 p = UCHAR_SCRIPT; 1091 v = u_getPropertyValueEnum(p, pname.data()); 1092 if (v == UCHAR_INVALID_CODE) { 1093 p = u_getPropertyEnum(pname.data()); 1094 if (p >= UCHAR_BINARY_START && p < UCHAR_BINARY_LIMIT) { 1095 v = 1; 1096 } else if (0 == uprv_comparePropertyNames(ANY, pname.data())) { 1097 set(MIN_VALUE, MAX_VALUE); 1098 return *this; 1099 } else if (0 == uprv_comparePropertyNames(ASCII, pname.data())) { 1100 set(0, 0x7F); 1101 return *this; 1102 } else if (0 == uprv_comparePropertyNames(ASSIGNED, pname.data())) { 1103 // [:Assigned:]=[:^Cn:] 1104 p = UCHAR_GENERAL_CATEGORY_MASK; 1105 v = U_GC_CN_MASK; 1106 invert = TRUE; 1107 } else { 1108 FAIL(ec); 1109 } 1110 } 1111 } 1112 } 1113 1114 applyIntPropertyValue(p, v, ec); 1115 if(invert) { 1116 complement(); 1117 } 1118 1119 if (isBogus() && U_SUCCESS(ec)) { 1120 // We likely ran out of memory. AHHH! 1121 ec = U_MEMORY_ALLOCATION_ERROR; 1122 } 1123 return *this; 1124} 1125 1126//---------------------------------------------------------------- 1127// Property set patterns 1128//---------------------------------------------------------------- 1129 1130/** 1131 * Return true if the given position, in the given pattern, appears 1132 * to be the start of a property set pattern. 1133 */ 1134UBool UnicodeSet::resemblesPropertyPattern(const UnicodeString& pattern, 1135 int32_t pos) { 1136 // Patterns are at least 5 characters long 1137 if ((pos+5) > pattern.length()) { 1138 return FALSE; 1139 } 1140 1141 // Look for an opening [:, [:^, \p, or \P 1142 return isPOSIXOpen(pattern, pos) || isPerlOpen(pattern, pos) || isNameOpen(pattern, pos); 1143} 1144 1145/** 1146 * Return true if the given iterator appears to point at a 1147 * property pattern. Regardless of the result, return with the 1148 * iterator unchanged. 1149 * @param chars iterator over the pattern characters. Upon return 1150 * it will be unchanged. 1151 * @param iterOpts RuleCharacterIterator options 1152 */ 1153UBool UnicodeSet::resemblesPropertyPattern(RuleCharacterIterator& chars, 1154 int32_t iterOpts) { 1155 // NOTE: literal will always be FALSE, because we don't parse escapes. 1156 UBool result = FALSE, literal; 1157 UErrorCode ec = U_ZERO_ERROR; 1158 iterOpts &= ~RuleCharacterIterator::PARSE_ESCAPES; 1159 RuleCharacterIterator::Pos pos; 1160 chars.getPos(pos); 1161 UChar32 c = chars.next(iterOpts, literal, ec); 1162 if (c == 0x5B /*'['*/ || c == 0x5C /*'\\'*/) { 1163 UChar32 d = chars.next(iterOpts & ~RuleCharacterIterator::SKIP_WHITESPACE, 1164 literal, ec); 1165 result = (c == 0x5B /*'['*/) ? (d == 0x3A /*':'*/) : 1166 (d == 0x4E /*'N'*/ || d == 0x70 /*'p'*/ || d == 0x50 /*'P'*/); 1167 } 1168 chars.setPos(pos); 1169 return result && U_SUCCESS(ec); 1170} 1171 1172/** 1173 * Parse the given property pattern at the given parse position. 1174 */ 1175UnicodeSet& UnicodeSet::applyPropertyPattern(const UnicodeString& pattern, 1176 ParsePosition& ppos, 1177 UErrorCode &ec) { 1178 int32_t pos = ppos.getIndex(); 1179 1180 UBool posix = FALSE; // true for [:pat:], false for \p{pat} \P{pat} \N{pat} 1181 UBool isName = FALSE; // true for \N{pat}, o/w false 1182 UBool invert = FALSE; 1183 1184 if (U_FAILURE(ec)) return *this; 1185 1186 // Minimum length is 5 characters, e.g. \p{L} 1187 if ((pos+5) > pattern.length()) { 1188 FAIL(ec); 1189 } 1190 1191 // On entry, ppos should point to one of the following locations: 1192 // Look for an opening [:, [:^, \p, or \P 1193 if (isPOSIXOpen(pattern, pos)) { 1194 posix = TRUE; 1195 pos += 2; 1196 pos = ICU_Utility::skipWhitespace(pattern, pos); 1197 if (pos < pattern.length() && pattern.charAt(pos) == COMPLEMENT) { 1198 ++pos; 1199 invert = TRUE; 1200 } 1201 } else if (isPerlOpen(pattern, pos) || isNameOpen(pattern, pos)) { 1202 UChar c = pattern.charAt(pos+1); 1203 invert = (c == UPPER_P); 1204 isName = (c == UPPER_N); 1205 pos += 2; 1206 pos = ICU_Utility::skipWhitespace(pattern, pos); 1207 if (pos == pattern.length() || pattern.charAt(pos++) != OPEN_BRACE) { 1208 // Syntax error; "\p" or "\P" not followed by "{" 1209 FAIL(ec); 1210 } 1211 } else { 1212 // Open delimiter not seen 1213 FAIL(ec); 1214 } 1215 1216 // Look for the matching close delimiter, either :] or } 1217 int32_t close; 1218 if (posix) { 1219 close = pattern.indexOf(POSIX_CLOSE, 2, pos); 1220 } else { 1221 close = pattern.indexOf(CLOSE_BRACE, pos); 1222 } 1223 if (close < 0) { 1224 // Syntax error; close delimiter missing 1225 FAIL(ec); 1226 } 1227 1228 // Look for an '=' sign. If this is present, we will parse a 1229 // medium \p{gc=Cf} or long \p{GeneralCategory=Format} 1230 // pattern. 1231 int32_t equals = pattern.indexOf(EQUALS, pos); 1232 UnicodeString propName, valueName; 1233 if (equals >= 0 && equals < close && !isName) { 1234 // Equals seen; parse medium/long pattern 1235 pattern.extractBetween(pos, equals, propName); 1236 pattern.extractBetween(equals+1, close, valueName); 1237 } 1238 1239 else { 1240 // Handle case where no '=' is seen, and \N{} 1241 pattern.extractBetween(pos, close, propName); 1242 1243 // Handle \N{name} 1244 if (isName) { 1245 // This is a little inefficient since it means we have to 1246 // parse NAME_PROP back to UCHAR_NAME even though we already 1247 // know it's UCHAR_NAME. If we refactor the API to 1248 // support args of (UProperty, char*) then we can remove 1249 // NAME_PROP and make this a little more efficient. 1250 valueName = propName; 1251 propName = UnicodeString(NAME_PROP, NAME_PROP_LENGTH, US_INV); 1252 } 1253 } 1254 1255 applyPropertyAlias(propName, valueName, ec); 1256 1257 if (U_SUCCESS(ec)) { 1258 if (invert) { 1259 complement(); 1260 } 1261 1262 // Move to the limit position after the close delimiter if the 1263 // parse succeeded. 1264 ppos.setIndex(close + (posix ? 2 : 1)); 1265 } 1266 1267 return *this; 1268} 1269 1270/** 1271 * Parse a property pattern. 1272 * @param chars iterator over the pattern characters. Upon return 1273 * it will be advanced to the first character after the parsed 1274 * pattern, or the end of the iteration if all characters are 1275 * parsed. 1276 * @param rebuiltPat the pattern that was parsed, rebuilt or 1277 * copied from the input pattern, as appropriate. 1278 */ 1279void UnicodeSet::applyPropertyPattern(RuleCharacterIterator& chars, 1280 UnicodeString& rebuiltPat, 1281 UErrorCode& ec) { 1282 if (U_FAILURE(ec)) return; 1283 UnicodeString pattern; 1284 chars.lookahead(pattern); 1285 ParsePosition pos(0); 1286 applyPropertyPattern(pattern, pos, ec); 1287 if (U_FAILURE(ec)) return; 1288 if (pos.getIndex() == 0) { 1289 // syntaxError(chars, "Invalid property pattern"); 1290 ec = U_MALFORMED_SET; 1291 return; 1292 } 1293 chars.jumpahead(pos.getIndex()); 1294 rebuiltPat.append(pattern, 0, pos.getIndex()); 1295} 1296 1297U_NAMESPACE_END 1298