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 "uinvchar.h" |
40 | #include "uprops.h" |
41 | #include "charstr.h" |
42 | #include "cstring.h" |
43 | #include "mutex.h" |
44 | #include "umutex.h" |
45 | #include "uassert.h" |
46 | #include "hash.h" |
47 | |
48 | U_NAMESPACE_USE |
49 | |
50 | // Special property set IDs |
51 | static const char ANY[] = "ANY" ; // [\u0000-\U0010FFFF] |
52 | static const char ASCII[] = "ASCII" ; // [\u0000-\u007F] |
53 | static const char ASSIGNED[] = "Assigned" ; // [:^Cn:] |
54 | |
55 | // Unicode name property alias |
56 | #define NAME_PROP "na" |
57 | #define NAME_PROP_LENGTH 2 |
58 | |
59 | // Cached sets ------------------------------------------------------------- *** |
60 | |
61 | U_CDECL_BEGIN |
62 | static UBool U_CALLCONV uset_cleanup(); |
63 | |
64 | static UnicodeSet *uni32Singleton; |
65 | static icu::UInitOnce uni32InitOnce {}; |
66 | |
67 | /** |
68 | * Cleanup function for UnicodeSet |
69 | */ |
70 | static UBool U_CALLCONV uset_cleanup() { |
71 | delete uni32Singleton; |
72 | uni32Singleton = nullptr; |
73 | uni32InitOnce.reset(); |
74 | return true; |
75 | } |
76 | |
77 | U_CDECL_END |
78 | |
79 | U_NAMESPACE_BEGIN |
80 | |
81 | namespace { |
82 | |
83 | // Cache some sets for other services -------------------------------------- *** |
84 | void U_CALLCONV createUni32Set(UErrorCode &errorCode) { |
85 | U_ASSERT(uni32Singleton == nullptr); |
86 | uni32Singleton = new UnicodeSet(UNICODE_STRING_SIMPLE("[:age=3.2:]" ), errorCode); |
87 | if(uni32Singleton==nullptr) { |
88 | errorCode=U_MEMORY_ALLOCATION_ERROR; |
89 | } else { |
90 | uni32Singleton->freeze(); |
91 | } |
92 | ucln_common_registerCleanup(UCLN_COMMON_USET, uset_cleanup); |
93 | } |
94 | |
95 | |
96 | U_CFUNC UnicodeSet * |
97 | uniset_getUnicode32Instance(UErrorCode &errorCode) { |
98 | umtx_initOnce(uni32InitOnce, &createUni32Set, errorCode); |
99 | return uni32Singleton; |
100 | } |
101 | |
102 | // helper functions for matching of pattern syntax pieces ------------------ *** |
103 | // these functions are parallel to the PERL_OPEN etc. strings above |
104 | |
105 | // using these functions is not only faster than UnicodeString::compare() and |
106 | // caseCompare(), but they also make UnicodeSet work for simple patterns when |
107 | // no Unicode properties data is available - when caseCompare() fails |
108 | |
109 | static inline UBool |
110 | isPerlOpen(const UnicodeString &pattern, int32_t pos) { |
111 | char16_t c; |
112 | return pattern.charAt(pos)==u'\\' && ((c=pattern.charAt(pos+1))==u'p' || c==u'P'); |
113 | } |
114 | |
115 | /*static inline UBool |
116 | isPerlClose(const UnicodeString &pattern, int32_t pos) { |
117 | return pattern.charAt(pos)==u'}'; |
118 | }*/ |
119 | |
120 | static inline UBool |
121 | isNameOpen(const UnicodeString &pattern, int32_t pos) { |
122 | return pattern.charAt(pos)==u'\\' && pattern.charAt(pos+1)==u'N'; |
123 | } |
124 | |
125 | static inline UBool |
126 | isPOSIXOpen(const UnicodeString &pattern, int32_t pos) { |
127 | return pattern.charAt(pos)==u'[' && pattern.charAt(pos+1)==u':'; |
128 | } |
129 | |
130 | /*static inline UBool |
131 | isPOSIXClose(const UnicodeString &pattern, int32_t pos) { |
132 | return pattern.charAt(pos)==u':' && pattern.charAt(pos+1)==u']'; |
133 | }*/ |
134 | |
135 | // TODO memory debugging provided inside uniset.cpp |
136 | // could be made available here but probably obsolete with use of modern |
137 | // memory leak checker tools |
138 | #define _dbgct(me) |
139 | |
140 | } // namespace |
141 | |
142 | //---------------------------------------------------------------- |
143 | // Constructors &c |
144 | //---------------------------------------------------------------- |
145 | |
146 | /** |
147 | * Constructs a set from the given pattern, optionally ignoring |
148 | * white space. See the class description for the syntax of the |
149 | * pattern language. |
150 | * @param pattern a string specifying what characters are in the set |
151 | */ |
152 | UnicodeSet::UnicodeSet(const UnicodeString& pattern, |
153 | UErrorCode& status) { |
154 | applyPattern(pattern, status); |
155 | _dbgct(this); |
156 | } |
157 | |
158 | //---------------------------------------------------------------- |
159 | // Public API |
160 | //---------------------------------------------------------------- |
161 | |
162 | UnicodeSet& UnicodeSet::applyPattern(const UnicodeString& pattern, |
163 | UErrorCode& status) { |
164 | // Equivalent to |
165 | // return applyPattern(pattern, USET_IGNORE_SPACE, nullptr, status); |
166 | // but without dependency on closeOver(). |
167 | ParsePosition pos(0); |
168 | applyPatternIgnoreSpace(pattern, pos, nullptr, status); |
169 | if (U_FAILURE(status)) return *this; |
170 | |
171 | int32_t i = pos.getIndex(); |
172 | // Skip over trailing whitespace |
173 | ICU_Utility::skipWhitespace(pattern, i, true); |
174 | if (i != pattern.length()) { |
175 | status = U_ILLEGAL_ARGUMENT_ERROR; |
176 | } |
177 | return *this; |
178 | } |
179 | |
180 | void |
181 | UnicodeSet::applyPatternIgnoreSpace(const UnicodeString& pattern, |
182 | ParsePosition& pos, |
183 | const SymbolTable* symbols, |
184 | UErrorCode& status) { |
185 | if (U_FAILURE(status)) { |
186 | return; |
187 | } |
188 | if (isFrozen()) { |
189 | status = U_NO_WRITE_PERMISSION; |
190 | return; |
191 | } |
192 | // Need to build the pattern in a temporary string because |
193 | // _applyPattern calls add() etc., which set pat to empty. |
194 | UnicodeString rebuiltPat; |
195 | RuleCharacterIterator chars(pattern, symbols, pos); |
196 | applyPattern(chars, symbols, rebuiltPat, USET_IGNORE_SPACE, nullptr, 0, status); |
197 | if (U_FAILURE(status)) return; |
198 | if (chars.inVariable()) { |
199 | // syntaxError(chars, "Extra chars in variable value"); |
200 | status = U_MALFORMED_SET; |
201 | return; |
202 | } |
203 | setPattern(rebuiltPat); |
204 | } |
205 | |
206 | /** |
207 | * Return true if the given position, in the given pattern, appears |
208 | * to be the start of a UnicodeSet pattern. |
209 | */ |
210 | UBool UnicodeSet::resemblesPattern(const UnicodeString& pattern, int32_t pos) { |
211 | return ((pos+1) < pattern.length() && |
212 | pattern.charAt(pos) == (char16_t)91/*[*/) || |
213 | resemblesPropertyPattern(pattern, pos); |
214 | } |
215 | |
216 | //---------------------------------------------------------------- |
217 | // Implementation: Pattern parsing |
218 | //---------------------------------------------------------------- |
219 | |
220 | namespace { |
221 | |
222 | /** |
223 | * A small all-inline class to manage a UnicodeSet pointer. Add |
224 | * operator->() etc. as needed. |
225 | */ |
226 | class UnicodeSetPointer { |
227 | UnicodeSet* p; |
228 | public: |
229 | inline UnicodeSetPointer() : p(0) {} |
230 | inline ~UnicodeSetPointer() { delete p; } |
231 | inline UnicodeSet* pointer() { return p; } |
232 | inline UBool allocate() { |
233 | if (p == 0) { |
234 | p = new UnicodeSet(); |
235 | } |
236 | return p != 0; |
237 | } |
238 | }; |
239 | |
240 | constexpr int32_t MAX_DEPTH = 100; |
241 | |
242 | } // namespace |
243 | |
244 | /** |
245 | * Parse the pattern from the given RuleCharacterIterator. The |
246 | * iterator is advanced over the parsed pattern. |
247 | * @param chars iterator over the pattern characters. Upon return |
248 | * it will be advanced to the first character after the parsed |
249 | * pattern, or the end of the iteration if all characters are |
250 | * parsed. |
251 | * @param symbols symbol table to use to parse and dereference |
252 | * variables, or null if none. |
253 | * @param rebuiltPat the pattern that was parsed, rebuilt or |
254 | * copied from the input pattern, as appropriate. |
255 | * @param options a bit mask of zero or more of the following: |
256 | * IGNORE_SPACE, CASE. |
257 | */ |
258 | void UnicodeSet::applyPattern(RuleCharacterIterator& chars, |
259 | const SymbolTable* symbols, |
260 | UnicodeString& rebuiltPat, |
261 | uint32_t options, |
262 | UnicodeSet& (UnicodeSet::*caseClosure)(int32_t attribute), |
263 | int32_t depth, |
264 | UErrorCode& ec) { |
265 | if (U_FAILURE(ec)) return; |
266 | if (depth > MAX_DEPTH) { |
267 | ec = U_ILLEGAL_ARGUMENT_ERROR; |
268 | return; |
269 | } |
270 | |
271 | // Syntax characters: [ ] ^ - & { } |
272 | |
273 | // Recognized special forms for chars, sets: c-c s-s s&s |
274 | |
275 | int32_t opts = RuleCharacterIterator::PARSE_VARIABLES | |
276 | RuleCharacterIterator::PARSE_ESCAPES; |
277 | if ((options & USET_IGNORE_SPACE) != 0) { |
278 | opts |= RuleCharacterIterator::SKIP_WHITESPACE; |
279 | } |
280 | |
281 | UnicodeString patLocal, buf; |
282 | UBool usePat = false; |
283 | UnicodeSetPointer scratch; |
284 | RuleCharacterIterator::Pos backup; |
285 | |
286 | // mode: 0=before [, 1=between [...], 2=after ] |
287 | // lastItem: 0=none, 1=char, 2=set |
288 | int8_t lastItem = 0, mode = 0; |
289 | UChar32 lastChar = 0; |
290 | char16_t op = 0; |
291 | |
292 | UBool invert = false; |
293 | |
294 | clear(); |
295 | |
296 | while (mode != 2 && !chars.atEnd()) { |
297 | U_ASSERT((lastItem == 0 && op == 0) || |
298 | (lastItem == 1 && (op == 0 || op == u'-')) || |
299 | (lastItem == 2 && (op == 0 || op == u'-' || op == u'&'))); |
300 | |
301 | UChar32 c = 0; |
302 | UBool literal = false; |
303 | UnicodeSet* nested = 0; // alias - do not delete |
304 | |
305 | // -------- Check for property pattern |
306 | |
307 | // setMode: 0=none, 1=unicodeset, 2=propertypat, 3=preparsed |
308 | int8_t setMode = 0; |
309 | if (resemblesPropertyPattern(chars, opts)) { |
310 | setMode = 2; |
311 | } |
312 | |
313 | // -------- Parse '[' of opening delimiter OR nested set. |
314 | // If there is a nested set, use `setMode' to define how |
315 | // the set should be parsed. If the '[' is part of the |
316 | // opening delimiter for this pattern, parse special |
317 | // strings "[", "[^", "[-", and "[^-". Check for stand-in |
318 | // characters representing a nested set in the symbol |
319 | // table. |
320 | |
321 | else { |
322 | // Prepare to backup if necessary |
323 | chars.getPos(backup); |
324 | c = chars.next(opts, literal, ec); |
325 | if (U_FAILURE(ec)) return; |
326 | |
327 | if (c == u'[' && !literal) { |
328 | if (mode == 1) { |
329 | chars.setPos(backup); // backup |
330 | setMode = 1; |
331 | } else { |
332 | // Handle opening '[' delimiter |
333 | mode = 1; |
334 | patLocal.append(u'['); |
335 | chars.getPos(backup); // prepare to backup |
336 | c = chars.next(opts, literal, ec); |
337 | if (U_FAILURE(ec)) return; |
338 | if (c == u'^' && !literal) { |
339 | invert = true; |
340 | patLocal.append(u'^'); |
341 | chars.getPos(backup); // prepare to backup |
342 | c = chars.next(opts, literal, ec); |
343 | if (U_FAILURE(ec)) return; |
344 | } |
345 | // Fall through to handle special leading '-'; |
346 | // otherwise restart loop for nested [], \p{}, etc. |
347 | if (c == u'-') { |
348 | literal = true; |
349 | // Fall through to handle literal '-' below |
350 | } else { |
351 | chars.setPos(backup); // backup |
352 | continue; |
353 | } |
354 | } |
355 | } else if (symbols != 0) { |
356 | const UnicodeFunctor *m = symbols->lookupMatcher(c); |
357 | if (m != 0) { |
358 | const UnicodeSet *ms = dynamic_cast<const UnicodeSet *>(m); |
359 | if (ms == nullptr) { |
360 | ec = U_MALFORMED_SET; |
361 | return; |
362 | } |
363 | // casting away const, but `nested' won't be modified |
364 | // (important not to modify stored set) |
365 | nested = const_cast<UnicodeSet*>(ms); |
366 | setMode = 3; |
367 | } |
368 | } |
369 | } |
370 | |
371 | // -------- Handle a nested set. This either is inline in |
372 | // the pattern or represented by a stand-in that has |
373 | // previously been parsed and was looked up in the symbol |
374 | // table. |
375 | |
376 | if (setMode != 0) { |
377 | if (lastItem == 1) { |
378 | if (op != 0) { |
379 | // syntaxError(chars, "Char expected after operator"); |
380 | ec = U_MALFORMED_SET; |
381 | return; |
382 | } |
383 | add(lastChar, lastChar); |
384 | _appendToPat(patLocal, lastChar, false); |
385 | lastItem = 0; |
386 | op = 0; |
387 | } |
388 | |
389 | if (op == u'-' || op == u'&') { |
390 | patLocal.append(op); |
391 | } |
392 | |
393 | if (nested == 0) { |
394 | // lazy allocation |
395 | if (!scratch.allocate()) { |
396 | ec = U_MEMORY_ALLOCATION_ERROR; |
397 | return; |
398 | } |
399 | nested = scratch.pointer(); |
400 | } |
401 | switch (setMode) { |
402 | case 1: |
403 | nested->applyPattern(chars, symbols, patLocal, options, caseClosure, depth + 1, ec); |
404 | break; |
405 | case 2: |
406 | chars.skipIgnored(opts); |
407 | nested->applyPropertyPattern(chars, patLocal, ec); |
408 | if (U_FAILURE(ec)) return; |
409 | break; |
410 | case 3: // `nested' already parsed |
411 | nested->_toPattern(patLocal, false); |
412 | break; |
413 | } |
414 | |
415 | usePat = true; |
416 | |
417 | if (mode == 0) { |
418 | // Entire pattern is a category; leave parse loop |
419 | *this = *nested; |
420 | mode = 2; |
421 | break; |
422 | } |
423 | |
424 | switch (op) { |
425 | case u'-': |
426 | removeAll(*nested); |
427 | break; |
428 | case u'&': |
429 | retainAll(*nested); |
430 | break; |
431 | case 0: |
432 | addAll(*nested); |
433 | break; |
434 | } |
435 | |
436 | op = 0; |
437 | lastItem = 2; |
438 | |
439 | continue; |
440 | } |
441 | |
442 | if (mode == 0) { |
443 | // syntaxError(chars, "Missing '['"); |
444 | ec = U_MALFORMED_SET; |
445 | return; |
446 | } |
447 | |
448 | // -------- Parse special (syntax) characters. If the |
449 | // current character is not special, or if it is escaped, |
450 | // then fall through and handle it below. |
451 | |
452 | if (!literal) { |
453 | switch (c) { |
454 | case u']': |
455 | if (lastItem == 1) { |
456 | add(lastChar, lastChar); |
457 | _appendToPat(patLocal, lastChar, false); |
458 | } |
459 | // Treat final trailing '-' as a literal |
460 | if (op == u'-') { |
461 | add(op, op); |
462 | patLocal.append(op); |
463 | } else if (op == u'&') { |
464 | // syntaxError(chars, "Trailing '&'"); |
465 | ec = U_MALFORMED_SET; |
466 | return; |
467 | } |
468 | patLocal.append(u']'); |
469 | mode = 2; |
470 | continue; |
471 | case u'-': |
472 | if (op == 0) { |
473 | if (lastItem != 0) { |
474 | op = (char16_t) c; |
475 | continue; |
476 | } else { |
477 | // Treat final trailing '-' as a literal |
478 | add(c, c); |
479 | c = chars.next(opts, literal, ec); |
480 | if (U_FAILURE(ec)) return; |
481 | if (c == u']' && !literal) { |
482 | patLocal.append(u"-]" , 2); |
483 | mode = 2; |
484 | continue; |
485 | } |
486 | } |
487 | } |
488 | // syntaxError(chars, "'-' not after char or set"); |
489 | ec = U_MALFORMED_SET; |
490 | return; |
491 | case u'&': |
492 | if (lastItem == 2 && op == 0) { |
493 | op = (char16_t) c; |
494 | continue; |
495 | } |
496 | // syntaxError(chars, "'&' not after set"); |
497 | ec = U_MALFORMED_SET; |
498 | return; |
499 | case u'^': |
500 | // syntaxError(chars, "'^' not after '['"); |
501 | ec = U_MALFORMED_SET; |
502 | return; |
503 | case u'{': |
504 | if (op != 0) { |
505 | // syntaxError(chars, "Missing operand after operator"); |
506 | ec = U_MALFORMED_SET; |
507 | return; |
508 | } |
509 | if (lastItem == 1) { |
510 | add(lastChar, lastChar); |
511 | _appendToPat(patLocal, lastChar, false); |
512 | } |
513 | lastItem = 0; |
514 | buf.truncate(0); |
515 | { |
516 | UBool ok = false; |
517 | while (!chars.atEnd()) { |
518 | c = chars.next(opts, literal, ec); |
519 | if (U_FAILURE(ec)) return; |
520 | if (c == u'}' && !literal) { |
521 | ok = true; |
522 | break; |
523 | } |
524 | buf.append(c); |
525 | } |
526 | if (!ok) { |
527 | // syntaxError(chars, "Invalid multicharacter string"); |
528 | ec = U_MALFORMED_SET; |
529 | return; |
530 | } |
531 | } |
532 | // We have new string. Add it to set and continue; |
533 | // we don't need to drop through to the further |
534 | // processing |
535 | add(buf); |
536 | patLocal.append(u'{'); |
537 | _appendToPat(patLocal, buf, false); |
538 | patLocal.append(u'}'); |
539 | continue; |
540 | case SymbolTable::SYMBOL_REF: |
541 | // symbols nosymbols |
542 | // [a-$] error error (ambiguous) |
543 | // [a$] anchor anchor |
544 | // [a-$x] var "x"* literal '$' |
545 | // [a-$.] error literal '$' |
546 | // *We won't get here in the case of var "x" |
547 | { |
548 | chars.getPos(backup); |
549 | c = chars.next(opts, literal, ec); |
550 | if (U_FAILURE(ec)) return; |
551 | UBool anchor = (c == u']' && !literal); |
552 | if (symbols == 0 && !anchor) { |
553 | c = SymbolTable::SYMBOL_REF; |
554 | chars.setPos(backup); |
555 | break; // literal '$' |
556 | } |
557 | if (anchor && op == 0) { |
558 | if (lastItem == 1) { |
559 | add(lastChar, lastChar); |
560 | _appendToPat(patLocal, lastChar, false); |
561 | } |
562 | add(U_ETHER); |
563 | usePat = true; |
564 | patLocal.append((char16_t) SymbolTable::SYMBOL_REF); |
565 | patLocal.append(u']'); |
566 | mode = 2; |
567 | continue; |
568 | } |
569 | // syntaxError(chars, "Unquoted '$'"); |
570 | ec = U_MALFORMED_SET; |
571 | return; |
572 | } |
573 | default: |
574 | break; |
575 | } |
576 | } |
577 | |
578 | // -------- Parse literal characters. This includes both |
579 | // escaped chars ("\u4E01") and non-syntax characters |
580 | // ("a"). |
581 | |
582 | switch (lastItem) { |
583 | case 0: |
584 | lastItem = 1; |
585 | lastChar = c; |
586 | break; |
587 | case 1: |
588 | if (op == u'-') { |
589 | if (lastChar >= c) { |
590 | // Don't allow redundant (a-a) or empty (b-a) ranges; |
591 | // these are most likely typos. |
592 | // syntaxError(chars, "Invalid range"); |
593 | ec = U_MALFORMED_SET; |
594 | return; |
595 | } |
596 | add(lastChar, c); |
597 | _appendToPat(patLocal, lastChar, false); |
598 | patLocal.append(op); |
599 | _appendToPat(patLocal, c, false); |
600 | lastItem = 0; |
601 | op = 0; |
602 | } else { |
603 | add(lastChar, lastChar); |
604 | _appendToPat(patLocal, lastChar, false); |
605 | lastChar = c; |
606 | } |
607 | break; |
608 | case 2: |
609 | if (op != 0) { |
610 | // syntaxError(chars, "Set expected after operator"); |
611 | ec = U_MALFORMED_SET; |
612 | return; |
613 | } |
614 | lastChar = c; |
615 | lastItem = 1; |
616 | break; |
617 | } |
618 | } |
619 | |
620 | if (mode != 2) { |
621 | // syntaxError(chars, "Missing ']'"); |
622 | ec = U_MALFORMED_SET; |
623 | return; |
624 | } |
625 | |
626 | chars.skipIgnored(opts); |
627 | |
628 | /** |
629 | * Handle global flags (invert, case insensitivity). If this |
630 | * pattern should be compiled case-insensitive, then we need |
631 | * to close over case BEFORE COMPLEMENTING. This makes |
632 | * patterns like /[^abc]/i work. |
633 | */ |
634 | if ((options & USET_CASE_MASK) != 0) { |
635 | (this->*caseClosure)(options); |
636 | } |
637 | if (invert) { |
638 | complement().removeAllStrings(); // code point complement |
639 | } |
640 | |
641 | // Use the rebuilt pattern (patLocal) only if necessary. Prefer the |
642 | // generated pattern. |
643 | if (usePat) { |
644 | rebuiltPat.append(patLocal); |
645 | } else { |
646 | _generatePattern(rebuiltPat, false); |
647 | } |
648 | if (isBogus() && U_SUCCESS(ec)) { |
649 | // We likely ran out of memory. AHHH! |
650 | ec = U_MEMORY_ALLOCATION_ERROR; |
651 | } |
652 | } |
653 | |
654 | //---------------------------------------------------------------- |
655 | // Property set implementation |
656 | //---------------------------------------------------------------- |
657 | |
658 | namespace { |
659 | |
660 | static UBool numericValueFilter(UChar32 ch, void* context) { |
661 | return u_getNumericValue(ch) == *(double*)context; |
662 | } |
663 | |
664 | static UBool generalCategoryMaskFilter(UChar32 ch, void* context) { |
665 | int32_t value = *(int32_t*)context; |
666 | return (U_GET_GC_MASK((UChar32) ch) & value) != 0; |
667 | } |
668 | |
669 | static UBool versionFilter(UChar32 ch, void* context) { |
670 | static const UVersionInfo none = { 0, 0, 0, 0 }; |
671 | UVersionInfo v; |
672 | u_charAge(ch, v); |
673 | UVersionInfo* version = (UVersionInfo*)context; |
674 | return uprv_memcmp(&v, &none, sizeof(v)) > 0 && uprv_memcmp(&v, version, sizeof(v)) <= 0; |
675 | } |
676 | |
677 | typedef struct { |
678 | UProperty prop; |
679 | int32_t value; |
680 | } IntPropertyContext; |
681 | |
682 | static UBool intPropertyFilter(UChar32 ch, void* context) { |
683 | IntPropertyContext* c = (IntPropertyContext*)context; |
684 | return u_getIntPropertyValue((UChar32) ch, c->prop) == c->value; |
685 | } |
686 | |
687 | static UBool scriptExtensionsFilter(UChar32 ch, void* context) { |
688 | return uscript_hasScript(ch, *(UScriptCode*)context); |
689 | } |
690 | |
691 | } // namespace |
692 | |
693 | /** |
694 | * Generic filter-based scanning code for UCD property UnicodeSets. |
695 | */ |
696 | void UnicodeSet::applyFilter(UnicodeSet::Filter filter, |
697 | void* context, |
698 | const UnicodeSet* inclusions, |
699 | UErrorCode &status) { |
700 | if (U_FAILURE(status)) return; |
701 | |
702 | // Logically, walk through all Unicode characters, noting the start |
703 | // and end of each range for which filter.contain(c) is |
704 | // true. Add each range to a set. |
705 | // |
706 | // To improve performance, use an inclusions set which |
707 | // encodes information about character ranges that are known |
708 | // to have identical properties. |
709 | // inclusions contains the first characters of |
710 | // same-value ranges for the given property. |
711 | |
712 | clear(); |
713 | |
714 | UChar32 startHasProperty = -1; |
715 | int32_t limitRange = inclusions->getRangeCount(); |
716 | |
717 | for (int j=0; j<limitRange; ++j) { |
718 | // get current range |
719 | UChar32 start = inclusions->getRangeStart(j); |
720 | UChar32 end = inclusions->getRangeEnd(j); |
721 | |
722 | // for all the code points in the range, process |
723 | for (UChar32 ch = start; ch <= end; ++ch) { |
724 | // only add to this UnicodeSet on inflection points -- |
725 | // where the hasProperty value changes to false |
726 | if ((*filter)(ch, context)) { |
727 | if (startHasProperty < 0) { |
728 | startHasProperty = ch; |
729 | } |
730 | } else if (startHasProperty >= 0) { |
731 | add(startHasProperty, ch-1); |
732 | startHasProperty = -1; |
733 | } |
734 | } |
735 | } |
736 | if (startHasProperty >= 0) { |
737 | add((UChar32)startHasProperty, (UChar32)0x10FFFF); |
738 | } |
739 | if (isBogus() && U_SUCCESS(status)) { |
740 | // We likely ran out of memory. AHHH! |
741 | status = U_MEMORY_ALLOCATION_ERROR; |
742 | } |
743 | } |
744 | |
745 | namespace { |
746 | |
747 | static UBool mungeCharName(char* dst, const char* src, int32_t dstCapacity) { |
748 | /* Note: we use ' ' in compiler code page */ |
749 | int32_t j = 0; |
750 | char ch; |
751 | --dstCapacity; /* make room for term. zero */ |
752 | while ((ch = *src++) != 0) { |
753 | if (ch == ' ' && (j==0 || (j>0 && dst[j-1]==' '))) { |
754 | continue; |
755 | } |
756 | if (j >= dstCapacity) return false; |
757 | dst[j++] = ch; |
758 | } |
759 | if (j > 0 && dst[j-1] == ' ') --j; |
760 | dst[j] = 0; |
761 | return true; |
762 | } |
763 | |
764 | } // namespace |
765 | |
766 | //---------------------------------------------------------------- |
767 | // Property set API |
768 | //---------------------------------------------------------------- |
769 | |
770 | #define FAIL(ec) UPRV_BLOCK_MACRO_BEGIN { \ |
771 | ec=U_ILLEGAL_ARGUMENT_ERROR; \ |
772 | return *this; \ |
773 | } UPRV_BLOCK_MACRO_END |
774 | |
775 | UnicodeSet& |
776 | UnicodeSet::applyIntPropertyValue(UProperty prop, int32_t value, UErrorCode& ec) { |
777 | if (U_FAILURE(ec) || isFrozen()) { return *this; } |
778 | if (prop == UCHAR_GENERAL_CATEGORY_MASK) { |
779 | const UnicodeSet* inclusions = CharacterProperties::getInclusionsForProperty(prop, ec); |
780 | applyFilter(generalCategoryMaskFilter, &value, inclusions, ec); |
781 | } else if (prop == UCHAR_SCRIPT_EXTENSIONS) { |
782 | const UnicodeSet* inclusions = CharacterProperties::getInclusionsForProperty(prop, ec); |
783 | UScriptCode script = (UScriptCode)value; |
784 | applyFilter(scriptExtensionsFilter, &script, inclusions, ec); |
785 | } else if (0 <= prop && prop < UCHAR_BINARY_LIMIT) { |
786 | if (value == 0 || value == 1) { |
787 | const USet *set = u_getBinaryPropertySet(prop, &ec); |
788 | if (U_FAILURE(ec)) { return *this; } |
789 | copyFrom(*UnicodeSet::fromUSet(set), true); |
790 | if (value == 0) { |
791 | complement().removeAllStrings(); // code point complement |
792 | } |
793 | } else { |
794 | clear(); |
795 | } |
796 | } else if (UCHAR_INT_START <= prop && prop < UCHAR_INT_LIMIT) { |
797 | const UnicodeSet* inclusions = CharacterProperties::getInclusionsForProperty(prop, ec); |
798 | IntPropertyContext c = {prop, value}; |
799 | applyFilter(intPropertyFilter, &c, inclusions, ec); |
800 | } else { |
801 | ec = U_ILLEGAL_ARGUMENT_ERROR; |
802 | } |
803 | return *this; |
804 | } |
805 | |
806 | UnicodeSet& |
807 | UnicodeSet::applyPropertyAlias(const UnicodeString& prop, |
808 | const UnicodeString& value, |
809 | UErrorCode& ec) { |
810 | if (U_FAILURE(ec) || isFrozen()) return *this; |
811 | |
812 | // prop and value used to be converted to char * using the default |
813 | // converter instead of the invariant conversion. |
814 | // This should not be necessary because all Unicode property and value |
815 | // names use only invariant characters. |
816 | // If there are any variant characters, then we won't find them anyway. |
817 | // Checking first avoids assertion failures in the conversion. |
818 | if( !uprv_isInvariantUString(prop.getBuffer(), prop.length()) || |
819 | !uprv_isInvariantUString(value.getBuffer(), value.length()) |
820 | ) { |
821 | FAIL(ec); |
822 | } |
823 | CharString pname, vname; |
824 | pname.appendInvariantChars(prop, ec); |
825 | vname.appendInvariantChars(value, ec); |
826 | if (U_FAILURE(ec)) return *this; |
827 | |
828 | UProperty p; |
829 | int32_t v; |
830 | UBool invert = false; |
831 | |
832 | if (value.length() > 0) { |
833 | p = u_getPropertyEnum(pname.data()); |
834 | if (p == UCHAR_INVALID_CODE) FAIL(ec); |
835 | |
836 | // Treat gc as gcm |
837 | if (p == UCHAR_GENERAL_CATEGORY) { |
838 | p = UCHAR_GENERAL_CATEGORY_MASK; |
839 | } |
840 | |
841 | if ((p >= UCHAR_BINARY_START && p < UCHAR_BINARY_LIMIT) || |
842 | (p >= UCHAR_INT_START && p < UCHAR_INT_LIMIT) || |
843 | (p >= UCHAR_MASK_START && p < UCHAR_MASK_LIMIT)) { |
844 | v = u_getPropertyValueEnum(p, vname.data()); |
845 | if (v == UCHAR_INVALID_CODE) { |
846 | // Handle numeric CCC |
847 | if (p == UCHAR_CANONICAL_COMBINING_CLASS || |
848 | p == UCHAR_TRAIL_CANONICAL_COMBINING_CLASS || |
849 | p == UCHAR_LEAD_CANONICAL_COMBINING_CLASS) { |
850 | char* end; |
851 | double val = uprv_strtod(vname.data(), &end); |
852 | // Anything between 0 and 255 is valid even if unused. |
853 | // Cast double->int only after range check. |
854 | // We catch NaN here because comparing it with both 0 and 255 will be false |
855 | // (as are all comparisons with NaN). |
856 | if (*end != 0 || !(0 <= val && val <= 255) || |
857 | (v = (int32_t)val) != val) { |
858 | // non-integral value or outside 0..255, or trailing junk |
859 | FAIL(ec); |
860 | } |
861 | } else { |
862 | FAIL(ec); |
863 | } |
864 | } |
865 | } |
866 | |
867 | else { |
868 | |
869 | switch (p) { |
870 | case UCHAR_NUMERIC_VALUE: |
871 | { |
872 | char* end; |
873 | double val = uprv_strtod(vname.data(), &end); |
874 | if (*end != 0) { |
875 | FAIL(ec); |
876 | } |
877 | applyFilter(numericValueFilter, &val, |
878 | CharacterProperties::getInclusionsForProperty(p, ec), ec); |
879 | return *this; |
880 | } |
881 | case UCHAR_NAME: |
882 | { |
883 | // Must munge name, since u_charFromName() does not do |
884 | // 'loose' matching. |
885 | char buf[128]; // it suffices that this be > uprv_getMaxCharNameLength |
886 | if (!mungeCharName(buf, vname.data(), sizeof(buf))) FAIL(ec); |
887 | UChar32 ch = u_charFromName(U_EXTENDED_CHAR_NAME, buf, &ec); |
888 | if (U_SUCCESS(ec)) { |
889 | clear(); |
890 | add(ch); |
891 | return *this; |
892 | } else { |
893 | FAIL(ec); |
894 | } |
895 | } |
896 | case UCHAR_UNICODE_1_NAME: |
897 | // ICU 49 deprecates the Unicode_1_Name property APIs. |
898 | FAIL(ec); |
899 | case UCHAR_AGE: |
900 | { |
901 | // Must munge name, since u_versionFromString() does not do |
902 | // 'loose' matching. |
903 | char buf[128]; |
904 | if (!mungeCharName(buf, vname.data(), sizeof(buf))) FAIL(ec); |
905 | UVersionInfo version; |
906 | u_versionFromString(version, buf); |
907 | applyFilter(versionFilter, &version, |
908 | CharacterProperties::getInclusionsForProperty(p, ec), ec); |
909 | return *this; |
910 | } |
911 | case UCHAR_SCRIPT_EXTENSIONS: |
912 | v = u_getPropertyValueEnum(UCHAR_SCRIPT, vname.data()); |
913 | if (v == UCHAR_INVALID_CODE) { |
914 | FAIL(ec); |
915 | } |
916 | // fall through to calling applyIntPropertyValue() |
917 | break; |
918 | default: |
919 | // p is a non-binary, non-enumerated property that we |
920 | // don't support (yet). |
921 | FAIL(ec); |
922 | } |
923 | } |
924 | } |
925 | |
926 | else { |
927 | // value is empty. Interpret as General Category, Script, or |
928 | // Binary property. |
929 | p = UCHAR_GENERAL_CATEGORY_MASK; |
930 | v = u_getPropertyValueEnum(p, pname.data()); |
931 | if (v == UCHAR_INVALID_CODE) { |
932 | p = UCHAR_SCRIPT; |
933 | v = u_getPropertyValueEnum(p, pname.data()); |
934 | if (v == UCHAR_INVALID_CODE) { |
935 | p = u_getPropertyEnum(pname.data()); |
936 | if (p >= UCHAR_BINARY_START && p < UCHAR_BINARY_LIMIT) { |
937 | v = 1; |
938 | } else if (0 == uprv_comparePropertyNames(ANY, pname.data())) { |
939 | set(MIN_VALUE, MAX_VALUE); |
940 | return *this; |
941 | } else if (0 == uprv_comparePropertyNames(ASCII, pname.data())) { |
942 | set(0, 0x7F); |
943 | return *this; |
944 | } else if (0 == uprv_comparePropertyNames(ASSIGNED, pname.data())) { |
945 | // [:Assigned:]=[:^Cn:] |
946 | p = UCHAR_GENERAL_CATEGORY_MASK; |
947 | v = U_GC_CN_MASK; |
948 | invert = true; |
949 | } else { |
950 | FAIL(ec); |
951 | } |
952 | } |
953 | } |
954 | } |
955 | |
956 | applyIntPropertyValue(p, v, ec); |
957 | if(invert) { |
958 | complement().removeAllStrings(); // code point complement |
959 | } |
960 | |
961 | if (isBogus() && U_SUCCESS(ec)) { |
962 | // We likely ran out of memory. AHHH! |
963 | ec = U_MEMORY_ALLOCATION_ERROR; |
964 | } |
965 | return *this; |
966 | } |
967 | |
968 | //---------------------------------------------------------------- |
969 | // Property set patterns |
970 | //---------------------------------------------------------------- |
971 | |
972 | /** |
973 | * Return true if the given position, in the given pattern, appears |
974 | * to be the start of a property set pattern. |
975 | */ |
976 | UBool UnicodeSet::resemblesPropertyPattern(const UnicodeString& pattern, |
977 | int32_t pos) { |
978 | // Patterns are at least 5 characters long |
979 | if ((pos+5) > pattern.length()) { |
980 | return false; |
981 | } |
982 | |
983 | // Look for an opening [:, [:^, \p, or \P |
984 | return isPOSIXOpen(pattern, pos) || isPerlOpen(pattern, pos) || isNameOpen(pattern, pos); |
985 | } |
986 | |
987 | /** |
988 | * Return true if the given iterator appears to point at a |
989 | * property pattern. Regardless of the result, return with the |
990 | * iterator unchanged. |
991 | * @param chars iterator over the pattern characters. Upon return |
992 | * it will be unchanged. |
993 | * @param iterOpts RuleCharacterIterator options |
994 | */ |
995 | UBool UnicodeSet::resemblesPropertyPattern(RuleCharacterIterator& chars, |
996 | int32_t iterOpts) { |
997 | // NOTE: literal will always be false, because we don't parse escapes. |
998 | UBool result = false, literal; |
999 | UErrorCode ec = U_ZERO_ERROR; |
1000 | iterOpts &= ~RuleCharacterIterator::PARSE_ESCAPES; |
1001 | RuleCharacterIterator::Pos pos; |
1002 | chars.getPos(pos); |
1003 | UChar32 c = chars.next(iterOpts, literal, ec); |
1004 | if (c == u'[' || c == u'\\') { |
1005 | UChar32 d = chars.next(iterOpts & ~RuleCharacterIterator::SKIP_WHITESPACE, |
1006 | literal, ec); |
1007 | result = (c == u'[') ? (d == u':') : |
1008 | (d == u'N' || d == u'p' || d == u'P'); |
1009 | } |
1010 | chars.setPos(pos); |
1011 | return result && U_SUCCESS(ec); |
1012 | } |
1013 | |
1014 | /** |
1015 | * Parse the given property pattern at the given parse position. |
1016 | */ |
1017 | UnicodeSet& UnicodeSet::applyPropertyPattern(const UnicodeString& pattern, |
1018 | ParsePosition& ppos, |
1019 | UErrorCode &ec) { |
1020 | int32_t pos = ppos.getIndex(); |
1021 | |
1022 | UBool posix = false; // true for [:pat:], false for \p{pat} \P{pat} \N{pat} |
1023 | UBool isName = false; // true for \N{pat}, o/w false |
1024 | UBool invert = false; |
1025 | |
1026 | if (U_FAILURE(ec)) return *this; |
1027 | |
1028 | // Minimum length is 5 characters, e.g. \p{L} |
1029 | if ((pos+5) > pattern.length()) { |
1030 | FAIL(ec); |
1031 | } |
1032 | |
1033 | // On entry, ppos should point to one of the following locations: |
1034 | // Look for an opening [:, [:^, \p, or \P |
1035 | if (isPOSIXOpen(pattern, pos)) { |
1036 | posix = true; |
1037 | pos += 2; |
1038 | pos = ICU_Utility::skipWhitespace(pattern, pos); |
1039 | if (pos < pattern.length() && pattern.charAt(pos) == u'^') { |
1040 | ++pos; |
1041 | invert = true; |
1042 | } |
1043 | } else if (isPerlOpen(pattern, pos) || isNameOpen(pattern, pos)) { |
1044 | char16_t c = pattern.charAt(pos+1); |
1045 | invert = (c == u'P'); |
1046 | isName = (c == u'N'); |
1047 | pos += 2; |
1048 | pos = ICU_Utility::skipWhitespace(pattern, pos); |
1049 | if (pos == pattern.length() || pattern.charAt(pos++) != u'{') { |
1050 | // Syntax error; "\p" or "\P" not followed by "{" |
1051 | FAIL(ec); |
1052 | } |
1053 | } else { |
1054 | // Open delimiter not seen |
1055 | FAIL(ec); |
1056 | } |
1057 | |
1058 | // Look for the matching close delimiter, either :] or } |
1059 | int32_t close; |
1060 | if (posix) { |
1061 | close = pattern.indexOf(u":]" , 2, pos); |
1062 | } else { |
1063 | close = pattern.indexOf(u'}', pos); |
1064 | } |
1065 | if (close < 0) { |
1066 | // Syntax error; close delimiter missing |
1067 | FAIL(ec); |
1068 | } |
1069 | |
1070 | // Look for an '=' sign. If this is present, we will parse a |
1071 | // medium \p{gc=Cf} or long \p{GeneralCategory=Format} |
1072 | // pattern. |
1073 | int32_t equals = pattern.indexOf(u'=', pos); |
1074 | UnicodeString propName, valueName; |
1075 | if (equals >= 0 && equals < close && !isName) { |
1076 | // Equals seen; parse medium/long pattern |
1077 | pattern.extractBetween(pos, equals, propName); |
1078 | pattern.extractBetween(equals+1, close, valueName); |
1079 | } |
1080 | |
1081 | else { |
1082 | // Handle case where no '=' is seen, and \N{} |
1083 | pattern.extractBetween(pos, close, propName); |
1084 | |
1085 | // Handle \N{name} |
1086 | if (isName) { |
1087 | // This is a little inefficient since it means we have to |
1088 | // parse NAME_PROP back to UCHAR_NAME even though we already |
1089 | // know it's UCHAR_NAME. If we refactor the API to |
1090 | // support args of (UProperty, char*) then we can remove |
1091 | // NAME_PROP and make this a little more efficient. |
1092 | valueName = propName; |
1093 | propName = UnicodeString(NAME_PROP, NAME_PROP_LENGTH, US_INV); |
1094 | } |
1095 | } |
1096 | |
1097 | applyPropertyAlias(propName, valueName, ec); |
1098 | |
1099 | if (U_SUCCESS(ec)) { |
1100 | if (invert) { |
1101 | complement().removeAllStrings(); // code point complement |
1102 | } |
1103 | |
1104 | // Move to the limit position after the close delimiter if the |
1105 | // parse succeeded. |
1106 | ppos.setIndex(close + (posix ? 2 : 1)); |
1107 | } |
1108 | |
1109 | return *this; |
1110 | } |
1111 | |
1112 | /** |
1113 | * Parse a property pattern. |
1114 | * @param chars iterator over the pattern characters. Upon return |
1115 | * it will be advanced to the first character after the parsed |
1116 | * pattern, or the end of the iteration if all characters are |
1117 | * parsed. |
1118 | * @param rebuiltPat the pattern that was parsed, rebuilt or |
1119 | * copied from the input pattern, as appropriate. |
1120 | */ |
1121 | void UnicodeSet::applyPropertyPattern(RuleCharacterIterator& chars, |
1122 | UnicodeString& rebuiltPat, |
1123 | UErrorCode& ec) { |
1124 | if (U_FAILURE(ec)) return; |
1125 | UnicodeString pattern; |
1126 | chars.lookahead(pattern); |
1127 | ParsePosition pos(0); |
1128 | applyPropertyPattern(pattern, pos, ec); |
1129 | if (U_FAILURE(ec)) return; |
1130 | if (pos.getIndex() == 0) { |
1131 | // syntaxError(chars, "Invalid property pattern"); |
1132 | ec = U_MALFORMED_SET; |
1133 | return; |
1134 | } |
1135 | chars.jumpahead(pos.getIndex()); |
1136 | rebuiltPat.append(pattern, 0, pos.getIndex()); |
1137 | } |
1138 | |
1139 | U_NAMESPACE_END |
1140 | |