1 | /* |
2 | * Copyright (C) 2009, 2013-2016 Apple Inc. All rights reserved. |
3 | * Copyright (C) 2010 Peter Varga ([email protected]), University of Szeged |
4 | * |
5 | * Redistribution and use in source and binary forms, with or without |
6 | * modification, are permitted provided that the following conditions |
7 | * are met: |
8 | * 1. Redistributions of source code must retain the above copyright |
9 | * notice, this list of conditions and the following disclaimer. |
10 | * 2. Redistributions in binary form must reproduce the above copyright |
11 | * notice, this list of conditions and the following disclaimer in the |
12 | * documentation and/or other materials provided with the distribution. |
13 | * |
14 | * THIS SOFTWARE IS PROVIDED BY APPLE INC. ``AS IS'' AND ANY |
15 | * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
16 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR |
17 | * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL APPLE INC. OR |
18 | * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, |
19 | * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, |
20 | * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR |
21 | * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY |
22 | * OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
23 | * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE |
24 | * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
25 | */ |
26 | |
27 | #include "config.h" |
28 | #include "YarrPattern.h" |
29 | |
30 | #include "Options.h" |
31 | #include "Yarr.h" |
32 | #include "YarrCanonicalize.h" |
33 | #include "YarrParser.h" |
34 | #include <wtf/DataLog.h> |
35 | #include <wtf/Optional.h> |
36 | #include <wtf/StackPointer.h> |
37 | #include <wtf/Threading.h> |
38 | #include <wtf/Vector.h> |
39 | |
40 | namespace JSC { namespace Yarr { |
41 | |
42 | #include "RegExpJitTables.h" |
43 | |
44 | class CharacterClassConstructor { |
45 | public: |
46 | CharacterClassConstructor(bool isCaseInsensitive, CanonicalMode canonicalMode) |
47 | : m_isCaseInsensitive(isCaseInsensitive) |
48 | , m_anyCharacter(false) |
49 | , m_characterWidths(CharacterClassWidths::Unknown) |
50 | , m_canonicalMode(canonicalMode) |
51 | { |
52 | } |
53 | |
54 | void reset() |
55 | { |
56 | m_matches.clear(); |
57 | m_ranges.clear(); |
58 | m_matchesUnicode.clear(); |
59 | m_rangesUnicode.clear(); |
60 | m_anyCharacter = false; |
61 | m_characterWidths = CharacterClassWidths::Unknown; |
62 | } |
63 | |
64 | void append(const CharacterClass* other) |
65 | { |
66 | for (size_t i = 0; i < other->m_matches.size(); ++i) |
67 | addSorted(m_matches, other->m_matches[i]); |
68 | for (size_t i = 0; i < other->m_ranges.size(); ++i) |
69 | addSortedRange(m_ranges, other->m_ranges[i].begin, other->m_ranges[i].end); |
70 | for (size_t i = 0; i < other->m_matchesUnicode.size(); ++i) |
71 | addSorted(m_matchesUnicode, other->m_matchesUnicode[i]); |
72 | for (size_t i = 0; i < other->m_rangesUnicode.size(); ++i) |
73 | addSortedRange(m_rangesUnicode, other->m_rangesUnicode[i].begin, other->m_rangesUnicode[i].end); |
74 | } |
75 | |
76 | void appendInverted(const CharacterClass* other) |
77 | { |
78 | auto addSortedInverted = [&](UChar32 min, UChar32 max, |
79 | const Vector<UChar32>& srcMatches, const Vector<CharacterRange>& srcRanges, |
80 | Vector<UChar32>& destMatches, Vector<CharacterRange>& destRanges) { |
81 | |
82 | auto addSortedMatchOrRange = [&](UChar32 lo, UChar32 hiPlusOne) { |
83 | if (lo < hiPlusOne) { |
84 | if (lo + 1 == hiPlusOne) |
85 | addSorted(destMatches, lo); |
86 | else |
87 | addSortedRange(destRanges, lo, hiPlusOne - 1); |
88 | } |
89 | }; |
90 | |
91 | UChar32 lo = min; |
92 | size_t matchesIndex = 0; |
93 | size_t rangesIndex = 0; |
94 | bool matchesRemaining = matchesIndex < srcMatches.size(); |
95 | bool rangesRemaining = rangesIndex < srcRanges.size(); |
96 | |
97 | if (!matchesRemaining && !rangesRemaining) { |
98 | addSortedMatchOrRange(min, max + 1); |
99 | return; |
100 | } |
101 | |
102 | while (matchesRemaining || rangesRemaining) { |
103 | UChar32 hiPlusOne; |
104 | UChar32 nextLo; |
105 | |
106 | if (matchesRemaining |
107 | && (!rangesRemaining || srcMatches[matchesIndex] < srcRanges[rangesIndex].begin)) { |
108 | hiPlusOne = srcMatches[matchesIndex]; |
109 | nextLo = hiPlusOne + 1; |
110 | ++matchesIndex; |
111 | matchesRemaining = matchesIndex < srcMatches.size(); |
112 | } else { |
113 | hiPlusOne = srcRanges[rangesIndex].begin; |
114 | nextLo = srcRanges[rangesIndex].end + 1; |
115 | ++rangesIndex; |
116 | rangesRemaining = rangesIndex < srcRanges.size(); |
117 | } |
118 | |
119 | addSortedMatchOrRange(lo, hiPlusOne); |
120 | |
121 | lo = nextLo; |
122 | } |
123 | |
124 | addSortedMatchOrRange(lo, max + 1); |
125 | }; |
126 | |
127 | addSortedInverted(0, 0x7f, other->m_matches, other->m_ranges, m_matches, m_ranges); |
128 | addSortedInverted(0x80, 0x10ffff, other->m_matchesUnicode, other->m_rangesUnicode, m_matchesUnicode, m_rangesUnicode); |
129 | } |
130 | |
131 | void putChar(UChar32 ch) |
132 | { |
133 | if (!m_isCaseInsensitive) { |
134 | addSorted(ch); |
135 | return; |
136 | } |
137 | |
138 | if (m_canonicalMode == CanonicalMode::UCS2 && isASCII(ch)) { |
139 | // Handle ASCII cases. |
140 | if (isASCIIAlpha(ch)) { |
141 | addSorted(m_matches, toASCIIUpper(ch)); |
142 | addSorted(m_matches, toASCIILower(ch)); |
143 | } else |
144 | addSorted(m_matches, ch); |
145 | return; |
146 | } |
147 | |
148 | // Add multiple matches, if necessary. |
149 | const CanonicalizationRange* info = canonicalRangeInfoFor(ch, m_canonicalMode); |
150 | if (info->type == CanonicalizeUnique) |
151 | addSorted(ch); |
152 | else |
153 | putUnicodeIgnoreCase(ch, info); |
154 | } |
155 | |
156 | void putUnicodeIgnoreCase(UChar32 ch, const CanonicalizationRange* info) |
157 | { |
158 | ASSERT(m_isCaseInsensitive); |
159 | ASSERT(ch >= info->begin && ch <= info->end); |
160 | ASSERT(info->type != CanonicalizeUnique); |
161 | if (info->type == CanonicalizeSet) { |
162 | for (const UChar32* set = canonicalCharacterSetInfo(info->value, m_canonicalMode); (ch = *set); ++set) |
163 | addSorted(ch); |
164 | } else { |
165 | addSorted(ch); |
166 | addSorted(getCanonicalPair(info, ch)); |
167 | } |
168 | } |
169 | |
170 | void putRange(UChar32 lo, UChar32 hi) |
171 | { |
172 | if (isASCII(lo)) { |
173 | char asciiLo = lo; |
174 | char asciiHi = std::min(hi, (UChar32)0x7f); |
175 | addSortedRange(m_ranges, lo, asciiHi); |
176 | |
177 | if (m_isCaseInsensitive) { |
178 | if ((asciiLo <= 'Z') && (asciiHi >= 'A')) |
179 | addSortedRange(m_ranges, std::max(asciiLo, 'A')+('a'-'A'), std::min(asciiHi, 'Z')+('a'-'A')); |
180 | if ((asciiLo <= 'z') && (asciiHi >= 'a')) |
181 | addSortedRange(m_ranges, std::max(asciiLo, 'a')+('A'-'a'), std::min(asciiHi, 'z')+('A'-'a')); |
182 | } |
183 | } |
184 | if (isASCII(hi)) |
185 | return; |
186 | |
187 | lo = std::max(lo, (UChar32)0x80); |
188 | addSortedRange(m_rangesUnicode, lo, hi); |
189 | |
190 | if (!m_isCaseInsensitive) |
191 | return; |
192 | |
193 | const CanonicalizationRange* info = canonicalRangeInfoFor(lo, m_canonicalMode); |
194 | while (true) { |
195 | // Handle the range [lo .. end] |
196 | UChar32 end = std::min<UChar32>(info->end, hi); |
197 | |
198 | switch (info->type) { |
199 | case CanonicalizeUnique: |
200 | // Nothing to do - no canonical equivalents. |
201 | break; |
202 | case CanonicalizeSet: { |
203 | UChar ch; |
204 | for (const UChar32* set = canonicalCharacterSetInfo(info->value, m_canonicalMode); (ch = *set); ++set) |
205 | addSorted(m_matchesUnicode, ch); |
206 | break; |
207 | } |
208 | case CanonicalizeRangeLo: |
209 | addSortedRange(m_rangesUnicode, lo + info->value, end + info->value); |
210 | break; |
211 | case CanonicalizeRangeHi: |
212 | addSortedRange(m_rangesUnicode, lo - info->value, end - info->value); |
213 | break; |
214 | case CanonicalizeAlternatingAligned: |
215 | // Use addSortedRange since there is likely an abutting range to combine with. |
216 | if (lo & 1) |
217 | addSortedRange(m_rangesUnicode, lo - 1, lo - 1); |
218 | if (!(end & 1)) |
219 | addSortedRange(m_rangesUnicode, end + 1, end + 1); |
220 | break; |
221 | case CanonicalizeAlternatingUnaligned: |
222 | // Use addSortedRange since there is likely an abutting range to combine with. |
223 | if (!(lo & 1)) |
224 | addSortedRange(m_rangesUnicode, lo - 1, lo - 1); |
225 | if (end & 1) |
226 | addSortedRange(m_rangesUnicode, end + 1, end + 1); |
227 | break; |
228 | } |
229 | |
230 | if (hi == end) |
231 | return; |
232 | |
233 | ++info; |
234 | lo = info->begin; |
235 | }; |
236 | |
237 | } |
238 | |
239 | std::unique_ptr<CharacterClass> charClass() |
240 | { |
241 | coalesceTables(); |
242 | |
243 | auto characterClass = std::make_unique<CharacterClass>(); |
244 | |
245 | characterClass->m_matches.swap(m_matches); |
246 | characterClass->m_ranges.swap(m_ranges); |
247 | characterClass->m_matchesUnicode.swap(m_matchesUnicode); |
248 | characterClass->m_rangesUnicode.swap(m_rangesUnicode); |
249 | characterClass->m_anyCharacter = anyCharacter(); |
250 | characterClass->m_characterWidths = characterWidths(); |
251 | |
252 | m_anyCharacter = false; |
253 | m_characterWidths = CharacterClassWidths::Unknown; |
254 | |
255 | return characterClass; |
256 | } |
257 | |
258 | private: |
259 | void addSorted(UChar32 ch) |
260 | { |
261 | addSorted(isASCII(ch) ? m_matches : m_matchesUnicode, ch); |
262 | } |
263 | |
264 | void addSorted(Vector<UChar32>& matches, UChar32 ch) |
265 | { |
266 | unsigned pos = 0; |
267 | unsigned range = matches.size(); |
268 | |
269 | m_characterWidths |= (U_IS_BMP(ch) ? CharacterClassWidths::HasBMPChars : CharacterClassWidths::HasNonBMPChars); |
270 | |
271 | // binary chop, find position to insert char. |
272 | while (range) { |
273 | unsigned index = range >> 1; |
274 | |
275 | int val = matches[pos+index] - ch; |
276 | if (!val) |
277 | return; |
278 | else if (val > 0) { |
279 | if (val == 1) { |
280 | UChar32 lo = ch; |
281 | UChar32 hi = ch + 1; |
282 | matches.remove(pos + index); |
283 | if (pos + index > 0 && matches[pos + index - 1] == ch - 1) { |
284 | lo = ch - 1; |
285 | matches.remove(pos + index - 1); |
286 | } |
287 | addSortedRange(isASCII(ch) ? m_ranges : m_rangesUnicode, lo, hi); |
288 | return; |
289 | } |
290 | range = index; |
291 | } else { |
292 | if (val == -1) { |
293 | UChar32 lo = ch - 1; |
294 | UChar32 hi = ch; |
295 | matches.remove(pos + index); |
296 | if (pos + index + 1 < matches.size() && matches[pos + index + 1] == ch + 1) { |
297 | hi = ch + 1; |
298 | matches.remove(pos + index + 1); |
299 | } |
300 | addSortedRange(isASCII(ch) ? m_ranges : m_rangesUnicode, lo, hi); |
301 | return; |
302 | } |
303 | pos += (index+1); |
304 | range -= (index+1); |
305 | } |
306 | } |
307 | |
308 | if (pos == matches.size()) |
309 | matches.append(ch); |
310 | else |
311 | matches.insert(pos, ch); |
312 | } |
313 | |
314 | void addSortedRange(Vector<CharacterRange>& ranges, UChar32 lo, UChar32 hi) |
315 | { |
316 | size_t end = ranges.size(); |
317 | |
318 | if (U_IS_BMP(lo)) |
319 | m_characterWidths |= CharacterClassWidths::HasBMPChars; |
320 | if (!U_IS_BMP(hi)) |
321 | m_characterWidths |= CharacterClassWidths::HasNonBMPChars; |
322 | |
323 | // Simple linear scan - I doubt there are that many ranges anyway... |
324 | // feel free to fix this with something faster (eg binary chop). |
325 | for (size_t i = 0; i < end; ++i) { |
326 | // does the new range fall before the current position in the array |
327 | if (hi < ranges[i].begin) { |
328 | // Concatenate appending ranges. |
329 | if (hi == (ranges[i].begin - 1)) { |
330 | ranges[i].begin = lo; |
331 | return; |
332 | } |
333 | ranges.insert(i, CharacterRange(lo, hi)); |
334 | return; |
335 | } |
336 | // Okay, since we didn't hit the last case, the end of the new range is definitely at or after the begining |
337 | // If the new range start at or before the end of the last range, then the overlap (if it starts one after the |
338 | // end of the last range they concatenate, which is just as good. |
339 | if (lo <= (ranges[i].end + 1)) { |
340 | // found an intersect! we'll replace this entry in the array. |
341 | ranges[i].begin = std::min(ranges[i].begin, lo); |
342 | ranges[i].end = std::max(ranges[i].end, hi); |
343 | |
344 | mergeRangesFrom(ranges, i); |
345 | return; |
346 | } |
347 | } |
348 | |
349 | // CharacterRange comes after all existing ranges. |
350 | ranges.append(CharacterRange(lo, hi)); |
351 | } |
352 | |
353 | void mergeRangesFrom(Vector<CharacterRange>& ranges, size_t index) |
354 | { |
355 | unsigned next = index + 1; |
356 | |
357 | // each iteration of the loop we will either remove something from the list, or break out of the loop. |
358 | while (next < ranges.size()) { |
359 | if (ranges[next].begin <= (ranges[index].end + 1)) { |
360 | // the next entry now overlaps / concatenates with this one. |
361 | ranges[index].end = std::max(ranges[index].end, ranges[next].end); |
362 | ranges.remove(next); |
363 | } else |
364 | break; |
365 | } |
366 | |
367 | } |
368 | |
369 | void coalesceTables() |
370 | { |
371 | auto coalesceMatchesAndRanges = [&](Vector<UChar32>& matches, Vector<CharacterRange>& ranges) { |
372 | |
373 | size_t matchesIndex = 0; |
374 | size_t rangesIndex = 0; |
375 | |
376 | while (matchesIndex < matches.size() && rangesIndex < ranges.size()) { |
377 | while (matchesIndex < matches.size() && matches[matchesIndex] < ranges[rangesIndex].begin - 1) |
378 | matchesIndex++; |
379 | |
380 | if (matchesIndex < matches.size() && matches[matchesIndex] == ranges[rangesIndex].begin - 1) { |
381 | ranges[rangesIndex].begin = matches[matchesIndex]; |
382 | matches.remove(matchesIndex); |
383 | } |
384 | |
385 | while (matchesIndex < matches.size() && matches[matchesIndex] < ranges[rangesIndex].end + 1) |
386 | matchesIndex++; |
387 | |
388 | if (matchesIndex < matches.size()) { |
389 | if (matches[matchesIndex] == ranges[rangesIndex].end + 1) { |
390 | ranges[rangesIndex].end = matches[matchesIndex]; |
391 | matches.remove(matchesIndex); |
392 | |
393 | mergeRangesFrom(ranges, rangesIndex); |
394 | } else |
395 | matchesIndex++; |
396 | } |
397 | } |
398 | }; |
399 | |
400 | coalesceMatchesAndRanges(m_matches, m_ranges); |
401 | coalesceMatchesAndRanges(m_matchesUnicode, m_rangesUnicode); |
402 | |
403 | if (!m_matches.size() && !m_matchesUnicode.size() |
404 | && m_ranges.size() == 1 && m_rangesUnicode.size() == 1 |
405 | && m_ranges[0].begin == 0 && m_ranges[0].end == 0x7f |
406 | && m_rangesUnicode[0].begin == 0x80 && m_rangesUnicode[0].end == 0x10ffff) |
407 | m_anyCharacter = true; |
408 | } |
409 | |
410 | bool hasNonBMPCharacters() |
411 | { |
412 | return m_characterWidths & CharacterClassWidths::HasNonBMPChars; |
413 | } |
414 | |
415 | CharacterClassWidths characterWidths() |
416 | { |
417 | return m_characterWidths; |
418 | } |
419 | |
420 | bool anyCharacter() |
421 | { |
422 | return m_anyCharacter; |
423 | } |
424 | |
425 | bool m_isCaseInsensitive : 1; |
426 | bool m_anyCharacter : 1; |
427 | CharacterClassWidths m_characterWidths; |
428 | |
429 | CanonicalMode m_canonicalMode; |
430 | |
431 | Vector<UChar32> m_matches; |
432 | Vector<CharacterRange> m_ranges; |
433 | Vector<UChar32> m_matchesUnicode; |
434 | Vector<CharacterRange> m_rangesUnicode; |
435 | }; |
436 | |
437 | class YarrPatternConstructor { |
438 | public: |
439 | YarrPatternConstructor(YarrPattern& pattern, void* stackLimit) |
440 | : m_pattern(pattern) |
441 | , m_characterClassConstructor(pattern.ignoreCase(), pattern.unicode() ? CanonicalMode::Unicode : CanonicalMode::UCS2) |
442 | , m_stackLimit(stackLimit) |
443 | { |
444 | auto body = std::make_unique<PatternDisjunction>(); |
445 | m_pattern.m_body = body.get(); |
446 | m_alternative = body->addNewAlternative(); |
447 | m_pattern.m_disjunctions.append(WTFMove(body)); |
448 | } |
449 | |
450 | ~YarrPatternConstructor() |
451 | { |
452 | } |
453 | |
454 | void resetForReparsing() |
455 | { |
456 | m_pattern.resetForReparsing(); |
457 | m_characterClassConstructor.reset(); |
458 | |
459 | auto body = std::make_unique<PatternDisjunction>(); |
460 | m_pattern.m_body = body.get(); |
461 | m_alternative = body->addNewAlternative(); |
462 | m_pattern.m_disjunctions.append(WTFMove(body)); |
463 | } |
464 | |
465 | void saveUnmatchedNamedForwardReferences() |
466 | { |
467 | m_unmatchedNamedForwardReferences.shrink(0); |
468 | |
469 | for (auto& entry : m_pattern.m_namedForwardReferences) { |
470 | if (!m_pattern.m_captureGroupNames.contains(entry)) |
471 | m_unmatchedNamedForwardReferences.append(entry); |
472 | } |
473 | } |
474 | |
475 | void assertionBOL() |
476 | { |
477 | if (!m_alternative->m_terms.size() && !m_invertParentheticalAssertion) { |
478 | m_alternative->m_startsWithBOL = true; |
479 | m_alternative->m_containsBOL = true; |
480 | m_pattern.m_containsBOL = true; |
481 | } |
482 | m_alternative->m_terms.append(PatternTerm::BOL()); |
483 | } |
484 | void assertionEOL() |
485 | { |
486 | m_alternative->m_terms.append(PatternTerm::EOL()); |
487 | } |
488 | void assertionWordBoundary(bool invert) |
489 | { |
490 | m_alternative->m_terms.append(PatternTerm::WordBoundary(invert)); |
491 | } |
492 | |
493 | void atomPatternCharacter(UChar32 ch) |
494 | { |
495 | // We handle case-insensitive checking of unicode characters which do have both |
496 | // cases by handling them as if they were defined using a CharacterClass. |
497 | if (!m_pattern.ignoreCase() || (isASCII(ch) && !m_pattern.unicode())) { |
498 | m_alternative->m_terms.append(PatternTerm(ch)); |
499 | return; |
500 | } |
501 | |
502 | const CanonicalizationRange* info = canonicalRangeInfoFor(ch, m_pattern.unicode() ? CanonicalMode::Unicode : CanonicalMode::UCS2); |
503 | if (info->type == CanonicalizeUnique) { |
504 | m_alternative->m_terms.append(PatternTerm(ch)); |
505 | return; |
506 | } |
507 | |
508 | m_characterClassConstructor.putUnicodeIgnoreCase(ch, info); |
509 | auto newCharacterClass = m_characterClassConstructor.charClass(); |
510 | m_alternative->m_terms.append(PatternTerm(newCharacterClass.get(), false)); |
511 | m_pattern.m_userCharacterClasses.append(WTFMove(newCharacterClass)); |
512 | } |
513 | |
514 | void atomBuiltInCharacterClass(BuiltInCharacterClassID classID, bool invert) |
515 | { |
516 | switch (classID) { |
517 | case BuiltInCharacterClassID::DigitClassID: |
518 | m_alternative->m_terms.append(PatternTerm(m_pattern.digitsCharacterClass(), invert)); |
519 | break; |
520 | case BuiltInCharacterClassID::SpaceClassID: |
521 | m_alternative->m_terms.append(PatternTerm(m_pattern.spacesCharacterClass(), invert)); |
522 | break; |
523 | case BuiltInCharacterClassID::WordClassID: |
524 | if (m_pattern.unicode() && m_pattern.ignoreCase()) |
525 | m_alternative->m_terms.append(PatternTerm(m_pattern.wordUnicodeIgnoreCaseCharCharacterClass(), invert)); |
526 | else |
527 | m_alternative->m_terms.append(PatternTerm(m_pattern.wordcharCharacterClass(), invert)); |
528 | break; |
529 | case BuiltInCharacterClassID::DotClassID: |
530 | ASSERT(!invert); |
531 | if (m_pattern.dotAll()) |
532 | m_alternative->m_terms.append(PatternTerm(m_pattern.anyCharacterClass(), false)); |
533 | else |
534 | m_alternative->m_terms.append(PatternTerm(m_pattern.newlineCharacterClass(), true)); |
535 | break; |
536 | default: |
537 | m_alternative->m_terms.append(PatternTerm(m_pattern.unicodeCharacterClassFor(classID), invert)); |
538 | break; |
539 | } |
540 | } |
541 | |
542 | void atomCharacterClassBegin(bool invert = false) |
543 | { |
544 | m_invertCharacterClass = invert; |
545 | } |
546 | |
547 | void atomCharacterClassAtom(UChar32 ch) |
548 | { |
549 | m_characterClassConstructor.putChar(ch); |
550 | } |
551 | |
552 | void atomCharacterClassRange(UChar32 begin, UChar32 end) |
553 | { |
554 | m_characterClassConstructor.putRange(begin, end); |
555 | } |
556 | |
557 | void atomCharacterClassBuiltIn(BuiltInCharacterClassID classID, bool invert) |
558 | { |
559 | ASSERT(classID != BuiltInCharacterClassID::DotClassID); |
560 | |
561 | switch (classID) { |
562 | case BuiltInCharacterClassID::DigitClassID: |
563 | m_characterClassConstructor.append(invert ? m_pattern.nondigitsCharacterClass() : m_pattern.digitsCharacterClass()); |
564 | break; |
565 | |
566 | case BuiltInCharacterClassID::SpaceClassID: |
567 | m_characterClassConstructor.append(invert ? m_pattern.nonspacesCharacterClass() : m_pattern.spacesCharacterClass()); |
568 | break; |
569 | |
570 | case BuiltInCharacterClassID::WordClassID: |
571 | if (m_pattern.unicode() && m_pattern.ignoreCase()) |
572 | m_characterClassConstructor.append(invert ? m_pattern.nonwordUnicodeIgnoreCaseCharCharacterClass() : m_pattern.wordUnicodeIgnoreCaseCharCharacterClass()); |
573 | else |
574 | m_characterClassConstructor.append(invert ? m_pattern.nonwordcharCharacterClass() : m_pattern.wordcharCharacterClass()); |
575 | break; |
576 | |
577 | default: |
578 | if (!invert) |
579 | m_characterClassConstructor.append(m_pattern.unicodeCharacterClassFor(classID)); |
580 | else |
581 | m_characterClassConstructor.appendInverted(m_pattern.unicodeCharacterClassFor(classID)); |
582 | } |
583 | } |
584 | |
585 | void atomCharacterClassEnd() |
586 | { |
587 | auto newCharacterClass = m_characterClassConstructor.charClass(); |
588 | |
589 | if (!m_invertCharacterClass && newCharacterClass.get()->m_anyCharacter) { |
590 | m_alternative->m_terms.append(PatternTerm(m_pattern.anyCharacterClass(), false)); |
591 | return; |
592 | } |
593 | m_alternative->m_terms.append(PatternTerm(newCharacterClass.get(), m_invertCharacterClass)); |
594 | m_pattern.m_userCharacterClasses.append(WTFMove(newCharacterClass)); |
595 | } |
596 | |
597 | void atomParenthesesSubpatternBegin(bool capture = true, Optional<String> optGroupName = WTF::nullopt) |
598 | { |
599 | unsigned subpatternId = m_pattern.m_numSubpatterns + 1; |
600 | if (capture) { |
601 | m_pattern.m_numSubpatterns++; |
602 | if (optGroupName) { |
603 | while (m_pattern.m_captureGroupNames.size() < subpatternId) |
604 | m_pattern.m_captureGroupNames.append(String()); |
605 | m_pattern.m_captureGroupNames.append(optGroupName.value()); |
606 | m_pattern.m_namedGroupToParenIndex.add(optGroupName.value(), subpatternId); |
607 | } |
608 | } else |
609 | ASSERT(!optGroupName); |
610 | |
611 | auto parenthesesDisjunction = std::make_unique<PatternDisjunction>(m_alternative); |
612 | m_alternative->m_terms.append(PatternTerm(PatternTerm::TypeParenthesesSubpattern, subpatternId, parenthesesDisjunction.get(), capture, false)); |
613 | m_alternative = parenthesesDisjunction->addNewAlternative(); |
614 | m_pattern.m_disjunctions.append(WTFMove(parenthesesDisjunction)); |
615 | } |
616 | |
617 | void atomParentheticalAssertionBegin(bool invert = false) |
618 | { |
619 | auto parenthesesDisjunction = std::make_unique<PatternDisjunction>(m_alternative); |
620 | m_alternative->m_terms.append(PatternTerm(PatternTerm::TypeParentheticalAssertion, m_pattern.m_numSubpatterns + 1, parenthesesDisjunction.get(), false, invert)); |
621 | m_alternative = parenthesesDisjunction->addNewAlternative(); |
622 | m_invertParentheticalAssertion = invert; |
623 | m_pattern.m_disjunctions.append(WTFMove(parenthesesDisjunction)); |
624 | } |
625 | |
626 | void atomParenthesesEnd() |
627 | { |
628 | ASSERT(m_alternative->m_parent); |
629 | ASSERT(m_alternative->m_parent->m_parent); |
630 | |
631 | PatternDisjunction* parenthesesDisjunction = m_alternative->m_parent; |
632 | m_alternative = m_alternative->m_parent->m_parent; |
633 | |
634 | PatternTerm& lastTerm = m_alternative->lastTerm(); |
635 | |
636 | unsigned numParenAlternatives = parenthesesDisjunction->m_alternatives.size(); |
637 | unsigned numBOLAnchoredAlts = 0; |
638 | |
639 | for (unsigned i = 0; i < numParenAlternatives; i++) { |
640 | // Bubble up BOL flags |
641 | if (parenthesesDisjunction->m_alternatives[i]->m_startsWithBOL) |
642 | numBOLAnchoredAlts++; |
643 | } |
644 | |
645 | if (numBOLAnchoredAlts) { |
646 | m_alternative->m_containsBOL = true; |
647 | // If all the alternatives in parens start with BOL, then so does this one |
648 | if (numBOLAnchoredAlts == numParenAlternatives) |
649 | m_alternative->m_startsWithBOL = true; |
650 | } |
651 | |
652 | lastTerm.parentheses.lastSubpatternId = m_pattern.m_numSubpatterns; |
653 | m_invertParentheticalAssertion = false; |
654 | } |
655 | |
656 | void atomBackReference(unsigned subpatternId) |
657 | { |
658 | ASSERT(subpatternId); |
659 | m_pattern.m_containsBackreferences = true; |
660 | m_pattern.m_maxBackReference = std::max(m_pattern.m_maxBackReference, subpatternId); |
661 | |
662 | if (subpatternId > m_pattern.m_numSubpatterns) { |
663 | m_alternative->m_terms.append(PatternTerm::ForwardReference()); |
664 | return; |
665 | } |
666 | |
667 | PatternAlternative* currentAlternative = m_alternative; |
668 | ASSERT(currentAlternative); |
669 | |
670 | // Note to self: if we waited until the AST was baked, we could also remove forwards refs |
671 | while ((currentAlternative = currentAlternative->m_parent->m_parent)) { |
672 | PatternTerm& term = currentAlternative->lastTerm(); |
673 | ASSERT((term.type == PatternTerm::TypeParenthesesSubpattern) || (term.type == PatternTerm::TypeParentheticalAssertion)); |
674 | |
675 | if ((term.type == PatternTerm::TypeParenthesesSubpattern) && term.capture() && (subpatternId == term.parentheses.subpatternId)) { |
676 | m_alternative->m_terms.append(PatternTerm::ForwardReference()); |
677 | return; |
678 | } |
679 | } |
680 | |
681 | m_alternative->m_terms.append(PatternTerm(subpatternId)); |
682 | } |
683 | |
684 | void atomNamedBackReference(const String& subpatternName) |
685 | { |
686 | ASSERT(m_pattern.m_namedGroupToParenIndex.find(subpatternName) != m_pattern.m_namedGroupToParenIndex.end()); |
687 | atomBackReference(m_pattern.m_namedGroupToParenIndex.get(subpatternName)); |
688 | } |
689 | |
690 | bool isValidNamedForwardReference(const String& subpatternName) |
691 | { |
692 | return !m_unmatchedNamedForwardReferences.contains(subpatternName); |
693 | } |
694 | |
695 | void atomNamedForwardReference(const String& subpatternName) |
696 | { |
697 | m_pattern.m_namedForwardReferences.appendIfNotContains(subpatternName); |
698 | m_alternative->m_terms.append(PatternTerm::ForwardReference()); |
699 | } |
700 | |
701 | // deep copy the argument disjunction. If filterStartsWithBOL is true, |
702 | // skip alternatives with m_startsWithBOL set true. |
703 | PatternDisjunction* copyDisjunction(PatternDisjunction* disjunction, bool filterStartsWithBOL = false) |
704 | { |
705 | std::unique_ptr<PatternDisjunction> newDisjunction; |
706 | for (unsigned alt = 0; alt < disjunction->m_alternatives.size(); ++alt) { |
707 | PatternAlternative* alternative = disjunction->m_alternatives[alt].get(); |
708 | if (!filterStartsWithBOL || !alternative->m_startsWithBOL) { |
709 | if (!newDisjunction) { |
710 | newDisjunction = std::make_unique<PatternDisjunction>(); |
711 | newDisjunction->m_parent = disjunction->m_parent; |
712 | } |
713 | PatternAlternative* newAlternative = newDisjunction->addNewAlternative(); |
714 | newAlternative->m_terms.reserveInitialCapacity(alternative->m_terms.size()); |
715 | for (unsigned i = 0; i < alternative->m_terms.size(); ++i) |
716 | newAlternative->m_terms.append(copyTerm(alternative->m_terms[i], filterStartsWithBOL)); |
717 | } |
718 | } |
719 | |
720 | if (!newDisjunction) |
721 | return 0; |
722 | |
723 | PatternDisjunction* copiedDisjunction = newDisjunction.get(); |
724 | m_pattern.m_disjunctions.append(WTFMove(newDisjunction)); |
725 | return copiedDisjunction; |
726 | } |
727 | |
728 | PatternTerm copyTerm(PatternTerm& term, bool filterStartsWithBOL = false) |
729 | { |
730 | if ((term.type != PatternTerm::TypeParenthesesSubpattern) && (term.type != PatternTerm::TypeParentheticalAssertion)) |
731 | return PatternTerm(term); |
732 | |
733 | PatternTerm termCopy = term; |
734 | termCopy.parentheses.disjunction = copyDisjunction(termCopy.parentheses.disjunction, filterStartsWithBOL); |
735 | m_pattern.m_hasCopiedParenSubexpressions = true; |
736 | return termCopy; |
737 | } |
738 | |
739 | void quantifyAtom(unsigned min, unsigned max, bool greedy) |
740 | { |
741 | ASSERT(min <= max); |
742 | ASSERT(m_alternative->m_terms.size()); |
743 | |
744 | if (!max) { |
745 | m_alternative->removeLastTerm(); |
746 | return; |
747 | } |
748 | |
749 | PatternTerm& term = m_alternative->lastTerm(); |
750 | ASSERT(term.type > PatternTerm::TypeAssertionWordBoundary); |
751 | ASSERT(term.quantityMinCount == 1 && term.quantityMaxCount == 1 && term.quantityType == QuantifierFixedCount); |
752 | |
753 | if (term.type == PatternTerm::TypeParentheticalAssertion) { |
754 | // If an assertion is quantified with a minimum count of zero, it can simply be removed. |
755 | // This arises from the RepeatMatcher behaviour in the spec. Matching an assertion never |
756 | // results in any input being consumed, however the continuation passed to the assertion |
757 | // (called in steps, 8c and 9 of the RepeatMatcher definition, ES5.1 15.10.2.5) will |
758 | // reject all zero length matches (see step 2.1). A match from the continuation of the |
759 | // expression will still be accepted regardless (via steps 8a and 11) - the upshot of all |
760 | // this is that matches from the assertion are not required, and won't be accepted anyway, |
761 | // so no need to ever run it. |
762 | if (!min) |
763 | m_alternative->removeLastTerm(); |
764 | // We never need to run an assertion more than once. Subsequent interations will be run |
765 | // with the same start index (since assertions are non-capturing) and the same captures |
766 | // (per step 4 of RepeatMatcher in ES5.1 15.10.2.5), and as such will always produce the |
767 | // same result and captures. If the first match succeeds then the subsequent (min - 1) |
768 | // matches will too. Any additional optional matches will fail (on the same basis as the |
769 | // minimum zero quantified assertions, above), but this will still result in a match. |
770 | return; |
771 | } |
772 | |
773 | if (min == max) |
774 | term.quantify(min, max, QuantifierFixedCount); |
775 | else if (!min || (term.type == PatternTerm::TypeParenthesesSubpattern && m_pattern.m_hasCopiedParenSubexpressions)) |
776 | term.quantify(min, max, greedy ? QuantifierGreedy : QuantifierNonGreedy); |
777 | else { |
778 | term.quantify(min, min, QuantifierFixedCount); |
779 | m_alternative->m_terms.append(copyTerm(term)); |
780 | // NOTE: this term is interesting from an analysis perspective, in that it can be ignored..... |
781 | m_alternative->lastTerm().quantify((max == quantifyInfinite) ? max : max - min, greedy ? QuantifierGreedy : QuantifierNonGreedy); |
782 | if (m_alternative->lastTerm().type == PatternTerm::TypeParenthesesSubpattern) |
783 | m_alternative->lastTerm().parentheses.isCopy = true; |
784 | } |
785 | } |
786 | |
787 | void disjunction() |
788 | { |
789 | m_alternative = m_alternative->m_parent->addNewAlternative(); |
790 | } |
791 | |
792 | ErrorCode setupAlternativeOffsets(PatternAlternative* alternative, unsigned currentCallFrameSize, unsigned initialInputPosition, unsigned& newCallFrameSize) WARN_UNUSED_RETURN |
793 | { |
794 | if (UNLIKELY(!isSafeToRecurse())) |
795 | return ErrorCode::TooManyDisjunctions; |
796 | |
797 | ErrorCode error = ErrorCode::NoError; |
798 | alternative->m_hasFixedSize = true; |
799 | Checked<unsigned, RecordOverflow> currentInputPosition = initialInputPosition; |
800 | |
801 | for (unsigned i = 0; i < alternative->m_terms.size(); ++i) { |
802 | PatternTerm& term = alternative->m_terms[i]; |
803 | |
804 | switch (term.type) { |
805 | case PatternTerm::TypeAssertionBOL: |
806 | case PatternTerm::TypeAssertionEOL: |
807 | case PatternTerm::TypeAssertionWordBoundary: |
808 | term.inputPosition = currentInputPosition.unsafeGet(); |
809 | break; |
810 | |
811 | case PatternTerm::TypeBackReference: |
812 | term.inputPosition = currentInputPosition.unsafeGet(); |
813 | term.frameLocation = currentCallFrameSize; |
814 | currentCallFrameSize += YarrStackSpaceForBackTrackInfoBackReference; |
815 | alternative->m_hasFixedSize = false; |
816 | break; |
817 | |
818 | case PatternTerm::TypeForwardReference: |
819 | break; |
820 | |
821 | case PatternTerm::TypePatternCharacter: |
822 | term.inputPosition = currentInputPosition.unsafeGet(); |
823 | if (term.quantityType != QuantifierFixedCount) { |
824 | term.frameLocation = currentCallFrameSize; |
825 | currentCallFrameSize += YarrStackSpaceForBackTrackInfoPatternCharacter; |
826 | alternative->m_hasFixedSize = false; |
827 | } else if (m_pattern.unicode()) { |
828 | Checked<unsigned, RecordOverflow> tempCount = term.quantityMaxCount; |
829 | tempCount *= U16_LENGTH(term.patternCharacter); |
830 | if (tempCount.hasOverflowed()) |
831 | return ErrorCode::OffsetTooLarge; |
832 | currentInputPosition += tempCount; |
833 | } else |
834 | currentInputPosition += term.quantityMaxCount; |
835 | break; |
836 | |
837 | case PatternTerm::TypeCharacterClass: |
838 | term.inputPosition = currentInputPosition.unsafeGet(); |
839 | if (term.quantityType != QuantifierFixedCount) { |
840 | term.frameLocation = currentCallFrameSize; |
841 | currentCallFrameSize += YarrStackSpaceForBackTrackInfoCharacterClass; |
842 | alternative->m_hasFixedSize = false; |
843 | } else if (m_pattern.unicode()) { |
844 | term.frameLocation = currentCallFrameSize; |
845 | currentCallFrameSize += YarrStackSpaceForBackTrackInfoCharacterClass; |
846 | if (term.characterClass->hasOneCharacterSize() && !term.invert()) { |
847 | Checked<unsigned, RecordOverflow> tempCount = term.quantityMaxCount; |
848 | tempCount *= term.characterClass->hasNonBMPCharacters() ? 2 : 1; |
849 | if (tempCount.hasOverflowed()) |
850 | return ErrorCode::OffsetTooLarge; |
851 | currentInputPosition += tempCount; |
852 | } else { |
853 | currentInputPosition += term.quantityMaxCount; |
854 | alternative->m_hasFixedSize = false; |
855 | } |
856 | } else |
857 | currentInputPosition += term.quantityMaxCount; |
858 | break; |
859 | |
860 | case PatternTerm::TypeParenthesesSubpattern: |
861 | // Note: for fixed once parentheses we will ensure at least the minimum is available; others are on their own. |
862 | term.frameLocation = currentCallFrameSize; |
863 | if (term.quantityMaxCount == 1 && !term.parentheses.isCopy) { |
864 | currentCallFrameSize += YarrStackSpaceForBackTrackInfoParenthesesOnce; |
865 | error = setupDisjunctionOffsets(term.parentheses.disjunction, currentCallFrameSize, currentInputPosition.unsafeGet(), currentCallFrameSize); |
866 | if (hasError(error)) |
867 | return error; |
868 | // If quantity is fixed, then pre-check its minimum size. |
869 | if (term.quantityType == QuantifierFixedCount) |
870 | currentInputPosition += term.parentheses.disjunction->m_minimumSize; |
871 | term.inputPosition = currentInputPosition.unsafeGet(); |
872 | } else if (term.parentheses.isTerminal) { |
873 | currentCallFrameSize += YarrStackSpaceForBackTrackInfoParenthesesTerminal; |
874 | error = setupDisjunctionOffsets(term.parentheses.disjunction, currentCallFrameSize, currentInputPosition.unsafeGet(), currentCallFrameSize); |
875 | if (hasError(error)) |
876 | return error; |
877 | term.inputPosition = currentInputPosition.unsafeGet(); |
878 | } else { |
879 | term.inputPosition = currentInputPosition.unsafeGet(); |
880 | currentCallFrameSize += YarrStackSpaceForBackTrackInfoParentheses; |
881 | error = setupDisjunctionOffsets(term.parentheses.disjunction, currentCallFrameSize, currentInputPosition.unsafeGet(), currentCallFrameSize); |
882 | if (hasError(error)) |
883 | return error; |
884 | } |
885 | // Fixed count of 1 could be accepted, if they have a fixed size *AND* if all alternatives are of the same length. |
886 | alternative->m_hasFixedSize = false; |
887 | break; |
888 | |
889 | case PatternTerm::TypeParentheticalAssertion: |
890 | term.inputPosition = currentInputPosition.unsafeGet(); |
891 | term.frameLocation = currentCallFrameSize; |
892 | error = setupDisjunctionOffsets(term.parentheses.disjunction, currentCallFrameSize + YarrStackSpaceForBackTrackInfoParentheticalAssertion, currentInputPosition.unsafeGet(), currentCallFrameSize); |
893 | if (hasError(error)) |
894 | return error; |
895 | break; |
896 | |
897 | case PatternTerm::TypeDotStarEnclosure: |
898 | ASSERT(!m_pattern.m_saveInitialStartValue); |
899 | alternative->m_hasFixedSize = false; |
900 | term.inputPosition = initialInputPosition; |
901 | m_pattern.m_initialStartValueFrameLocation = currentCallFrameSize; |
902 | currentCallFrameSize += YarrStackSpaceForDotStarEnclosure; |
903 | m_pattern.m_saveInitialStartValue = true; |
904 | break; |
905 | } |
906 | if (currentInputPosition.hasOverflowed()) |
907 | return ErrorCode::OffsetTooLarge; |
908 | } |
909 | |
910 | alternative->m_minimumSize = (currentInputPosition - initialInputPosition).unsafeGet(); |
911 | newCallFrameSize = currentCallFrameSize; |
912 | return error; |
913 | } |
914 | |
915 | ErrorCode setupDisjunctionOffsets(PatternDisjunction* disjunction, unsigned initialCallFrameSize, unsigned initialInputPosition, unsigned& callFrameSize) |
916 | { |
917 | if (UNLIKELY(!isSafeToRecurse())) |
918 | return ErrorCode::TooManyDisjunctions; |
919 | |
920 | if ((disjunction != m_pattern.m_body) && (disjunction->m_alternatives.size() > 1)) |
921 | initialCallFrameSize += YarrStackSpaceForBackTrackInfoAlternative; |
922 | |
923 | unsigned minimumInputSize = UINT_MAX; |
924 | unsigned maximumCallFrameSize = 0; |
925 | bool hasFixedSize = true; |
926 | ErrorCode error = ErrorCode::NoError; |
927 | |
928 | for (unsigned alt = 0; alt < disjunction->m_alternatives.size(); ++alt) { |
929 | PatternAlternative* alternative = disjunction->m_alternatives[alt].get(); |
930 | unsigned currentAlternativeCallFrameSize; |
931 | error = setupAlternativeOffsets(alternative, initialCallFrameSize, initialInputPosition, currentAlternativeCallFrameSize); |
932 | if (hasError(error)) |
933 | return error; |
934 | minimumInputSize = std::min(minimumInputSize, alternative->m_minimumSize); |
935 | maximumCallFrameSize = std::max(maximumCallFrameSize, currentAlternativeCallFrameSize); |
936 | hasFixedSize &= alternative->m_hasFixedSize; |
937 | if (alternative->m_minimumSize > INT_MAX) |
938 | m_pattern.m_containsUnsignedLengthPattern = true; |
939 | } |
940 | |
941 | ASSERT(minimumInputSize != UINT_MAX); |
942 | ASSERT(maximumCallFrameSize >= initialCallFrameSize); |
943 | |
944 | disjunction->m_hasFixedSize = hasFixedSize; |
945 | disjunction->m_minimumSize = minimumInputSize; |
946 | disjunction->m_callFrameSize = maximumCallFrameSize; |
947 | callFrameSize = maximumCallFrameSize; |
948 | return error; |
949 | } |
950 | |
951 | ErrorCode setupOffsets() |
952 | { |
953 | // FIXME: Yarr should not use the stack to handle subpatterns (rdar://problem/26436314). |
954 | unsigned ignoredCallFrameSize; |
955 | return setupDisjunctionOffsets(m_pattern.m_body, 0, 0, ignoredCallFrameSize); |
956 | } |
957 | |
958 | // This optimization identifies sets of parentheses that we will never need to backtrack. |
959 | // In these cases we do not need to store state from prior iterations. |
960 | // We can presently avoid backtracking for: |
961 | // * where the parens are at the end of the regular expression (last term in any of the |
962 | // alternatives of the main body disjunction). |
963 | // * where the parens are non-capturing, and quantified unbounded greedy (*). |
964 | // * where the parens do not contain any capturing subpatterns. |
965 | void checkForTerminalParentheses() |
966 | { |
967 | // This check is much too crude; should be just checking whether the candidate |
968 | // node contains nested capturing subpatterns, not the whole expression! |
969 | if (m_pattern.m_numSubpatterns) |
970 | return; |
971 | |
972 | Vector<std::unique_ptr<PatternAlternative>>& alternatives = m_pattern.m_body->m_alternatives; |
973 | for (size_t i = 0; i < alternatives.size(); ++i) { |
974 | Vector<PatternTerm>& terms = alternatives[i]->m_terms; |
975 | if (terms.size()) { |
976 | PatternTerm& term = terms.last(); |
977 | if (term.type == PatternTerm::TypeParenthesesSubpattern |
978 | && term.quantityType == QuantifierGreedy |
979 | && term.quantityMinCount == 0 |
980 | && term.quantityMaxCount == quantifyInfinite |
981 | && !term.capture()) |
982 | term.parentheses.isTerminal = true; |
983 | } |
984 | } |
985 | } |
986 | |
987 | void optimizeBOL() |
988 | { |
989 | // Look for expressions containing beginning of line (^) anchoring and unroll them. |
990 | // e.g. /^a|^b|c/ becomes /^a|^b|c/ which is executed once followed by /c/ which loops |
991 | // This code relies on the parsing code tagging alternatives with m_containsBOL and |
992 | // m_startsWithBOL and rolling those up to containing alternatives. |
993 | // At this point, this is only valid for non-multiline expressions. |
994 | PatternDisjunction* disjunction = m_pattern.m_body; |
995 | |
996 | if (!m_pattern.m_containsBOL || m_pattern.multiline()) |
997 | return; |
998 | |
999 | PatternDisjunction* loopDisjunction = copyDisjunction(disjunction, true); |
1000 | |
1001 | // Set alternatives in disjunction to "onceThrough" |
1002 | for (unsigned alt = 0; alt < disjunction->m_alternatives.size(); ++alt) |
1003 | disjunction->m_alternatives[alt]->setOnceThrough(); |
1004 | |
1005 | if (loopDisjunction) { |
1006 | // Move alternatives from loopDisjunction to disjunction |
1007 | for (unsigned alt = 0; alt < loopDisjunction->m_alternatives.size(); ++alt) |
1008 | disjunction->m_alternatives.append(loopDisjunction->m_alternatives[alt].release()); |
1009 | |
1010 | loopDisjunction->m_alternatives.clear(); |
1011 | } |
1012 | } |
1013 | |
1014 | bool containsCapturingTerms(PatternAlternative* alternative, size_t firstTermIndex, size_t endIndex) |
1015 | { |
1016 | Vector<PatternTerm>& terms = alternative->m_terms; |
1017 | |
1018 | ASSERT(endIndex <= terms.size()); |
1019 | for (size_t termIndex = firstTermIndex; termIndex < endIndex; ++termIndex) { |
1020 | PatternTerm& term = terms[termIndex]; |
1021 | |
1022 | if (term.m_capture) |
1023 | return true; |
1024 | |
1025 | if (term.type == PatternTerm::TypeParenthesesSubpattern) { |
1026 | PatternDisjunction* nestedDisjunction = term.parentheses.disjunction; |
1027 | for (unsigned alt = 0; alt < nestedDisjunction->m_alternatives.size(); ++alt) { |
1028 | if (containsCapturingTerms(nestedDisjunction->m_alternatives[alt].get(), 0, nestedDisjunction->m_alternatives[alt]->m_terms.size())) |
1029 | return true; |
1030 | } |
1031 | } |
1032 | } |
1033 | |
1034 | return false; |
1035 | } |
1036 | |
1037 | // This optimization identifies alternatives in the form of |
1038 | // [^].*[?]<expression>.*[$] for expressions that don't have any |
1039 | // capturing terms. The alternative is changed to <expression> |
1040 | // followed by processing of the dot stars to find and adjust the |
1041 | // beginning and the end of the match. |
1042 | void optimizeDotStarWrappedExpressions() |
1043 | { |
1044 | Vector<std::unique_ptr<PatternAlternative>>& alternatives = m_pattern.m_body->m_alternatives; |
1045 | if (alternatives.size() != 1) |
1046 | return; |
1047 | |
1048 | CharacterClass* dotCharacterClass = m_pattern.dotAll() ? m_pattern.anyCharacterClass() : m_pattern.newlineCharacterClass(); |
1049 | PatternAlternative* alternative = alternatives[0].get(); |
1050 | Vector<PatternTerm>& terms = alternative->m_terms; |
1051 | if (terms.size() >= 3) { |
1052 | bool startsWithBOL = false; |
1053 | bool endsWithEOL = false; |
1054 | size_t termIndex, firstExpressionTerm; |
1055 | |
1056 | termIndex = 0; |
1057 | if (terms[termIndex].type == PatternTerm::TypeAssertionBOL) { |
1058 | startsWithBOL = true; |
1059 | ++termIndex; |
1060 | } |
1061 | |
1062 | PatternTerm& firstNonAnchorTerm = terms[termIndex]; |
1063 | if (firstNonAnchorTerm.type != PatternTerm::TypeCharacterClass |
1064 | || firstNonAnchorTerm.characterClass != dotCharacterClass |
1065 | || firstNonAnchorTerm.quantityMinCount |
1066 | || firstNonAnchorTerm.quantityMaxCount != quantifyInfinite) |
1067 | return; |
1068 | |
1069 | firstExpressionTerm = termIndex + 1; |
1070 | |
1071 | termIndex = terms.size() - 1; |
1072 | if (terms[termIndex].type == PatternTerm::TypeAssertionEOL) { |
1073 | endsWithEOL = true; |
1074 | --termIndex; |
1075 | } |
1076 | |
1077 | PatternTerm& lastNonAnchorTerm = terms[termIndex]; |
1078 | if (lastNonAnchorTerm.type != PatternTerm::TypeCharacterClass |
1079 | || lastNonAnchorTerm.characterClass != dotCharacterClass |
1080 | || lastNonAnchorTerm.quantityType != QuantifierGreedy |
1081 | || lastNonAnchorTerm.quantityMinCount |
1082 | || lastNonAnchorTerm.quantityMaxCount != quantifyInfinite) |
1083 | return; |
1084 | |
1085 | size_t endIndex = termIndex; |
1086 | if (firstExpressionTerm >= endIndex) |
1087 | return; |
1088 | |
1089 | if (!containsCapturingTerms(alternative, firstExpressionTerm, endIndex)) { |
1090 | for (termIndex = terms.size() - 1; termIndex >= endIndex; --termIndex) |
1091 | terms.remove(termIndex); |
1092 | |
1093 | for (termIndex = firstExpressionTerm; termIndex > 0; --termIndex) |
1094 | terms.remove(termIndex - 1); |
1095 | |
1096 | terms.append(PatternTerm(startsWithBOL, endsWithEOL)); |
1097 | |
1098 | m_pattern.m_containsBOL = false; |
1099 | } |
1100 | } |
1101 | } |
1102 | |
1103 | private: |
1104 | bool isSafeToRecurse() const |
1105 | { |
1106 | if (!m_stackLimit) |
1107 | return true; |
1108 | ASSERT(Thread::current().stack().isGrowingDownward()); |
1109 | int8_t* curr = reinterpret_cast<int8_t*>(currentStackPointer()); |
1110 | int8_t* limit = reinterpret_cast<int8_t*>(m_stackLimit); |
1111 | return curr >= limit; |
1112 | } |
1113 | |
1114 | YarrPattern& m_pattern; |
1115 | PatternAlternative* m_alternative; |
1116 | CharacterClassConstructor m_characterClassConstructor; |
1117 | Vector<String> m_unmatchedNamedForwardReferences; |
1118 | void* m_stackLimit; |
1119 | bool m_invertCharacterClass; |
1120 | bool m_invertParentheticalAssertion { false }; |
1121 | }; |
1122 | |
1123 | ErrorCode YarrPattern::compile(const String& patternString, void* stackLimit) |
1124 | { |
1125 | YarrPatternConstructor constructor(*this, stackLimit); |
1126 | |
1127 | { |
1128 | ErrorCode error = parse(constructor, patternString, unicode()); |
1129 | if (hasError(error)) |
1130 | return error; |
1131 | } |
1132 | |
1133 | // If the pattern contains illegal backreferences reset & reparse. |
1134 | // Quoting Netscape's "What's new in JavaScript 1.2", |
1135 | // "Note: if the number of left parentheses is less than the number specified |
1136 | // in \#, the \# is taken as an octal escape as described in the next row." |
1137 | if (containsIllegalBackReference() || containsIllegalNamedForwardReferences()) { |
1138 | if (unicode()) |
1139 | return ErrorCode::InvalidBackreference; |
1140 | |
1141 | unsigned numSubpatterns = m_numSubpatterns; |
1142 | |
1143 | constructor.saveUnmatchedNamedForwardReferences(); |
1144 | constructor.resetForReparsing(); |
1145 | ErrorCode error = parse(constructor, patternString, unicode(), numSubpatterns); |
1146 | ASSERT_UNUSED(error, !hasError(error)); |
1147 | ASSERT(numSubpatterns == m_numSubpatterns); |
1148 | } |
1149 | |
1150 | constructor.checkForTerminalParentheses(); |
1151 | constructor.optimizeDotStarWrappedExpressions(); |
1152 | constructor.optimizeBOL(); |
1153 | |
1154 | { |
1155 | ErrorCode error = constructor.setupOffsets(); |
1156 | if (hasError(error)) |
1157 | return error; |
1158 | } |
1159 | |
1160 | if (Options::dumpCompiledRegExpPatterns()) |
1161 | dumpPattern(patternString); |
1162 | |
1163 | return ErrorCode::NoError; |
1164 | } |
1165 | |
1166 | YarrPattern::YarrPattern(const String& pattern, OptionSet<Flags> flags, ErrorCode& error, void* stackLimit) |
1167 | : m_containsBackreferences(false) |
1168 | , m_containsBOL(false) |
1169 | , m_containsUnsignedLengthPattern(false) |
1170 | , m_hasCopiedParenSubexpressions(false) |
1171 | , m_saveInitialStartValue(false) |
1172 | , m_flags(flags) |
1173 | { |
1174 | ASSERT(m_flags != Flags::DeletedValue); |
1175 | error = compile(pattern, stackLimit); |
1176 | } |
1177 | |
1178 | void indentForNestingLevel(PrintStream& out, unsigned nestingDepth) |
1179 | { |
1180 | out.print(" " ); |
1181 | for (; nestingDepth; --nestingDepth) |
1182 | out.print(" " ); |
1183 | } |
1184 | |
1185 | void dumpUChar32(PrintStream& out, UChar32 c) |
1186 | { |
1187 | if (c >= ' '&& c <= 0xff) |
1188 | out.printf("'%c'" , static_cast<char>(c)); |
1189 | else |
1190 | out.printf("0x%04x" , c); |
1191 | } |
1192 | |
1193 | void dumpCharacterClass(PrintStream& out, YarrPattern* pattern, CharacterClass* characterClass) |
1194 | { |
1195 | if (characterClass == pattern->anyCharacterClass()) |
1196 | out.print("<any character>" ); |
1197 | else if (characterClass == pattern->newlineCharacterClass()) |
1198 | out.print("<newline>" ); |
1199 | else if (characterClass == pattern->digitsCharacterClass()) |
1200 | out.print("<digits>" ); |
1201 | else if (characterClass == pattern->spacesCharacterClass()) |
1202 | out.print("<whitespace>" ); |
1203 | else if (characterClass == pattern->wordcharCharacterClass()) |
1204 | out.print("<word>" ); |
1205 | else if (characterClass == pattern->wordUnicodeIgnoreCaseCharCharacterClass()) |
1206 | out.print("<unicode word ignore case>" ); |
1207 | else if (characterClass == pattern->nondigitsCharacterClass()) |
1208 | out.print("<non-digits>" ); |
1209 | else if (characterClass == pattern->nonspacesCharacterClass()) |
1210 | out.print("<non-whitespace>" ); |
1211 | else if (characterClass == pattern->nonwordcharCharacterClass()) |
1212 | out.print("<non-word>" ); |
1213 | else if (characterClass == pattern->nonwordUnicodeIgnoreCaseCharCharacterClass()) |
1214 | out.print("<unicode non-word ignore case>" ); |
1215 | else { |
1216 | bool needMatchesRangesSeperator = false; |
1217 | |
1218 | auto dumpMatches = [&] (const char* prefix, Vector<UChar32> matches) { |
1219 | size_t matchesSize = matches.size(); |
1220 | if (matchesSize) { |
1221 | if (needMatchesRangesSeperator) |
1222 | out.print("," ); |
1223 | needMatchesRangesSeperator = true; |
1224 | |
1225 | out.print(prefix, ":(" ); |
1226 | for (size_t i = 0; i < matchesSize; ++i) { |
1227 | if (i) |
1228 | out.print("," ); |
1229 | dumpUChar32(out, matches[i]); |
1230 | } |
1231 | out.print(")" ); |
1232 | } |
1233 | }; |
1234 | |
1235 | auto dumpRanges = [&] (const char* prefix, Vector<CharacterRange> ranges) { |
1236 | size_t rangeSize = ranges.size(); |
1237 | if (rangeSize) { |
1238 | if (needMatchesRangesSeperator) |
1239 | out.print("," ); |
1240 | needMatchesRangesSeperator = true; |
1241 | |
1242 | out.print(prefix, " ranges:(" ); |
1243 | for (size_t i = 0; i < rangeSize; ++i) { |
1244 | if (i) |
1245 | out.print("," ); |
1246 | CharacterRange range = ranges[i]; |
1247 | out.print("(" ); |
1248 | dumpUChar32(out, range.begin); |
1249 | out.print(".." ); |
1250 | dumpUChar32(out, range.end); |
1251 | out.print(")" ); |
1252 | } |
1253 | out.print(")" ); |
1254 | } |
1255 | }; |
1256 | |
1257 | out.print("[" ); |
1258 | dumpMatches("ASCII" , characterClass->m_matches); |
1259 | dumpRanges("ASCII" , characterClass->m_ranges); |
1260 | dumpMatches("Unicode" , characterClass->m_matchesUnicode); |
1261 | dumpRanges("Unicode" , characterClass->m_rangesUnicode); |
1262 | out.print("]" ); |
1263 | } |
1264 | } |
1265 | |
1266 | void PatternAlternative::dump(PrintStream& out, YarrPattern* thisPattern, unsigned nestingDepth) |
1267 | { |
1268 | out.print("minimum size: " , m_minimumSize); |
1269 | if (m_hasFixedSize) |
1270 | out.print(",fixed size" ); |
1271 | if (m_onceThrough) |
1272 | out.print(",once through" ); |
1273 | if (m_startsWithBOL) |
1274 | out.print(",starts with ^" ); |
1275 | if (m_containsBOL) |
1276 | out.print(",contains ^" ); |
1277 | out.print("\n" ); |
1278 | |
1279 | for (size_t i = 0; i < m_terms.size(); ++i) |
1280 | m_terms[i].dump(out, thisPattern, nestingDepth); |
1281 | } |
1282 | |
1283 | void PatternTerm::dumpQuantifier(PrintStream& out) |
1284 | { |
1285 | if (quantityType == QuantifierFixedCount && quantityMinCount == 1 && quantityMaxCount == 1) |
1286 | return; |
1287 | out.print(" {" , quantityMinCount.unsafeGet()); |
1288 | if (quantityMinCount != quantityMaxCount) { |
1289 | if (quantityMaxCount == UINT_MAX) |
1290 | out.print(",..." ); |
1291 | else |
1292 | out.print("," , quantityMaxCount.unsafeGet()); |
1293 | } |
1294 | out.print("}" ); |
1295 | if (quantityType == QuantifierGreedy) |
1296 | out.print(" greedy" ); |
1297 | else if (quantityType == QuantifierNonGreedy) |
1298 | out.print(" non-greedy" ); |
1299 | } |
1300 | |
1301 | void PatternTerm::dump(PrintStream& out, YarrPattern* thisPattern, unsigned nestingDepth) |
1302 | { |
1303 | indentForNestingLevel(out, nestingDepth); |
1304 | |
1305 | if (type != TypeParenthesesSubpattern && type != TypeParentheticalAssertion) { |
1306 | if (invert()) |
1307 | out.print("not " ); |
1308 | } |
1309 | |
1310 | switch (type) { |
1311 | case TypeAssertionBOL: |
1312 | out.println("BOL" ); |
1313 | break; |
1314 | case TypeAssertionEOL: |
1315 | out.println("EOL" ); |
1316 | break; |
1317 | case TypeAssertionWordBoundary: |
1318 | out.println("word boundary" ); |
1319 | break; |
1320 | case TypePatternCharacter: |
1321 | out.printf("character " ); |
1322 | out.printf("inputPosition %u " , inputPosition); |
1323 | if (thisPattern->ignoreCase() && isASCIIAlpha(patternCharacter)) { |
1324 | dumpUChar32(out, toASCIIUpper(patternCharacter)); |
1325 | out.print("/" ); |
1326 | dumpUChar32(out, toASCIILower(patternCharacter)); |
1327 | } else |
1328 | dumpUChar32(out, patternCharacter); |
1329 | dumpQuantifier(out); |
1330 | if (quantityType != QuantifierFixedCount) |
1331 | out.print(",frame location " , frameLocation); |
1332 | out.println(); |
1333 | break; |
1334 | case TypeCharacterClass: |
1335 | out.print("character class " ); |
1336 | out.printf("inputPosition %u " , inputPosition); |
1337 | dumpCharacterClass(out, thisPattern, characterClass); |
1338 | dumpQuantifier(out); |
1339 | if (quantityType != QuantifierFixedCount || thisPattern->unicode()) |
1340 | out.print(",frame location " , frameLocation); |
1341 | out.println(); |
1342 | break; |
1343 | case TypeBackReference: |
1344 | out.print("back reference to subpattern #" , backReferenceSubpatternId); |
1345 | out.println(",frame location " , frameLocation); |
1346 | break; |
1347 | case TypeForwardReference: |
1348 | out.println("forward reference" ); |
1349 | break; |
1350 | case TypeParenthesesSubpattern: |
1351 | if (m_capture) |
1352 | out.print("captured " ); |
1353 | else |
1354 | out.print("non-captured " ); |
1355 | |
1356 | FALLTHROUGH; |
1357 | case TypeParentheticalAssertion: |
1358 | if (m_invert) |
1359 | out.print("inverted " ); |
1360 | |
1361 | if (type == TypeParenthesesSubpattern) |
1362 | out.print("subpattern" ); |
1363 | else if (type == TypeParentheticalAssertion) |
1364 | out.print("assertion" ); |
1365 | |
1366 | if (m_capture) |
1367 | out.print(" #" , parentheses.subpatternId); |
1368 | |
1369 | dumpQuantifier(out); |
1370 | |
1371 | if (parentheses.isCopy) |
1372 | out.print(",copy" ); |
1373 | |
1374 | if (parentheses.isTerminal) |
1375 | out.print(",terminal" ); |
1376 | |
1377 | out.println(",frame location " , frameLocation); |
1378 | |
1379 | if (parentheses.disjunction->m_alternatives.size() > 1) { |
1380 | indentForNestingLevel(out, nestingDepth + 1); |
1381 | unsigned alternativeFrameLocation = frameLocation; |
1382 | if (quantityMaxCount == 1 && !parentheses.isCopy) |
1383 | alternativeFrameLocation += YarrStackSpaceForBackTrackInfoParenthesesOnce; |
1384 | else if (parentheses.isTerminal) |
1385 | alternativeFrameLocation += YarrStackSpaceForBackTrackInfoParenthesesTerminal; |
1386 | else |
1387 | alternativeFrameLocation += YarrStackSpaceForBackTrackInfoParentheses; |
1388 | out.println("alternative list,frame location " , alternativeFrameLocation); |
1389 | } |
1390 | |
1391 | parentheses.disjunction->dump(out, thisPattern, nestingDepth + 1); |
1392 | break; |
1393 | case TypeDotStarEnclosure: |
1394 | out.println(".* enclosure,frame location " , thisPattern->m_initialStartValueFrameLocation); |
1395 | break; |
1396 | } |
1397 | } |
1398 | |
1399 | void PatternDisjunction::dump(PrintStream& out, YarrPattern* thisPattern, unsigned nestingDepth = 0) |
1400 | { |
1401 | unsigned alternativeCount = m_alternatives.size(); |
1402 | for (unsigned i = 0; i < alternativeCount; ++i) { |
1403 | indentForNestingLevel(out, nestingDepth); |
1404 | if (alternativeCount > 1) |
1405 | out.print("alternative #" , i, ": " ); |
1406 | m_alternatives[i].get()->dump(out, thisPattern, nestingDepth + (alternativeCount > 1)); |
1407 | } |
1408 | } |
1409 | |
1410 | void YarrPattern::dumpPatternString(PrintStream& out, const String& patternString) |
1411 | { |
1412 | out.print("/" , patternString, "/" ); |
1413 | |
1414 | if (global()) |
1415 | out.print("g" ); |
1416 | if (ignoreCase()) |
1417 | out.print("i" ); |
1418 | if (multiline()) |
1419 | out.print("m" ); |
1420 | if (unicode()) |
1421 | out.print("u" ); |
1422 | if (sticky()) |
1423 | out.print("y" ); |
1424 | } |
1425 | |
1426 | void YarrPattern::dumpPattern(const String& patternString) |
1427 | { |
1428 | dumpPattern(WTF::dataFile(), patternString); |
1429 | } |
1430 | |
1431 | void YarrPattern::dumpPattern(PrintStream& out, const String& patternString) |
1432 | { |
1433 | out.print("RegExp pattern for " ); |
1434 | dumpPatternString(out, patternString); |
1435 | |
1436 | if (m_flags) { |
1437 | bool printSeperator = false; |
1438 | out.print(" (" ); |
1439 | if (global()) { |
1440 | out.print("global" ); |
1441 | printSeperator = true; |
1442 | } |
1443 | if (ignoreCase()) { |
1444 | if (printSeperator) |
1445 | out.print("|" ); |
1446 | out.print("ignore case" ); |
1447 | printSeperator = true; |
1448 | } |
1449 | if (multiline()) { |
1450 | if (printSeperator) |
1451 | out.print("|" ); |
1452 | out.print("multiline" ); |
1453 | printSeperator = true; |
1454 | } |
1455 | if (unicode()) { |
1456 | if (printSeperator) |
1457 | out.print("|" ); |
1458 | out.print("unicode" ); |
1459 | printSeperator = true; |
1460 | } |
1461 | if (sticky()) { |
1462 | if (printSeperator) |
1463 | out.print("|" ); |
1464 | out.print("sticky" ); |
1465 | printSeperator = true; |
1466 | } |
1467 | out.print(")" ); |
1468 | } |
1469 | out.print(":\n" ); |
1470 | if (m_body->m_callFrameSize) |
1471 | out.print(" callframe size: " , m_body->m_callFrameSize, "\n" ); |
1472 | m_body->dump(out, this); |
1473 | } |
1474 | |
1475 | std::unique_ptr<CharacterClass> anycharCreate() |
1476 | { |
1477 | auto characterClass = std::make_unique<CharacterClass>(); |
1478 | characterClass->m_ranges.append(CharacterRange(0x00, 0x7f)); |
1479 | characterClass->m_rangesUnicode.append(CharacterRange(0x0080, 0x10ffff)); |
1480 | characterClass->m_characterWidths = CharacterClassWidths::HasBothBMPAndNonBMP; |
1481 | characterClass->m_anyCharacter = true; |
1482 | return characterClass; |
1483 | } |
1484 | |
1485 | } } // namespace JSC::Yarr |
1486 | |