• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Copyright (C) 2009 Apple Inc. All rights reserved.
3  * Copyright (C) 2010 Peter Varga (pvarga@inf.u-szeged.hu), 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 "Yarr.h"
31 #include "YarrParser.h"
32 #include <wtf/Vector.h>
33 
34 using namespace WTF;
35 
36 namespace JSC { namespace Yarr {
37 
38 #include "RegExpJitTables.h"
39 
40 class CharacterClassConstructor {
41 public:
CharacterClassConstructor(bool isCaseInsensitive=false)42     CharacterClassConstructor(bool isCaseInsensitive = false)
43         : m_isCaseInsensitive(isCaseInsensitive)
44     {
45     }
46 
reset()47     void reset()
48     {
49         m_matches.clear();
50         m_ranges.clear();
51         m_matchesUnicode.clear();
52         m_rangesUnicode.clear();
53     }
54 
append(const CharacterClass * other)55     void append(const CharacterClass* other)
56     {
57         for (size_t i = 0; i < other->m_matches.size(); ++i)
58             addSorted(m_matches, other->m_matches[i]);
59         for (size_t i = 0; i < other->m_ranges.size(); ++i)
60             addSortedRange(m_ranges, other->m_ranges[i].begin, other->m_ranges[i].end);
61         for (size_t i = 0; i < other->m_matchesUnicode.size(); ++i)
62             addSorted(m_matchesUnicode, other->m_matchesUnicode[i]);
63         for (size_t i = 0; i < other->m_rangesUnicode.size(); ++i)
64             addSortedRange(m_rangesUnicode, other->m_rangesUnicode[i].begin, other->m_rangesUnicode[i].end);
65     }
66 
putChar(UChar ch)67     void putChar(UChar ch)
68     {
69         if (ch <= 0x7f) {
70             if (m_isCaseInsensitive && isASCIIAlpha(ch)) {
71                 addSorted(m_matches, toASCIIUpper(ch));
72                 addSorted(m_matches, toASCIILower(ch));
73             } else
74                 addSorted(m_matches, ch);
75         } else {
76             UChar upper, lower;
77             if (m_isCaseInsensitive && ((upper = Unicode::toUpper(ch)) != (lower = Unicode::toLower(ch)))) {
78                 addSorted(m_matchesUnicode, upper);
79                 addSorted(m_matchesUnicode, lower);
80             } else
81                 addSorted(m_matchesUnicode, ch);
82         }
83     }
84 
85     // returns true if this character has another case, and 'ch' is the upper case form.
isUnicodeUpper(UChar ch)86     static inline bool isUnicodeUpper(UChar ch)
87     {
88         return ch != Unicode::toLower(ch);
89     }
90 
91     // returns true if this character has another case, and 'ch' is the lower case form.
isUnicodeLower(UChar ch)92     static inline bool isUnicodeLower(UChar ch)
93     {
94         return ch != Unicode::toUpper(ch);
95     }
96 
putRange(UChar lo,UChar hi)97     void putRange(UChar lo, UChar hi)
98     {
99         if (lo <= 0x7f) {
100             char asciiLo = lo;
101             char asciiHi = std::min(hi, (UChar)0x7f);
102             addSortedRange(m_ranges, lo, asciiHi);
103 
104             if (m_isCaseInsensitive) {
105                 if ((asciiLo <= 'Z') && (asciiHi >= 'A'))
106                     addSortedRange(m_ranges, std::max(asciiLo, 'A')+('a'-'A'), std::min(asciiHi, 'Z')+('a'-'A'));
107                 if ((asciiLo <= 'z') && (asciiHi >= 'a'))
108                     addSortedRange(m_ranges, std::max(asciiLo, 'a')+('A'-'a'), std::min(asciiHi, 'z')+('A'-'a'));
109             }
110         }
111         if (hi >= 0x80) {
112             uint32_t unicodeCurr = std::max(lo, (UChar)0x80);
113             addSortedRange(m_rangesUnicode, unicodeCurr, hi);
114 
115             if (m_isCaseInsensitive) {
116                 while (unicodeCurr <= hi) {
117                     // If the upper bound of the range (hi) is 0xffff, the increments to
118                     // unicodeCurr in this loop may take it to 0x10000.  This is fine
119                     // (if so we won't re-enter the loop, since the loop condition above
120                     // will definitely fail) - but this does mean we cannot use a UChar
121                     // to represent unicodeCurr, we must use a 32-bit value instead.
122                     ASSERT(unicodeCurr <= 0xffff);
123 
124                     if (isUnicodeUpper(unicodeCurr)) {
125                         UChar lowerCaseRangeBegin = Unicode::toLower(unicodeCurr);
126                         UChar lowerCaseRangeEnd = lowerCaseRangeBegin;
127                         while ((++unicodeCurr <= hi) && isUnicodeUpper(unicodeCurr) && (Unicode::toLower(unicodeCurr) == (lowerCaseRangeEnd + 1)))
128                             lowerCaseRangeEnd++;
129                         addSortedRange(m_rangesUnicode, lowerCaseRangeBegin, lowerCaseRangeEnd);
130                     } else if (isUnicodeLower(unicodeCurr)) {
131                         UChar upperCaseRangeBegin = Unicode::toUpper(unicodeCurr);
132                         UChar upperCaseRangeEnd = upperCaseRangeBegin;
133                         while ((++unicodeCurr <= hi) && isUnicodeLower(unicodeCurr) && (Unicode::toUpper(unicodeCurr) == (upperCaseRangeEnd + 1)))
134                             upperCaseRangeEnd++;
135                         addSortedRange(m_rangesUnicode, upperCaseRangeBegin, upperCaseRangeEnd);
136                     } else
137                         ++unicodeCurr;
138                 }
139             }
140         }
141     }
142 
charClass()143     CharacterClass* charClass()
144     {
145         CharacterClass* characterClass = new CharacterClass(0);
146 
147         characterClass->m_matches.append(m_matches);
148         characterClass->m_ranges.append(m_ranges);
149         characterClass->m_matchesUnicode.append(m_matchesUnicode);
150         characterClass->m_rangesUnicode.append(m_rangesUnicode);
151 
152         reset();
153 
154         return characterClass;
155     }
156 
157 private:
addSorted(Vector<UChar> & matches,UChar ch)158     void addSorted(Vector<UChar>& matches, UChar ch)
159     {
160         unsigned pos = 0;
161         unsigned range = matches.size();
162 
163         // binary chop, find position to insert char.
164         while (range) {
165             unsigned index = range >> 1;
166 
167             int val = matches[pos+index] - ch;
168             if (!val)
169                 return;
170             else if (val > 0)
171                 range = index;
172             else {
173                 pos += (index+1);
174                 range -= (index+1);
175             }
176         }
177 
178         if (pos == matches.size())
179             matches.append(ch);
180         else
181             matches.insert(pos, ch);
182     }
183 
addSortedRange(Vector<CharacterRange> & ranges,UChar lo,UChar hi)184     void addSortedRange(Vector<CharacterRange>& ranges, UChar lo, UChar hi)
185     {
186         unsigned end = ranges.size();
187 
188         // Simple linear scan - I doubt there are that many ranges anyway...
189         // feel free to fix this with something faster (eg binary chop).
190         for (unsigned i = 0; i < end; ++i) {
191             // does the new range fall before the current position in the array
192             if (hi < ranges[i].begin) {
193                 // optional optimization: concatenate appending ranges? - may not be worthwhile.
194                 if (hi == (ranges[i].begin - 1)) {
195                     ranges[i].begin = lo;
196                     return;
197                 }
198                 ranges.insert(i, CharacterRange(lo, hi));
199                 return;
200             }
201             // Okay, since we didn't hit the last case, the end of the new range is definitely at or after the begining
202             // If the new range start at or before the end of the last range, then the overlap (if it starts one after the
203             // end of the last range they concatenate, which is just as good.
204             if (lo <= (ranges[i].end + 1)) {
205                 // found an intersect! we'll replace this entry in the array.
206                 ranges[i].begin = std::min(ranges[i].begin, lo);
207                 ranges[i].end = std::max(ranges[i].end, hi);
208 
209                 // now check if the new range can subsume any subsequent ranges.
210                 unsigned next = i+1;
211                 // each iteration of the loop we will either remove something from the list, or break the loop.
212                 while (next < ranges.size()) {
213                     if (ranges[next].begin <= (ranges[i].end + 1)) {
214                         // the next entry now overlaps / concatenates this one.
215                         ranges[i].end = std::max(ranges[i].end, ranges[next].end);
216                         ranges.remove(next);
217                     } else
218                         break;
219                 }
220 
221                 return;
222             }
223         }
224 
225         // CharacterRange comes after all existing ranges.
226         ranges.append(CharacterRange(lo, hi));
227     }
228 
229     bool m_isCaseInsensitive;
230 
231     Vector<UChar> m_matches;
232     Vector<CharacterRange> m_ranges;
233     Vector<UChar> m_matchesUnicode;
234     Vector<CharacterRange> m_rangesUnicode;
235 };
236 
237 struct BeginCharHelper {
BeginCharHelperJSC::Yarr::BeginCharHelper238     BeginCharHelper(Vector<BeginChar>* beginChars, bool isCaseInsensitive = false)
239         : m_beginChars(beginChars)
240         , m_isCaseInsensitive(isCaseInsensitive)
241     {}
242 
addBeginCharJSC::Yarr::BeginCharHelper243     void addBeginChar(BeginChar beginChar, Vector<TermChain>* hotTerms, QuantifierType quantityType, unsigned quantityCount)
244     {
245         if (quantityType == QuantifierFixedCount && quantityCount > 1) {
246             // We duplicate the first found character if the quantity of the term is more than one. eg.: /a{3}/
247             beginChar.value |= beginChar.value << 16;
248             beginChar.mask |= beginChar.mask << 16;
249             addCharacter(beginChar);
250         } else if (quantityType == QuantifierFixedCount && quantityCount == 1 && hotTerms->size())
251             // In case of characters with fixed quantifier we should check the next character as well.
252             linkHotTerms(beginChar, hotTerms);
253         else
254             // In case of greedy matching the next character checking is unnecessary therefore we just store
255             // the first character.
256             addCharacter(beginChar);
257     }
258 
259     // Merge two following BeginChars in the vector to reduce the number of character checks.
mergeJSC::Yarr::BeginCharHelper260     void merge(unsigned size)
261     {
262         for (unsigned i = 0; i < size; i++) {
263             BeginChar* curr = &m_beginChars->at(i);
264             BeginChar* next = &m_beginChars->at(i + 1);
265 
266             // If the current and the next size of value is different we should skip the merge process
267             // because the 16bit and 32bit values are unmergable.
268             if (curr->value <= 0xFFFF && next->value > 0xFFFF)
269                 continue;
270 
271             unsigned diff = curr->value ^ next->value;
272 
273             curr->mask |= diff;
274             curr->value |= curr->mask;
275 
276             m_beginChars->remove(i + 1);
277             size--;
278         }
279     }
280 
281 private:
addCharacterJSC::Yarr::BeginCharHelper282     void addCharacter(BeginChar beginChar)
283     {
284         unsigned pos = 0;
285         unsigned range = m_beginChars->size();
286 
287         // binary chop, find position to insert char.
288         while (range) {
289             unsigned index = range >> 1;
290 
291             int val = m_beginChars->at(pos+index).value - beginChar.value;
292             if (!val)
293                 return;
294             if (val < 0)
295                 range = index;
296             else {
297                 pos += (index+1);
298                 range -= (index+1);
299             }
300         }
301 
302         if (pos == m_beginChars->size())
303             m_beginChars->append(beginChar);
304         else
305             m_beginChars->insert(pos, beginChar);
306     }
307 
308     // Create BeginChar objects by appending each terms from a hotTerms vector to an existing BeginChar object.
linkHotTermsJSC::Yarr::BeginCharHelper309     void linkHotTerms(BeginChar beginChar, Vector<TermChain>* hotTerms)
310     {
311         for (unsigned i = 0; i < hotTerms->size(); i++) {
312             PatternTerm hotTerm = hotTerms->at(i).term;
313             ASSERT(hotTerm.type == PatternTerm::TypePatternCharacter);
314 
315             UChar characterNext = hotTerm.patternCharacter;
316 
317             // Append a character to an existing BeginChar object.
318             if (characterNext <= 0x7f) {
319                 unsigned mask = 0;
320 
321                 if (m_isCaseInsensitive && isASCIIAlpha(characterNext)) {
322                     mask = 32;
323                     characterNext = toASCIILower(characterNext);
324                 }
325 
326                 addCharacter(BeginChar(beginChar.value | (characterNext << 16), beginChar.mask | (mask << 16)));
327             } else {
328                 UChar upper, lower;
329                 if (m_isCaseInsensitive && ((upper = Unicode::toUpper(characterNext)) != (lower = Unicode::toLower(characterNext)))) {
330                     addCharacter(BeginChar(beginChar.value | (upper << 16), beginChar.mask));
331                     addCharacter(BeginChar(beginChar.value | (lower << 16), beginChar.mask));
332                 } else
333                     addCharacter(BeginChar(beginChar.value | (characterNext << 16), beginChar.mask));
334             }
335         }
336     }
337 
338     Vector<BeginChar>* m_beginChars;
339     bool m_isCaseInsensitive;
340 };
341 
342 class YarrPatternConstructor {
343 public:
YarrPatternConstructor(YarrPattern & pattern)344     YarrPatternConstructor(YarrPattern& pattern)
345         : m_pattern(pattern)
346         , m_characterClassConstructor(pattern.m_ignoreCase)
347         , m_beginCharHelper(&pattern.m_beginChars, pattern.m_ignoreCase)
348         , m_invertParentheticalAssertion(false)
349     {
350         m_pattern.m_body = new PatternDisjunction();
351         m_alternative = m_pattern.m_body->addNewAlternative();
352         m_pattern.m_disjunctions.append(m_pattern.m_body);
353     }
354 
~YarrPatternConstructor()355     ~YarrPatternConstructor()
356     {
357     }
358 
reset()359     void reset()
360     {
361         m_pattern.reset();
362         m_characterClassConstructor.reset();
363 
364         m_pattern.m_body = new PatternDisjunction();
365         m_alternative = m_pattern.m_body->addNewAlternative();
366         m_pattern.m_disjunctions.append(m_pattern.m_body);
367     }
368 
assertionBOL()369     void assertionBOL()
370     {
371         if (!m_alternative->m_terms.size() & !m_invertParentheticalAssertion) {
372             m_alternative->m_startsWithBOL = true;
373             m_alternative->m_containsBOL = true;
374             m_pattern.m_containsBOL = true;
375         }
376         m_alternative->m_terms.append(PatternTerm::BOL());
377     }
assertionEOL()378     void assertionEOL()
379     {
380         m_alternative->m_terms.append(PatternTerm::EOL());
381     }
assertionWordBoundary(bool invert)382     void assertionWordBoundary(bool invert)
383     {
384         m_alternative->m_terms.append(PatternTerm::WordBoundary(invert));
385     }
386 
atomPatternCharacter(UChar ch)387     void atomPatternCharacter(UChar ch)
388     {
389         // We handle case-insensitive checking of unicode characters which do have both
390         // cases by handling them as if they were defined using a CharacterClass.
391         if (m_pattern.m_ignoreCase && !isASCII(ch) && (Unicode::toUpper(ch) != Unicode::toLower(ch))) {
392             atomCharacterClassBegin();
393             atomCharacterClassAtom(ch);
394             atomCharacterClassEnd();
395         } else
396             m_alternative->m_terms.append(PatternTerm(ch));
397     }
398 
atomBuiltInCharacterClass(BuiltInCharacterClassID classID,bool invert)399     void atomBuiltInCharacterClass(BuiltInCharacterClassID classID, bool invert)
400     {
401         switch (classID) {
402         case DigitClassID:
403             m_alternative->m_terms.append(PatternTerm(m_pattern.digitsCharacterClass(), invert));
404             break;
405         case SpaceClassID:
406             m_alternative->m_terms.append(PatternTerm(m_pattern.spacesCharacterClass(), invert));
407             break;
408         case WordClassID:
409             m_alternative->m_terms.append(PatternTerm(m_pattern.wordcharCharacterClass(), invert));
410             break;
411         case NewlineClassID:
412             m_alternative->m_terms.append(PatternTerm(m_pattern.newlineCharacterClass(), invert));
413             break;
414         }
415     }
416 
atomCharacterClassBegin(bool invert=false)417     void atomCharacterClassBegin(bool invert = false)
418     {
419         m_invertCharacterClass = invert;
420     }
421 
atomCharacterClassAtom(UChar ch)422     void atomCharacterClassAtom(UChar ch)
423     {
424         m_characterClassConstructor.putChar(ch);
425     }
426 
atomCharacterClassRange(UChar begin,UChar end)427     void atomCharacterClassRange(UChar begin, UChar end)
428     {
429         m_characterClassConstructor.putRange(begin, end);
430     }
431 
atomCharacterClassBuiltIn(BuiltInCharacterClassID classID,bool invert)432     void atomCharacterClassBuiltIn(BuiltInCharacterClassID classID, bool invert)
433     {
434         ASSERT(classID != NewlineClassID);
435 
436         switch (classID) {
437         case DigitClassID:
438             m_characterClassConstructor.append(invert ? m_pattern.nondigitsCharacterClass() : m_pattern.digitsCharacterClass());
439             break;
440 
441         case SpaceClassID:
442             m_characterClassConstructor.append(invert ? m_pattern.nonspacesCharacterClass() : m_pattern.spacesCharacterClass());
443             break;
444 
445         case WordClassID:
446             m_characterClassConstructor.append(invert ? m_pattern.nonwordcharCharacterClass() : m_pattern.wordcharCharacterClass());
447             break;
448 
449         default:
450             ASSERT_NOT_REACHED();
451         }
452     }
453 
atomCharacterClassEnd()454     void atomCharacterClassEnd()
455     {
456         CharacterClass* newCharacterClass = m_characterClassConstructor.charClass();
457         m_pattern.m_userCharacterClasses.append(newCharacterClass);
458         m_alternative->m_terms.append(PatternTerm(newCharacterClass, m_invertCharacterClass));
459     }
460 
atomParenthesesSubpatternBegin(bool capture=true)461     void atomParenthesesSubpatternBegin(bool capture = true)
462     {
463         unsigned subpatternId = m_pattern.m_numSubpatterns + 1;
464         if (capture)
465             m_pattern.m_numSubpatterns++;
466 
467         PatternDisjunction* parenthesesDisjunction = new PatternDisjunction(m_alternative);
468         m_pattern.m_disjunctions.append(parenthesesDisjunction);
469         m_alternative->m_terms.append(PatternTerm(PatternTerm::TypeParenthesesSubpattern, subpatternId, parenthesesDisjunction, capture, false));
470         m_alternative = parenthesesDisjunction->addNewAlternative();
471     }
472 
atomParentheticalAssertionBegin(bool invert=false)473     void atomParentheticalAssertionBegin(bool invert = false)
474     {
475         PatternDisjunction* parenthesesDisjunction = new PatternDisjunction(m_alternative);
476         m_pattern.m_disjunctions.append(parenthesesDisjunction);
477         m_alternative->m_terms.append(PatternTerm(PatternTerm::TypeParentheticalAssertion, m_pattern.m_numSubpatterns + 1, parenthesesDisjunction, false, invert));
478         m_alternative = parenthesesDisjunction->addNewAlternative();
479         m_invertParentheticalAssertion = invert;
480     }
481 
atomParenthesesEnd()482     void atomParenthesesEnd()
483     {
484         ASSERT(m_alternative->m_parent);
485         ASSERT(m_alternative->m_parent->m_parent);
486 
487         PatternDisjunction* parenthesesDisjunction = m_alternative->m_parent;
488         m_alternative = m_alternative->m_parent->m_parent;
489 
490         PatternTerm& lastTerm = m_alternative->lastTerm();
491 
492         unsigned numParenAlternatives = parenthesesDisjunction->m_alternatives.size();
493         unsigned numBOLAnchoredAlts = 0;
494         bool containsEmptyAlternative = false;
495 
496         for (unsigned i = 0; i < numParenAlternatives; i++) {
497             if (!parenthesesDisjunction->m_alternatives[i]->m_terms.size() && numParenAlternatives > 1) {
498                 PatternAlternative* altToRemove = parenthesesDisjunction->m_alternatives[i];
499                 parenthesesDisjunction->m_alternatives.remove(i);
500                 delete altToRemove;
501                 --numParenAlternatives;
502 
503                 containsEmptyAlternative = true;
504                 continue;
505             }
506 
507             // Bubble up BOL flags
508             if (parenthesesDisjunction->m_alternatives[i]->m_startsWithBOL)
509                 numBOLAnchoredAlts++;
510         }
511 
512         if (numBOLAnchoredAlts) {
513             m_alternative->m_containsBOL = true;
514             // If all the alternatives in parens start with BOL, then so does this one
515             if (numBOLAnchoredAlts == numParenAlternatives)
516                 m_alternative->m_startsWithBOL = true;
517         }
518 
519         lastTerm.parentheses.lastSubpatternId = m_pattern.m_numSubpatterns;
520         m_invertParentheticalAssertion = false;
521 
522         if (containsEmptyAlternative) {
523             // Backup and remove the current disjunction's alternatives.
524             Vector<PatternAlternative*> alternatives;
525             alternatives.append(parenthesesDisjunction->m_alternatives);
526             parenthesesDisjunction->m_alternatives.clear();
527             PatternAlternative* alternative = parenthesesDisjunction->addNewAlternative();
528 
529             // Insert a new non-capturing parentheses.
530             unsigned subpatternId = m_pattern.m_numSubpatterns + 1;
531             PatternDisjunction* newDisjunction = new PatternDisjunction(alternative);
532             m_pattern.m_disjunctions.append(newDisjunction);
533             alternative->m_terms.append(PatternTerm(PatternTerm::TypeParenthesesSubpattern, subpatternId, newDisjunction, false, false));
534             newDisjunction->m_alternatives.append(alternatives);
535 
536             // Set the quantifier of the new parentheses to '?' and set the inherited properties.
537             PatternTerm& disjunctionTerm = alternative->lastTerm();
538             disjunctionTerm.quantify(1, QuantifierGreedy);
539             disjunctionTerm.parentheses.lastSubpatternId = m_pattern.m_numSubpatterns;
540             alternative->m_containsBOL = m_alternative->m_containsBOL;
541             alternative->m_startsWithBOL = m_alternative->m_startsWithBOL;
542         }
543     }
544 
atomBackReference(unsigned subpatternId)545     void atomBackReference(unsigned subpatternId)
546     {
547         ASSERT(subpatternId);
548         m_pattern.m_containsBackreferences = true;
549         m_pattern.m_maxBackReference = std::max(m_pattern.m_maxBackReference, subpatternId);
550 
551         if (subpatternId > m_pattern.m_numSubpatterns) {
552             m_alternative->m_terms.append(PatternTerm::ForwardReference());
553             return;
554         }
555 
556         PatternAlternative* currentAlternative = m_alternative;
557         ASSERT(currentAlternative);
558 
559         // Note to self: if we waited until the AST was baked, we could also remove forwards refs
560         while ((currentAlternative = currentAlternative->m_parent->m_parent)) {
561             PatternTerm& term = currentAlternative->lastTerm();
562             ASSERT((term.type == PatternTerm::TypeParenthesesSubpattern) || (term.type == PatternTerm::TypeParentheticalAssertion));
563 
564             if ((term.type == PatternTerm::TypeParenthesesSubpattern) && term.capture() && (subpatternId == term.parentheses.subpatternId)) {
565                 m_alternative->m_terms.append(PatternTerm::ForwardReference());
566                 return;
567             }
568         }
569 
570         m_alternative->m_terms.append(PatternTerm(subpatternId));
571     }
572 
573     // deep copy the argument disjunction.  If filterStartsWithBOL is true,
574     // skip alternatives with m_startsWithBOL set true.
copyDisjunction(PatternDisjunction * disjunction,bool filterStartsWithBOL=false)575     PatternDisjunction* copyDisjunction(PatternDisjunction* disjunction, bool filterStartsWithBOL = false)
576     {
577         PatternDisjunction* newDisjunction = 0;
578         for (unsigned alt = 0; alt < disjunction->m_alternatives.size(); ++alt) {
579             PatternAlternative* alternative = disjunction->m_alternatives[alt];
580             if (!filterStartsWithBOL || !alternative->m_startsWithBOL) {
581                 if (!newDisjunction) {
582                     newDisjunction = new PatternDisjunction();
583                     newDisjunction->m_parent = disjunction->m_parent;
584                 }
585                 PatternAlternative* newAlternative = newDisjunction->addNewAlternative();
586                 for (unsigned i = 0; i < alternative->m_terms.size(); ++i)
587                     newAlternative->m_terms.append(copyTerm(alternative->m_terms[i], filterStartsWithBOL));
588             }
589         }
590 
591         if (newDisjunction)
592             m_pattern.m_disjunctions.append(newDisjunction);
593         return newDisjunction;
594     }
595 
copyTerm(PatternTerm & term,bool filterStartsWithBOL=false)596     PatternTerm copyTerm(PatternTerm& term, bool filterStartsWithBOL = false)
597     {
598         if ((term.type != PatternTerm::TypeParenthesesSubpattern) && (term.type != PatternTerm::TypeParentheticalAssertion))
599             return PatternTerm(term);
600 
601         PatternTerm termCopy = term;
602         termCopy.parentheses.disjunction = copyDisjunction(termCopy.parentheses.disjunction, filterStartsWithBOL);
603         return termCopy;
604     }
605 
quantifyAtom(unsigned min,unsigned max,bool greedy)606     void quantifyAtom(unsigned min, unsigned max, bool greedy)
607     {
608         ASSERT(min <= max);
609         ASSERT(m_alternative->m_terms.size());
610 
611         if (!max) {
612             m_alternative->removeLastTerm();
613             return;
614         }
615 
616         PatternTerm& term = m_alternative->lastTerm();
617         ASSERT(term.type > PatternTerm::TypeAssertionWordBoundary);
618         ASSERT((term.quantityCount == 1) && (term.quantityType == QuantifierFixedCount));
619 
620         // For any assertion with a zero minimum, not matching is valid and has no effect,
621         // remove it.  Otherwise, we need to match as least once, but there is no point
622         // matching more than once, so remove the quantifier.  It is not entirely clear
623         // from the spec whether or not this behavior is correct, but I believe this
624         // matches Firefox. :-/
625         if (term.type == PatternTerm::TypeParentheticalAssertion) {
626             if (!min)
627                 m_alternative->removeLastTerm();
628             return;
629         }
630 
631         if (min == 0)
632             term.quantify(max, greedy   ? QuantifierGreedy : QuantifierNonGreedy);
633         else if (min == max)
634             term.quantify(min, QuantifierFixedCount);
635         else {
636             term.quantify(min, QuantifierFixedCount);
637             m_alternative->m_terms.append(copyTerm(term));
638             // NOTE: this term is interesting from an analysis perspective, in that it can be ignored.....
639             m_alternative->lastTerm().quantify((max == quantifyInfinite) ? max : max - min, greedy ? QuantifierGreedy : QuantifierNonGreedy);
640             if (m_alternative->lastTerm().type == PatternTerm::TypeParenthesesSubpattern)
641                 m_alternative->lastTerm().parentheses.isCopy = true;
642         }
643     }
644 
disjunction()645     void disjunction()
646     {
647         m_alternative = m_alternative->m_parent->addNewAlternative();
648     }
649 
setupAlternativeOffsets(PatternAlternative * alternative,unsigned currentCallFrameSize,unsigned initialInputPosition)650     unsigned setupAlternativeOffsets(PatternAlternative* alternative, unsigned currentCallFrameSize, unsigned initialInputPosition)
651     {
652         alternative->m_hasFixedSize = true;
653         unsigned currentInputPosition = initialInputPosition;
654 
655         for (unsigned i = 0; i < alternative->m_terms.size(); ++i) {
656             PatternTerm& term = alternative->m_terms[i];
657 
658             switch (term.type) {
659             case PatternTerm::TypeAssertionBOL:
660             case PatternTerm::TypeAssertionEOL:
661             case PatternTerm::TypeAssertionWordBoundary:
662                 term.inputPosition = currentInputPosition;
663                 break;
664 
665             case PatternTerm::TypeBackReference:
666                 term.inputPosition = currentInputPosition;
667                 term.frameLocation = currentCallFrameSize;
668                 currentCallFrameSize += YarrStackSpaceForBackTrackInfoBackReference;
669                 alternative->m_hasFixedSize = false;
670                 break;
671 
672             case PatternTerm::TypeForwardReference:
673                 break;
674 
675             case PatternTerm::TypePatternCharacter:
676                 term.inputPosition = currentInputPosition;
677                 if (term.quantityType != QuantifierFixedCount) {
678                     term.frameLocation = currentCallFrameSize;
679                     currentCallFrameSize += YarrStackSpaceForBackTrackInfoPatternCharacter;
680                     alternative->m_hasFixedSize = false;
681                 } else
682                     currentInputPosition += term.quantityCount;
683                 break;
684 
685             case PatternTerm::TypeCharacterClass:
686                 term.inputPosition = currentInputPosition;
687                 if (term.quantityType != QuantifierFixedCount) {
688                     term.frameLocation = currentCallFrameSize;
689                     currentCallFrameSize += YarrStackSpaceForBackTrackInfoCharacterClass;
690                     alternative->m_hasFixedSize = false;
691                 } else
692                     currentInputPosition += term.quantityCount;
693                 break;
694 
695             case PatternTerm::TypeParenthesesSubpattern:
696                 // Note: for fixed once parentheses we will ensure at least the minimum is available; others are on their own.
697                 term.frameLocation = currentCallFrameSize;
698                 if (term.quantityCount == 1 && !term.parentheses.isCopy) {
699                     if (term.quantityType != QuantifierFixedCount)
700                         currentCallFrameSize += YarrStackSpaceForBackTrackInfoParenthesesOnce;
701                     currentCallFrameSize = setupDisjunctionOffsets(term.parentheses.disjunction, currentCallFrameSize, currentInputPosition);
702                     // If quantity is fixed, then pre-check its minimum size.
703                     if (term.quantityType == QuantifierFixedCount)
704                         currentInputPosition += term.parentheses.disjunction->m_minimumSize;
705                     term.inputPosition = currentInputPosition;
706                 } else if (term.parentheses.isTerminal) {
707                     currentCallFrameSize += YarrStackSpaceForBackTrackInfoParenthesesTerminal;
708                     currentCallFrameSize = setupDisjunctionOffsets(term.parentheses.disjunction, currentCallFrameSize, currentInputPosition);
709                     term.inputPosition = currentInputPosition;
710                 } else {
711                     term.inputPosition = currentInputPosition;
712                     setupDisjunctionOffsets(term.parentheses.disjunction, 0, currentInputPosition);
713                     currentCallFrameSize += YarrStackSpaceForBackTrackInfoParentheses;
714                 }
715                 // Fixed count of 1 could be accepted, if they have a fixed size *AND* if all alternatives are of the same length.
716                 alternative->m_hasFixedSize = false;
717                 break;
718 
719             case PatternTerm::TypeParentheticalAssertion:
720                 term.inputPosition = currentInputPosition;
721                 term.frameLocation = currentCallFrameSize;
722                 currentCallFrameSize = setupDisjunctionOffsets(term.parentheses.disjunction, currentCallFrameSize + YarrStackSpaceForBackTrackInfoParentheticalAssertion, currentInputPosition);
723                 break;
724             }
725         }
726 
727         alternative->m_minimumSize = currentInputPosition - initialInputPosition;
728         return currentCallFrameSize;
729     }
730 
setupDisjunctionOffsets(PatternDisjunction * disjunction,unsigned initialCallFrameSize,unsigned initialInputPosition)731     unsigned setupDisjunctionOffsets(PatternDisjunction* disjunction, unsigned initialCallFrameSize, unsigned initialInputPosition)
732     {
733         if ((disjunction != m_pattern.m_body) && (disjunction->m_alternatives.size() > 1))
734             initialCallFrameSize += YarrStackSpaceForBackTrackInfoAlternative;
735 
736         unsigned minimumInputSize = UINT_MAX;
737         unsigned maximumCallFrameSize = 0;
738         bool hasFixedSize = true;
739 
740         for (unsigned alt = 0; alt < disjunction->m_alternatives.size(); ++alt) {
741             PatternAlternative* alternative = disjunction->m_alternatives[alt];
742             unsigned currentAlternativeCallFrameSize = setupAlternativeOffsets(alternative, initialCallFrameSize, initialInputPosition);
743             minimumInputSize = min(minimumInputSize, alternative->m_minimumSize);
744             maximumCallFrameSize = max(maximumCallFrameSize, currentAlternativeCallFrameSize);
745             hasFixedSize &= alternative->m_hasFixedSize;
746         }
747 
748         ASSERT(minimumInputSize != UINT_MAX);
749         ASSERT(maximumCallFrameSize >= initialCallFrameSize);
750 
751         disjunction->m_hasFixedSize = hasFixedSize;
752         disjunction->m_minimumSize = minimumInputSize;
753         disjunction->m_callFrameSize = maximumCallFrameSize;
754         return maximumCallFrameSize;
755     }
756 
setupOffsets()757     void setupOffsets()
758     {
759         setupDisjunctionOffsets(m_pattern.m_body, 0, 0);
760     }
761 
762     // This optimization identifies sets of parentheses that we will never need to backtrack.
763     // In these cases we do not need to store state from prior iterations.
764     // We can presently avoid backtracking for:
765     //   * where the parens are at the end of the regular expression (last term in any of the
766     //     alternatives of the main body disjunction).
767     //   * where the parens are non-capturing, and quantified unbounded greedy (*).
768     //   * where the parens do not contain any capturing subpatterns.
checkForTerminalParentheses()769     void checkForTerminalParentheses()
770     {
771         // This check is much too crude; should be just checking whether the candidate
772         // node contains nested capturing subpatterns, not the whole expression!
773         if (m_pattern.m_numSubpatterns)
774             return;
775 
776         Vector<PatternAlternative*>& alternatives = m_pattern.m_body->m_alternatives;
777         for (size_t i = 0; i < alternatives.size(); ++i) {
778             Vector<PatternTerm>& terms = alternatives[i]->m_terms;
779             if (terms.size()) {
780                 PatternTerm& term = terms.last();
781                 if (term.type == PatternTerm::TypeParenthesesSubpattern
782                     && term.quantityType == QuantifierGreedy
783                     && term.quantityCount == quantifyInfinite
784                     && !term.capture())
785                     term.parentheses.isTerminal = true;
786             }
787         }
788     }
789 
optimizeBOL()790     void optimizeBOL()
791     {
792         // Look for expressions containing beginning of line (^) anchoring and unroll them.
793         // e.g. /^a|^b|c/ becomes /^a|^b|c/ which is executed once followed by /c/ which loops
794         // This code relies on the parsing code tagging alternatives with m_containsBOL and
795         // m_startsWithBOL and rolling those up to containing alternatives.
796         // At this point, this is only valid for non-multiline expressions.
797         PatternDisjunction* disjunction = m_pattern.m_body;
798 
799         if (!m_pattern.m_containsBOL || m_pattern.m_multiline)
800             return;
801 
802         PatternDisjunction* loopDisjunction = copyDisjunction(disjunction, true);
803 
804         // Set alternatives in disjunction to "onceThrough"
805         for (unsigned alt = 0; alt < disjunction->m_alternatives.size(); ++alt)
806             disjunction->m_alternatives[alt]->setOnceThrough();
807 
808         if (loopDisjunction) {
809             // Move alternatives from loopDisjunction to disjunction
810             for (unsigned alt = 0; alt < loopDisjunction->m_alternatives.size(); ++alt)
811                 disjunction->m_alternatives.append(loopDisjunction->m_alternatives[alt]);
812 
813             loopDisjunction->m_alternatives.clear();
814         }
815     }
816 
817     // This function collects the terms which are potentially matching the first number of depth characters in the result.
818     // If this function returns false then it found at least one term which makes the beginning character
819     // look-up optimization inefficient.
setupDisjunctionBeginTerms(PatternDisjunction * disjunction,Vector<TermChain> * beginTerms,unsigned depth)820     bool setupDisjunctionBeginTerms(PatternDisjunction* disjunction, Vector<TermChain>* beginTerms, unsigned depth)
821     {
822         for (unsigned alt = 0; alt < disjunction->m_alternatives.size(); ++alt) {
823             PatternAlternative* alternative = disjunction->m_alternatives[alt];
824 
825             if (!setupAlternativeBeginTerms(alternative, beginTerms, 0, depth))
826                 return false;
827         }
828 
829         return true;
830     }
831 
setupAlternativeBeginTerms(PatternAlternative * alternative,Vector<TermChain> * beginTerms,unsigned termIndex,unsigned depth)832     bool setupAlternativeBeginTerms(PatternAlternative* alternative, Vector<TermChain>* beginTerms, unsigned termIndex, unsigned depth)
833     {
834         bool checkNext = true;
835         unsigned numTerms = alternative->m_terms.size();
836 
837         while (checkNext && termIndex < numTerms) {
838             PatternTerm term = alternative->m_terms[termIndex];
839             checkNext = false;
840 
841             switch (term.type) {
842             case PatternTerm::TypeAssertionBOL:
843             case PatternTerm::TypeAssertionEOL:
844             case PatternTerm::TypeAssertionWordBoundary:
845                 return false;
846 
847             case PatternTerm::TypeBackReference:
848             case PatternTerm::TypeForwardReference:
849                 return false;
850 
851             case PatternTerm::TypePatternCharacter:
852                 if (termIndex != numTerms - 1) {
853                     beginTerms->append(TermChain(term));
854                     termIndex++;
855                     checkNext = true;
856                 } else if (term.quantityType == QuantifierFixedCount) {
857                     beginTerms->append(TermChain(term));
858                     if (depth < 2 && termIndex < numTerms - 1 && term.quantityCount == 1)
859                         if (!setupAlternativeBeginTerms(alternative, &beginTerms->last().hotTerms, termIndex + 1, depth + 1))
860                             return false;
861                 }
862 
863                 break;
864 
865             case PatternTerm::TypeCharacterClass:
866                 return false;
867 
868             case PatternTerm::TypeParentheticalAssertion:
869                 if (term.invert())
870                     return false;
871 
872             case PatternTerm::TypeParenthesesSubpattern:
873                 if (term.quantityType != QuantifierFixedCount) {
874                     if (termIndex == numTerms - 1)
875                         break;
876 
877                     termIndex++;
878                     checkNext = true;
879                 }
880 
881                 if (!setupDisjunctionBeginTerms(term.parentheses.disjunction, beginTerms, depth))
882                     return false;
883 
884                 break;
885             }
886         }
887 
888         return true;
889     }
890 
setupBeginChars()891     void setupBeginChars()
892     {
893         Vector<TermChain> beginTerms;
894         bool containsFixedCharacter = false;
895 
896         if ((!m_pattern.m_body->m_hasFixedSize || m_pattern.m_body->m_alternatives.size() > 1)
897                 && setupDisjunctionBeginTerms(m_pattern.m_body, &beginTerms, 0)) {
898             unsigned size = beginTerms.size();
899 
900             // If we haven't collected any terms we should abort the preparation of beginning character look-up optimization.
901             if (!size)
902                 return;
903 
904             m_pattern.m_containsBeginChars = true;
905 
906             for (unsigned i = 0; i < size; i++) {
907                 PatternTerm term = beginTerms[i].term;
908 
909                 // We have just collected PatternCharacter terms, other terms are not allowed.
910                 ASSERT(term.type == PatternTerm::TypePatternCharacter);
911 
912                 if (term.quantityType == QuantifierFixedCount)
913                     containsFixedCharacter = true;
914 
915                 UChar character = term.patternCharacter;
916                 unsigned mask = 0;
917 
918                 if (character <= 0x7f) {
919                     if (m_pattern.m_ignoreCase && isASCIIAlpha(character)) {
920                         mask = 32;
921                         character = toASCIILower(character);
922                     }
923 
924                     m_beginCharHelper.addBeginChar(BeginChar(character, mask), &beginTerms[i].hotTerms, term.quantityType, term.quantityCount);
925                 } else {
926                     UChar upper, lower;
927                     if (m_pattern.m_ignoreCase && ((upper = Unicode::toUpper(character)) != (lower = Unicode::toLower(character)))) {
928                         m_beginCharHelper.addBeginChar(BeginChar(upper, mask), &beginTerms[i].hotTerms, term.quantityType, term.quantityCount);
929                         m_beginCharHelper.addBeginChar(BeginChar(lower, mask), &beginTerms[i].hotTerms, term.quantityType, term.quantityCount);
930                     } else
931                         m_beginCharHelper.addBeginChar(BeginChar(character, mask), &beginTerms[i].hotTerms, term.quantityType, term.quantityCount);
932                 }
933             }
934 
935             // If the pattern doesn't contain terms with fixed quantifiers then the beginning character look-up optimization is inefficient.
936             if (!containsFixedCharacter) {
937                 m_pattern.m_containsBeginChars = false;
938                 return;
939             }
940 
941             size = m_pattern.m_beginChars.size();
942 
943             if (size > 2)
944                 m_beginCharHelper.merge(size - 1);
945             else if (size <= 1)
946                 m_pattern.m_containsBeginChars = false;
947         }
948     }
949 
950 private:
951     YarrPattern& m_pattern;
952     PatternAlternative* m_alternative;
953     CharacterClassConstructor m_characterClassConstructor;
954     BeginCharHelper m_beginCharHelper;
955     bool m_invertCharacterClass;
956     bool m_invertParentheticalAssertion;
957 };
958 
compile(const UString & patternString)959 const char* YarrPattern::compile(const UString& patternString)
960 {
961     YarrPatternConstructor constructor(*this);
962 
963     if (const char* error = parse(constructor, patternString))
964         return error;
965 
966     // If the pattern contains illegal backreferences reset & reparse.
967     // Quoting Netscape's "What's new in JavaScript 1.2",
968     //      "Note: if the number of left parentheses is less than the number specified
969     //       in \#, the \# is taken as an octal escape as described in the next row."
970     if (containsIllegalBackReference()) {
971         unsigned numSubpatterns = m_numSubpatterns;
972 
973         constructor.reset();
974 #if !ASSERT_DISABLED
975         const char* error =
976 #endif
977             parse(constructor, patternString, numSubpatterns);
978 
979         ASSERT(!error);
980         ASSERT(numSubpatterns == m_numSubpatterns);
981     }
982 
983     constructor.checkForTerminalParentheses();
984     constructor.optimizeBOL();
985 
986     constructor.setupOffsets();
987     constructor.setupBeginChars();
988 
989     return 0;
990 }
991 
YarrPattern(const UString & pattern,bool ignoreCase,bool multiline,const char ** error)992 YarrPattern::YarrPattern(const UString& pattern, bool ignoreCase, bool multiline, const char** error)
993     : m_ignoreCase(ignoreCase)
994     , m_multiline(multiline)
995     , m_containsBackreferences(false)
996     , m_containsBeginChars(false)
997     , m_containsBOL(false)
998     , m_numSubpatterns(0)
999     , m_maxBackReference(0)
1000     , newlineCached(0)
1001     , digitsCached(0)
1002     , spacesCached(0)
1003     , wordcharCached(0)
1004     , nondigitsCached(0)
1005     , nonspacesCached(0)
1006     , nonwordcharCached(0)
1007 {
1008     *error = compile(pattern);
1009 }
1010 
1011 } }
1012