1 // Copyright 2019 the V8 project authors. All rights reserved.
2 // Use of this source code is governed by a BSD-style license that can be
3 // found in the LICENSE file.
4
5 #ifndef V8_REGEXP_REGEXP_COMPILER_H_
6 #define V8_REGEXP_REGEXP_COMPILER_H_
7
8 #include <bitset>
9
10 #include "src/base/small-vector.h"
11 #include "src/base/strings.h"
12 #include "src/regexp/regexp-flags.h"
13 #include "src/regexp/regexp-nodes.h"
14
15 namespace v8 {
16 namespace internal {
17
18 class DynamicBitSet;
19 class Isolate;
20
21 namespace regexp_compiler_constants {
22
23 // The '2' variant is has inclusive from and exclusive to.
24 // This covers \s as defined in ECMA-262 5.1, 15.10.2.12,
25 // which include WhiteSpace (7.2) or LineTerminator (7.3) values.
26 constexpr base::uc32 kRangeEndMarker = 0x110000;
27 constexpr int kSpaceRanges[] = {
28 '\t', '\r' + 1, ' ', ' ' + 1, 0x00A0, 0x00A1, 0x1680,
29 0x1681, 0x2000, 0x200B, 0x2028, 0x202A, 0x202F, 0x2030,
30 0x205F, 0x2060, 0x3000, 0x3001, 0xFEFF, 0xFF00, kRangeEndMarker};
31 constexpr int kSpaceRangeCount = arraysize(kSpaceRanges);
32
33 constexpr int kWordRanges[] = {'0', '9' + 1, 'A', 'Z' + 1, '_',
34 '_' + 1, 'a', 'z' + 1, kRangeEndMarker};
35 constexpr int kWordRangeCount = arraysize(kWordRanges);
36 constexpr int kDigitRanges[] = {'0', '9' + 1, kRangeEndMarker};
37 constexpr int kDigitRangeCount = arraysize(kDigitRanges);
38 constexpr int kSurrogateRanges[] = {kLeadSurrogateStart,
39 kLeadSurrogateStart + 1, kRangeEndMarker};
40 constexpr int kSurrogateRangeCount = arraysize(kSurrogateRanges);
41 constexpr int kLineTerminatorRanges[] = {0x000A, 0x000B, 0x000D, 0x000E,
42 0x2028, 0x202A, kRangeEndMarker};
43 constexpr int kLineTerminatorRangeCount = arraysize(kLineTerminatorRanges);
44
45 // More makes code generation slower, less makes V8 benchmark score lower.
46 constexpr int kMaxLookaheadForBoyerMoore = 8;
47 // In a 3-character pattern you can maximally step forwards 3 characters
48 // at a time, which is not always enough to pay for the extra logic.
49 constexpr int kPatternTooShortForBoyerMoore = 2;
50
51 } // namespace regexp_compiler_constants
52
NeedsUnicodeCaseEquivalents(RegExpFlags flags)53 inline bool NeedsUnicodeCaseEquivalents(RegExpFlags flags) {
54 // Both unicode and ignore_case flags are set. We need to use ICU to find
55 // the closure over case equivalents.
56 return IsUnicode(flags) && IsIgnoreCase(flags);
57 }
58
59 // Details of a quick mask-compare check that can look ahead in the
60 // input stream.
61 class QuickCheckDetails {
62 public:
QuickCheckDetails()63 QuickCheckDetails()
64 : characters_(0), mask_(0), value_(0), cannot_match_(false) {}
QuickCheckDetails(int characters)65 explicit QuickCheckDetails(int characters)
66 : characters_(characters), mask_(0), value_(0), cannot_match_(false) {}
67 bool Rationalize(bool one_byte);
68 // Merge in the information from another branch of an alternation.
69 void Merge(QuickCheckDetails* other, int from_index);
70 // Advance the current position by some amount.
71 void Advance(int by, bool one_byte);
72 void Clear();
cannot_match()73 bool cannot_match() { return cannot_match_; }
set_cannot_match()74 void set_cannot_match() { cannot_match_ = true; }
75 struct Position {
Positionv8::internal::QuickCheckDetails::Position76 Position() : mask(0), value(0), determines_perfectly(false) {}
77 base::uc32 mask;
78 base::uc32 value;
79 bool determines_perfectly;
80 };
characters()81 int characters() { return characters_; }
set_characters(int characters)82 void set_characters(int characters) { characters_ = characters; }
positions(int index)83 Position* positions(int index) {
84 DCHECK_LE(0, index);
85 DCHECK_GT(characters_, index);
86 return positions_ + index;
87 }
mask()88 uint32_t mask() { return mask_; }
value()89 uint32_t value() { return value_; }
90
91 private:
92 // How many characters do we have quick check information from. This is
93 // the same for all branches of a choice node.
94 int characters_;
95 Position positions_[4];
96 // These values are the condensate of the above array after Rationalize().
97 uint32_t mask_;
98 uint32_t value_;
99 // If set to true, there is no way this quick check can match at all.
100 // E.g., if it requires to be at the start of the input, and isn't.
101 bool cannot_match_;
102 };
103
104 // Improve the speed that we scan for an initial point where a non-anchored
105 // regexp can match by using a Boyer-Moore-like table. This is done by
106 // identifying non-greedy non-capturing loops in the nodes that eat any
107 // character one at a time. For example in the middle of the regexp
108 // /foo[\s\S]*?bar/ we find such a loop. There is also such a loop implicitly
109 // inserted at the start of any non-anchored regexp.
110 //
111 // When we have found such a loop we look ahead in the nodes to find the set of
112 // characters that can come at given distances. For example for the regexp
113 // /.?foo/ we know that there are at least 3 characters ahead of us, and the
114 // sets of characters that can occur are [any, [f, o], [o]]. We find a range in
115 // the lookahead info where the set of characters is reasonably constrained. In
116 // our example this is from index 1 to 2 (0 is not constrained). We can now
117 // look 3 characters ahead and if we don't find one of [f, o] (the union of
118 // [f, o] and [o]) then we can skip forwards by the range size (in this case 2).
119 //
120 // For Unicode input strings we do the same, but modulo 128.
121 //
122 // We also look at the first string fed to the regexp and use that to get a hint
123 // of the character frequencies in the inputs. This affects the assessment of
124 // whether the set of characters is 'reasonably constrained'.
125 //
126 // We also have another lookahead mechanism (called quick check in the code),
127 // which uses a wide load of multiple characters followed by a mask and compare
128 // to determine whether a match is possible at this point.
129 enum ContainedInLattice {
130 kNotYet = 0,
131 kLatticeIn = 1,
132 kLatticeOut = 2,
133 kLatticeUnknown = 3 // Can also mean both in and out.
134 };
135
Combine(ContainedInLattice a, ContainedInLattice b)136 inline ContainedInLattice Combine(ContainedInLattice a, ContainedInLattice b) {
137 return static_cast<ContainedInLattice>(a | b);
138 }
139
140 class BoyerMoorePositionInfo : public ZoneObject {
141 public:
at(int i) const142 bool at(int i) const { return map_[i]; }
143
144 static constexpr int kMapSize = 128;
145 static constexpr int kMask = kMapSize - 1;
146
map_count() const147 int map_count() const { return map_count_; }
148
149 void Set(int character);
150 void SetInterval(const Interval& interval);
151 void SetAll();
152
is_non_word()153 bool is_non_word() { return w_ == kLatticeOut; }
is_word()154 bool is_word() { return w_ == kLatticeIn; }
155
156 using Bitset = std::bitset<kMapSize>;
raw_bitset() const157 Bitset raw_bitset() const { return map_; }
158
159 private:
160 Bitset map_;
161 int map_count_ = 0; // Number of set bits in the map.
162 ContainedInLattice w_ = kNotYet; // The \w character class.
163 };
164
165 class BoyerMooreLookahead : public ZoneObject {
166 public:
167 BoyerMooreLookahead(int length, RegExpCompiler* compiler, Zone* zone);
168
length()169 int length() { return length_; }
max_char()170 int max_char() { return max_char_; }
compiler()171 RegExpCompiler* compiler() { return compiler_; }
172
Count(int map_number)173 int Count(int map_number) { return bitmaps_->at(map_number)->map_count(); }
174
at(int i)175 BoyerMoorePositionInfo* at(int i) { return bitmaps_->at(i); }
176
Set(int map_number, int character)177 void Set(int map_number, int character) {
178 if (character > max_char_) return;
179 BoyerMoorePositionInfo* info = bitmaps_->at(map_number);
180 info->Set(character);
181 }
182
SetInterval(int map_number, const Interval& interval)183 void SetInterval(int map_number, const Interval& interval) {
184 if (interval.from() > max_char_) return;
185 BoyerMoorePositionInfo* info = bitmaps_->at(map_number);
186 if (interval.to() > max_char_) {
187 info->SetInterval(Interval(interval.from(), max_char_));
188 } else {
189 info->SetInterval(interval);
190 }
191 }
192
SetAll(int map_number)193 void SetAll(int map_number) { bitmaps_->at(map_number)->SetAll(); }
194
SetRest(int from_map)195 void SetRest(int from_map) {
196 for (int i = from_map; i < length_; i++) SetAll(i);
197 }
198 void EmitSkipInstructions(RegExpMacroAssembler* masm);
199
200 private:
201 // This is the value obtained by EatsAtLeast. If we do not have at least this
202 // many characters left in the sample string then the match is bound to fail.
203 // Therefore it is OK to read a character this far ahead of the current match
204 // point.
205 int length_;
206 RegExpCompiler* compiler_;
207 // 0xff for Latin1, 0xffff for UTF-16.
208 int max_char_;
209 ZoneList<BoyerMoorePositionInfo*>* bitmaps_;
210
211 int GetSkipTable(int min_lookahead, int max_lookahead,
212 Handle<ByteArray> boolean_skip_table);
213 bool FindWorthwhileInterval(int* from, int* to);
214 int FindBestInterval(int max_number_of_chars, int old_biggest_points,
215 int* from, int* to);
216 };
217
218 // There are many ways to generate code for a node. This class encapsulates
219 // the current way we should be generating. In other words it encapsulates
220 // the current state of the code generator. The effect of this is that we
221 // generate code for paths that the matcher can take through the regular
222 // expression. A given node in the regexp can be code-generated several times
223 // as it can be part of several traces. For example for the regexp:
224 // /foo(bar|ip)baz/ the code to match baz will be generated twice, once as part
225 // of the foo-bar-baz trace and once as part of the foo-ip-baz trace. The code
226 // to match foo is generated only once (the traces have a common prefix). The
227 // code to store the capture is deferred and generated (twice) after the places
228 // where baz has been matched.
229 class Trace {
230 public:
231 // A value for a property that is either known to be true, know to be false,
232 // or not known.
233 enum TriBool { UNKNOWN = -1, FALSE_VALUE = 0, TRUE_VALUE = 1 };
234
235 class DeferredAction {
236 public:
DeferredAction(ActionNode::ActionType action_type, int reg)237 DeferredAction(ActionNode::ActionType action_type, int reg)
238 : action_type_(action_type), reg_(reg), next_(nullptr) {}
next()239 DeferredAction* next() { return next_; }
240 bool Mentions(int reg);
reg()241 int reg() { return reg_; }
action_type()242 ActionNode::ActionType action_type() { return action_type_; }
243
244 private:
245 ActionNode::ActionType action_type_;
246 int reg_;
247 DeferredAction* next_;
248 friend class Trace;
249 };
250
251 class DeferredCapture : public DeferredAction {
252 public:
DeferredCapture(int reg, bool is_capture, Trace* trace)253 DeferredCapture(int reg, bool is_capture, Trace* trace)
254 : DeferredAction(ActionNode::STORE_POSITION, reg),
255 cp_offset_(trace->cp_offset()),
256 is_capture_(is_capture) {}
cp_offset()257 int cp_offset() { return cp_offset_; }
is_capture()258 bool is_capture() { return is_capture_; }
259
260 private:
261 int cp_offset_;
262 bool is_capture_;
set_cp_offset(int cp_offset)263 void set_cp_offset(int cp_offset) { cp_offset_ = cp_offset; }
264 };
265
266 class DeferredSetRegisterForLoop : public DeferredAction {
267 public:
DeferredSetRegisterForLoop(int reg, int value)268 DeferredSetRegisterForLoop(int reg, int value)
269 : DeferredAction(ActionNode::SET_REGISTER_FOR_LOOP, reg),
270 value_(value) {}
value()271 int value() { return value_; }
272
273 private:
274 int value_;
275 };
276
277 class DeferredClearCaptures : public DeferredAction {
278 public:
DeferredClearCaptures(Interval range)279 explicit DeferredClearCaptures(Interval range)
280 : DeferredAction(ActionNode::CLEAR_CAPTURES, -1), range_(range) {}
range()281 Interval range() { return range_; }
282
283 private:
284 Interval range_;
285 };
286
287 class DeferredIncrementRegister : public DeferredAction {
288 public:
DeferredIncrementRegister(int reg)289 explicit DeferredIncrementRegister(int reg)
290 : DeferredAction(ActionNode::INCREMENT_REGISTER, reg) {}
291 };
292
Trace()293 Trace()
294 : cp_offset_(0),
295 actions_(nullptr),
296 backtrack_(nullptr),
297 stop_node_(nullptr),
298 loop_label_(nullptr),
299 characters_preloaded_(0),
300 bound_checked_up_to_(0),
301 flush_budget_(100),
302 at_start_(UNKNOWN) {}
303
304 // End the trace. This involves flushing the deferred actions in the trace
305 // and pushing a backtrack location onto the backtrack stack. Once this is
306 // done we can start a new trace or go to one that has already been
307 // generated.
308 void Flush(RegExpCompiler* compiler, RegExpNode* successor);
cp_offset()309 int cp_offset() { return cp_offset_; }
actions()310 DeferredAction* actions() { return actions_; }
311 // A trivial trace is one that has no deferred actions or other state that
312 // affects the assumptions used when generating code. There is no recorded
313 // backtrack location in a trivial trace, so with a trivial trace we will
314 // generate code that, on a failure to match, gets the backtrack location
315 // from the backtrack stack rather than using a direct jump instruction. We
316 // always start code generation with a trivial trace and non-trivial traces
317 // are created as we emit code for nodes or add to the list of deferred
318 // actions in the trace. The location of the code generated for a node using
319 // a trivial trace is recorded in a label in the node so that gotos can be
320 // generated to that code.
is_trivial()321 bool is_trivial() {
322 return backtrack_ == nullptr && actions_ == nullptr && cp_offset_ == 0 &&
323 characters_preloaded_ == 0 && bound_checked_up_to_ == 0 &&
324 quick_check_performed_.characters() == 0 && at_start_ == UNKNOWN;
325 }
at_start()326 TriBool at_start() { return at_start_; }
set_at_start(TriBool at_start)327 void set_at_start(TriBool at_start) { at_start_ = at_start; }
backtrack()328 Label* backtrack() { return backtrack_; }
loop_label()329 Label* loop_label() { return loop_label_; }
stop_node()330 RegExpNode* stop_node() { return stop_node_; }
characters_preloaded()331 int characters_preloaded() { return characters_preloaded_; }
bound_checked_up_to()332 int bound_checked_up_to() { return bound_checked_up_to_; }
flush_budget()333 int flush_budget() { return flush_budget_; }
quick_check_performed()334 QuickCheckDetails* quick_check_performed() { return &quick_check_performed_; }
335 bool mentions_reg(int reg);
336 // Returns true if a deferred position store exists to the specified
337 // register and stores the offset in the out-parameter. Otherwise
338 // returns false.
339 bool GetStoredPosition(int reg, int* cp_offset);
340 // These set methods and AdvanceCurrentPositionInTrace should be used only on
341 // new traces - the intention is that traces are immutable after creation.
add_action(DeferredAction* new_action)342 void add_action(DeferredAction* new_action) {
343 DCHECK(new_action->next_ == nullptr);
344 new_action->next_ = actions_;
345 actions_ = new_action;
346 }
set_backtrack(Label* backtrack)347 void set_backtrack(Label* backtrack) { backtrack_ = backtrack; }
set_stop_node(RegExpNode* node)348 void set_stop_node(RegExpNode* node) { stop_node_ = node; }
set_loop_label(Label* label)349 void set_loop_label(Label* label) { loop_label_ = label; }
set_characters_preloaded(int count)350 void set_characters_preloaded(int count) { characters_preloaded_ = count; }
set_bound_checked_up_to(int to)351 void set_bound_checked_up_to(int to) { bound_checked_up_to_ = to; }
set_flush_budget(int to)352 void set_flush_budget(int to) { flush_budget_ = to; }
set_quick_check_performed(QuickCheckDetails* d)353 void set_quick_check_performed(QuickCheckDetails* d) {
354 quick_check_performed_ = *d;
355 }
356 void InvalidateCurrentCharacter();
357 void AdvanceCurrentPositionInTrace(int by, RegExpCompiler* compiler);
358
359 private:
360 int FindAffectedRegisters(DynamicBitSet* affected_registers, Zone* zone);
361 void PerformDeferredActions(RegExpMacroAssembler* macro, int max_register,
362 const DynamicBitSet& affected_registers,
363 DynamicBitSet* registers_to_pop,
364 DynamicBitSet* registers_to_clear, Zone* zone);
365 void RestoreAffectedRegisters(RegExpMacroAssembler* macro, int max_register,
366 const DynamicBitSet& registers_to_pop,
367 const DynamicBitSet& registers_to_clear);
368 int cp_offset_;
369 DeferredAction* actions_;
370 Label* backtrack_;
371 RegExpNode* stop_node_;
372 Label* loop_label_;
373 int characters_preloaded_;
374 int bound_checked_up_to_;
375 QuickCheckDetails quick_check_performed_;
376 int flush_budget_;
377 TriBool at_start_;
378 };
379
380 class GreedyLoopState {
381 public:
382 explicit GreedyLoopState(bool not_at_start);
383
label()384 Label* label() { return &label_; }
counter_backtrack_trace()385 Trace* counter_backtrack_trace() { return &counter_backtrack_trace_; }
386
387 private:
388 Label label_;
389 Trace counter_backtrack_trace_;
390 };
391
392 struct PreloadState {
393 static const int kEatsAtLeastNotYetInitialized = -1;
394 bool preload_is_current_;
395 bool preload_has_checked_bounds_;
396 int preload_characters_;
397 int eats_at_least_;
initv8::internal::PreloadState398 void init() { eats_at_least_ = kEatsAtLeastNotYetInitialized; }
399 };
400
401 // Analysis performs assertion propagation and computes eats_at_least_ values.
402 // See the comments on AssertionPropagator and EatsAtLeastPropagator for more
403 // details.
404 RegExpError AnalyzeRegExp(Isolate* isolate, bool is_one_byte, RegExpFlags flags,
405 RegExpNode* node);
406
407 class FrequencyCollator {
408 public:
FrequencyCollator()409 FrequencyCollator() : total_samples_(0) {
410 for (int i = 0; i < RegExpMacroAssembler::kTableSize; i++) {
411 frequencies_[i] = CharacterFrequency(i);
412 }
413 }
414
CountCharacter(int character)415 void CountCharacter(int character) {
416 int index = (character & RegExpMacroAssembler::kTableMask);
417 frequencies_[index].Increment();
418 total_samples_++;
419 }
420
421 // Does not measure in percent, but rather per-128 (the table size from the
422 // regexp macro assembler).
Frequency(int in_character)423 int Frequency(int in_character) {
424 DCHECK((in_character & RegExpMacroAssembler::kTableMask) == in_character);
425 if (total_samples_ < 1) return 1; // Division by zero.
426 int freq_in_per128 =
427 (frequencies_[in_character].counter() * 128) / total_samples_;
428 return freq_in_per128;
429 }
430
431 private:
432 class CharacterFrequency {
433 public:
CharacterFrequency()434 CharacterFrequency() : counter_(0), character_(-1) {}
CharacterFrequency(int character)435 explicit CharacterFrequency(int character)
436 : counter_(0), character_(character) {}
437
Increment()438 void Increment() { counter_++; }
counter()439 int counter() { return counter_; }
character()440 int character() { return character_; }
441
442 private:
443 int counter_;
444 int character_;
445 };
446
447 private:
448 CharacterFrequency frequencies_[RegExpMacroAssembler::kTableSize];
449 int total_samples_;
450 };
451
452 class RegExpCompiler {
453 public:
454 RegExpCompiler(Isolate* isolate, Zone* zone, int capture_count,
455 RegExpFlags flags, bool is_one_byte);
456
AllocateRegister()457 int AllocateRegister() {
458 if (next_register_ >= RegExpMacroAssembler::kMaxRegister) {
459 reg_exp_too_big_ = true;
460 return next_register_;
461 }
462 return next_register_++;
463 }
464
465 // Lookarounds to match lone surrogates for unicode character class matches
466 // are never nested. We can therefore reuse registers.
UnicodeLookaroundStackRegister()467 int UnicodeLookaroundStackRegister() {
468 if (unicode_lookaround_stack_register_ == kNoRegister) {
469 unicode_lookaround_stack_register_ = AllocateRegister();
470 }
471 return unicode_lookaround_stack_register_;
472 }
473
UnicodeLookaroundPositionRegister()474 int UnicodeLookaroundPositionRegister() {
475 if (unicode_lookaround_position_register_ == kNoRegister) {
476 unicode_lookaround_position_register_ = AllocateRegister();
477 }
478 return unicode_lookaround_position_register_;
479 }
480
481 struct CompilationResult final {
CompilationResultv8::internal::RegExpCompiler::final482 explicit CompilationResult(RegExpError err) : error(err) {}
CompilationResultv8::internal::RegExpCompiler::final483 CompilationResult(Handle<Object> code, int registers)
484 : code(code), num_registers(registers) {}
485
RegExpTooBigv8::internal::RegExpCompiler::final486 static CompilationResult RegExpTooBig() {
487 return CompilationResult(RegExpError::kTooLarge);
488 }
489
Succeededv8::internal::RegExpCompiler::final490 bool Succeeded() const { return error == RegExpError::kNone; }
491
492 const RegExpError error = RegExpError::kNone;
493 Handle<Object> code;
494 int num_registers = 0;
495 };
496
497 CompilationResult Assemble(Isolate* isolate, RegExpMacroAssembler* assembler,
498 RegExpNode* start, int capture_count,
499 Handle<String> pattern);
500
501 // Preprocessing is the final step of node creation before analysis
502 // and assembly. It includes:
503 // - Wrapping the body of the regexp in capture 0.
504 // - Inserting the implicit .* before/after the regexp if necessary.
505 // - If the input is a one-byte string, filtering out nodes that can't match.
506 // - Fixing up regexp matches that start within a surrogate pair.
507 RegExpNode* PreprocessRegExp(RegExpCompileData* data, RegExpFlags flags,
508 bool is_one_byte);
509
510 // If the regexp matching starts within a surrogate pair, step back to the
511 // lead surrogate and start matching from there.
512 RegExpNode* OptionallyStepBackToLeadSurrogate(RegExpNode* on_success);
513
AddWork(RegExpNode* node)514 inline void AddWork(RegExpNode* node) {
515 if (!node->on_work_list() && !node->label()->is_bound()) {
516 node->set_on_work_list(true);
517 work_list_->push_back(node);
518 }
519 }
520
521 static const int kImplementationOffset = 0;
522 static const int kNumberOfRegistersOffset = 0;
523 static const int kCodeOffset = 1;
524
macro_assembler()525 RegExpMacroAssembler* macro_assembler() { return macro_assembler_; }
accept()526 EndNode* accept() { return accept_; }
527
528 static const int kMaxRecursion = 100;
recursion_depth()529 inline int recursion_depth() { return recursion_depth_; }
IncrementRecursionDepth()530 inline void IncrementRecursionDepth() { recursion_depth_++; }
DecrementRecursionDepth()531 inline void DecrementRecursionDepth() { recursion_depth_--; }
532
flags() const533 RegExpFlags flags() const { return flags_; }
534
SetRegExpTooBig()535 void SetRegExpTooBig() { reg_exp_too_big_ = true; }
536
one_byte()537 inline bool one_byte() { return one_byte_; }
optimize()538 inline bool optimize() { return optimize_; }
set_optimize(bool value)539 inline void set_optimize(bool value) { optimize_ = value; }
limiting_recursion()540 inline bool limiting_recursion() { return limiting_recursion_; }
set_limiting_recursion(bool value)541 inline void set_limiting_recursion(bool value) {
542 limiting_recursion_ = value;
543 }
read_backward()544 bool read_backward() { return read_backward_; }
set_read_backward(bool value)545 void set_read_backward(bool value) { read_backward_ = value; }
frequency_collator()546 FrequencyCollator* frequency_collator() { return &frequency_collator_; }
547
current_expansion_factor()548 int current_expansion_factor() { return current_expansion_factor_; }
set_current_expansion_factor(int value)549 void set_current_expansion_factor(int value) {
550 current_expansion_factor_ = value;
551 }
552
553 // The recursive nature of ToNode node generation means we may run into stack
554 // overflow issues. We introduce periodic checks to detect these, and the
555 // tick counter helps limit overhead of these checks.
556 // TODO(jgruber): This is super hacky and should be replaced by an abort
557 // mechanism or iterative node generation.
ToNodeMaybeCheckForStackOverflow()558 void ToNodeMaybeCheckForStackOverflow() {
559 if ((to_node_overflow_check_ticks_++ % 16 == 0)) {
560 ToNodeCheckForStackOverflow();
561 }
562 }
563 void ToNodeCheckForStackOverflow();
564
isolate() const565 Isolate* isolate() const { return isolate_; }
zone() const566 Zone* zone() const { return zone_; }
567
568 static const int kNoRegister = -1;
569
570 private:
571 EndNode* accept_;
572 int next_register_;
573 int unicode_lookaround_stack_register_;
574 int unicode_lookaround_position_register_;
575 ZoneVector<RegExpNode*>* work_list_;
576 int recursion_depth_;
577 const RegExpFlags flags_;
578 RegExpMacroAssembler* macro_assembler_;
579 bool one_byte_;
580 bool reg_exp_too_big_;
581 bool limiting_recursion_;
582 int to_node_overflow_check_ticks_ = 0;
583 bool optimize_;
584 bool read_backward_;
585 int current_expansion_factor_;
586 FrequencyCollator frequency_collator_;
587 Isolate* isolate_;
588 Zone* zone_;
589 };
590
591 // Categorizes character ranges into BMP, non-BMP, lead, and trail surrogates.
592 class UnicodeRangeSplitter {
593 public:
594 V8_EXPORT_PRIVATE UnicodeRangeSplitter(ZoneList<CharacterRange>* base);
595
596 static constexpr int kInitialSize = 8;
597 using CharacterRangeVector = base::SmallVector<CharacterRange, kInitialSize>;
598
bmp() const599 const CharacterRangeVector* bmp() const { return &bmp_; }
lead_surrogates() const600 const CharacterRangeVector* lead_surrogates() const {
601 return &lead_surrogates_;
602 }
trail_surrogates() const603 const CharacterRangeVector* trail_surrogates() const {
604 return &trail_surrogates_;
605 }
non_bmp() const606 const CharacterRangeVector* non_bmp() const { return &non_bmp_; }
607
608 private:
609 void AddRange(CharacterRange range);
610
611 CharacterRangeVector bmp_;
612 CharacterRangeVector lead_surrogates_;
613 CharacterRangeVector trail_surrogates_;
614 CharacterRangeVector non_bmp_;
615 };
616
617 // We need to check for the following characters: 0x39C 0x3BC 0x178.
618 // TODO(jgruber): Move to CharacterRange.
619 bool RangeContainsLatin1Equivalents(CharacterRange range);
620
621 } // namespace internal
622 } // namespace v8
623
624 #endif // V8_REGEXP_REGEXP_COMPILER_H_
625