AuroraMiddleware/v8
1
0
Fork 0
Code Issues 2 Pull Requests Releases Wiki Activity

RegExp: Add the ability to switch flags on and off within the regexp.

Browse Source 
This is a reland of https://chromium-review.googlesource.com/c/v8/v8/+/752522
which was itself a reupload of
https://chromium-review.googlesource.com/c/v8/v8/+/571746 where reviews took
place.

R=yangguo@chromium.org

Bug: 
Change-Id: Ia4dbdd6e9a362e272753ff10dc66b7f72d81ee20
Reviewed-on: https://chromium-review.googlesource.com/753596
Reviewed-by: Yang Guo <yangguo@chromium.org>
Commit-Queue: Erik Corry <erikcorry@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49129}
This commit is contained in:
Erik 2017-11-05 11:15:28 +01:00 committed by Commit Bot
parent 6ebb5641ff
commit e83ee94565
12 changed files with 917 additions and 333 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
src/flag-definitions.h
View File

@ -959,6 +959,7 @@ DEFINE_BOOL(serialization_statistics, false,
// Regexp
DEFINE_BOOL(regexp_optimization, true, "generate optimized regexp code")
DEFINE_BOOL(regexp_mode_modifiers, false, "enable inline flags in regexp.")
// Testing flags test/cctest/test-{flags,api,serialization}.cc
DEFINE_BOOL(testing_bool_flag, true, "testing_bool_flag")

315
src/regexp/jsregexp.cc
View File

@ -132,17 +132,17 @@ MaybeHandle<Object> RegExpImpl::Compile(Handle<JSRegExp> re,
bool has_been_compiled = false;
if (parse_result.simple && !(flags & JSRegExp::kIgnoreCase) &&
!(flags & JSRegExp::kSticky) &&
if (parse_result.simple && !IgnoreCase(flags) && !IsSticky(flags) &&
pattern->length() <= kPatternTooShortForBoyerMoore) {
// Parse-tree is a single atom that is equal to the pattern.
AtomCompile(re, pattern, flags, pattern);
has_been_compiled = true;
} else if (parse_result.tree->IsAtom() && !(flags & JSRegExp::kIgnoreCase) &&
!(flags & JSRegExp::kSticky) && parse_result.capture_count == 0) {
} else if (parse_result.tree->IsAtom() && !IsSticky(flags) &&
parse_result.capture_count == 0) {
RegExpAtom* atom = parse_result.tree->AsAtom();
Vector<const uc16> atom_pattern = atom->data();
if (atom_pattern.length() <= kPatternTooShortForBoyerMoore) {
if (!IgnoreCase(atom->flags()) &&
atom_pattern.length() <= kPatternTooShortForBoyerMoore) {
Handle<String> atom_string;
ASSIGN_RETURN_ON_EXCEPTION(
isolate, atom_string,
@ -622,7 +622,7 @@ RegExpImpl::GlobalCache::GlobalCache(Handle<JSRegExp> regexp,
}
}
DCHECK_NE(0, regexp->GetFlags() & JSRegExp::kGlobal);
DCHECK(IsGlobal(regexp->GetFlags()));
if (!interpreted) {
register_array_size_ =
Max(registers_per_match_, Isolate::kJSRegexpStaticOffsetsVectorSize);
@ -653,8 +653,7 @@ RegExpImpl::GlobalCache::GlobalCache(Handle<JSRegExp> regexp,
}
int RegExpImpl::GlobalCache::AdvanceZeroLength(int last_index) {
if ((regexp_->GetFlags() & JSRegExp::kUnicode) != 0 &&
last_index + 1 < subject_->length() &&
if (IsUnicode(regexp_->GetFlags()) && last_index + 1 < subject_->length() &&
unibrow::Utf16::IsLeadSurrogate(subject_->Get(last_index)) &&
unibrow::Utf16::IsTrailSurrogate(subject_->Get(last_index + 1))) {
// Advance over the surrogate pair.
@ -916,7 +915,7 @@ class FrequencyCollator {
class RegExpCompiler {
public:
RegExpCompiler(Isolate* isolate, Zone* zone, int capture_count,
JSRegExp::Flags flags, bool is_one_byte);
bool is_one_byte);
int AllocateRegister() {
if (next_register_ >= RegExpMacroAssembler::kMaxRegister) {
@ -968,13 +967,6 @@ class RegExpCompiler {
void SetRegExpTooBig() { reg_exp_too_big_ = true; }
inline bool ignore_case() { return (flags_ & JSRegExp::kIgnoreCase) != 0; }
inline bool unicode() { return (flags_ & JSRegExp::kUnicode) != 0; }
// Both unicode and ignore_case flags are set. We need to use ICU to find
// the closure over case equivalents.
inline bool needs_unicode_case_equivalents() {
return unicode() && ignore_case();
}
inline bool one_byte() { return one_byte_; }
inline bool optimize() { return optimize_; }
inline void set_optimize(bool value) { optimize_ = value; }
@ -1004,7 +996,6 @@ class RegExpCompiler {
std::vector<RegExpNode*>* work_list_;
int recursion_depth_;
RegExpMacroAssembler* macro_assembler_;
JSRegExp::Flags flags_;
bool one_byte_;
bool reg_exp_too_big_;
bool limiting_recursion_;
@ -1036,13 +1027,12 @@ static RegExpEngine::CompilationResult IrregexpRegExpTooBig(Isolate* isolate) {
// Attempts to compile the regexp using an Irregexp code generator. Returns
// a fixed array or a null handle depending on whether it succeeded.
RegExpCompiler::RegExpCompiler(Isolate* isolate, Zone* zone, int capture_count,
JSRegExp::Flags flags, bool one_byte)
bool one_byte)
: next_register_(2 * (capture_count + 1)),
unicode_lookaround_stack_register_(kNoRegister),
unicode_lookaround_position_register_(kNoRegister),
work_list_(nullptr),
recursion_depth_(0),
flags_(flags),
one_byte_(one_byte),
reg_exp_too_big_(false),
limiting_recursion_(false),
@ -2503,7 +2493,7 @@ void TextNode::GetQuickCheckDetails(QuickCheckDetails* details,
QuickCheckDetails::Position* pos =
details->positions(characters_filled_in);
uc16 c = quarks[i];
if (compiler->ignore_case()) {
if (elm.atom()->ignore_case()) {
unibrow::uchar chars[unibrow::Ecma262UnCanonicalize::kMaxWidth];
int length = GetCaseIndependentLetters(isolate, c,
compiler->one_byte(), chars);
@ -2711,18 +2701,16 @@ class VisitMarker {
NodeInfo* info_;
};
RegExpNode* SeqRegExpNode::FilterOneByte(int depth, bool ignore_case) {
RegExpNode* SeqRegExpNode::FilterOneByte(int depth) {
if (info()->replacement_calculated) return replacement();
if (depth < 0) return this;
DCHECK(!info()->visited);
VisitMarker marker(info());
return FilterSuccessor(depth - 1, ignore_case);
return FilterSuccessor(depth - 1);
}
RegExpNode* SeqRegExpNode::FilterSuccessor(int depth, bool ignore_case) {
RegExpNode* next = on_success_->FilterOneByte(depth - 1, ignore_case);
RegExpNode* SeqRegExpNode::FilterSuccessor(int depth) {
RegExpNode* next = on_success_->FilterOneByte(depth - 1);
if (next == nullptr) return set_replacement(nullptr);
on_success_ = next;
return set_replacement(this);
@ -2745,8 +2733,7 @@ static bool RangesContainLatin1Equivalents(ZoneList<CharacterRange>* ranges) {
return false;
}
RegExpNode* TextNode::FilterOneByte(int depth, bool ignore_case) {
RegExpNode* TextNode::FilterOneByte(int depth) {
if (info()->replacement_calculated) return replacement();
if (depth < 0) return this;
DCHECK(!info()->visited);
@ -2759,7 +2746,7 @@ RegExpNode* TextNode::FilterOneByte(int depth, bool ignore_case) {
for (int j = 0; j < quarks.length(); j++) {
uint16_t c = quarks[j];
if (c <= String::kMaxOneByteCharCode) continue;
if (!ignore_case) return set_replacement(nullptr);
if (!IgnoreCase(elm.atom()->flags())) return set_replacement(nullptr);
// Here, we need to check for characters whose upper and lower cases
// are outside the Latin-1 range.
uint16_t converted = unibrow::Latin1::ConvertNonLatin1ToLatin1(c);
@ -2781,42 +2768,41 @@ RegExpNode* TextNode::FilterOneByte(int depth, bool ignore_case) {
ranges->at(0).from() == 0 &&
ranges->at(0).to() >= String::kMaxOneByteCharCode) {
// This will be handled in a later filter.
if (ignore_case && RangesContainLatin1Equivalents(ranges)) continue;
if (IgnoreCase(cc->flags()) && RangesContainLatin1Equivalents(ranges))
continue;
return set_replacement(nullptr);
}
} else {
if (range_count == 0 ||
ranges->at(0).from() > String::kMaxOneByteCharCode) {
// This will be handled in a later filter.
if (ignore_case && RangesContainLatin1Equivalents(ranges)) continue;
if (IgnoreCase(cc->flags()) && RangesContainLatin1Equivalents(ranges))
continue;
return set_replacement(nullptr);
}
}
}
}
return FilterSuccessor(depth - 1, ignore_case);
return FilterSuccessor(depth - 1);
}
RegExpNode* LoopChoiceNode::FilterOneByte(int depth, bool ignore_case) {
RegExpNode* LoopChoiceNode::FilterOneByte(int depth) {
if (info()->replacement_calculated) return replacement();
if (depth < 0) return this;
if (info()->visited) return this;
{
VisitMarker marker(info());
RegExpNode* continue_replacement =
continue_node_->FilterOneByte(depth - 1, ignore_case);
RegExpNode* continue_replacement = continue_node_->FilterOneByte(depth - 1);
// If we can't continue after the loop then there is no sense in doing the
// loop.
if (continue_replacement == nullptr) return set_replacement(nullptr);
}
return ChoiceNode::FilterOneByte(depth - 1, ignore_case);
return ChoiceNode::FilterOneByte(depth - 1);
}
RegExpNode* ChoiceNode::FilterOneByte(int depth, bool ignore_case) {
RegExpNode* ChoiceNode::FilterOneByte(int depth) {
if (info()->replacement_calculated) return replacement();
if (depth < 0) return this;
if (info()->visited) return this;
@ -2836,8 +2822,7 @@ RegExpNode* ChoiceNode::FilterOneByte(int depth, bool ignore_case) {
RegExpNode* survivor = nullptr;
for (int i = 0; i < choice_count; i++) {
GuardedAlternative alternative = alternatives_->at(i);
RegExpNode* replacement =
alternative.node()->FilterOneByte(depth - 1, ignore_case);
RegExpNode* replacement = alternative.node()->FilterOneByte(depth - 1);
DCHECK(replacement != this); // No missing EMPTY_MATCH_CHECK.
if (replacement != nullptr) {
alternatives_->at(i).set_node(replacement);
@ -2857,7 +2842,7 @@ RegExpNode* ChoiceNode::FilterOneByte(int depth, bool ignore_case) {
new(zone()) ZoneList<GuardedAlternative>(surviving, zone());
for (int i = 0; i < choice_count; i++) {
RegExpNode* replacement =
alternatives_->at(i).node()->FilterOneByte(depth - 1, ignore_case);
alternatives_->at(i).node()->FilterOneByte(depth - 1);
if (replacement != nullptr) {
alternatives_->at(i).set_node(replacement);
new_alternatives->Add(alternatives_->at(i), zone());
@ -2867,9 +2852,7 @@ RegExpNode* ChoiceNode::FilterOneByte(int depth, bool ignore_case) {
return this;
}
RegExpNode* NegativeLookaroundChoiceNode::FilterOneByte(int depth,
bool ignore_case) {
RegExpNode* NegativeLookaroundChoiceNode::FilterOneByte(int depth) {
if (info()->replacement_calculated) return replacement();
if (depth < 0) return this;
if (info()->visited) return this;
@ -2877,12 +2860,12 @@ RegExpNode* NegativeLookaroundChoiceNode::FilterOneByte(int depth,
// Alternative 0 is the negative lookahead, alternative 1 is what comes
// afterwards.
RegExpNode* node = alternatives_->at(1).node();
RegExpNode* replacement = node->FilterOneByte(depth - 1, ignore_case);
RegExpNode* replacement = node->FilterOneByte(depth - 1);
if (replacement == nullptr) return set_replacement(nullptr);
alternatives_->at(1).set_node(replacement);
RegExpNode* neg_node = alternatives_->at(0).node();
RegExpNode* neg_replacement = neg_node->FilterOneByte(depth - 1, ignore_case);
RegExpNode* neg_replacement = neg_node->FilterOneByte(depth - 1);
// If the negative lookahead is always going to fail then
// we don't need to check it.
if (neg_replacement == nullptr) return set_replacement(replacement);
@ -3199,6 +3182,7 @@ void TextNode::TextEmitPass(RegExpCompiler* compiler,
TextElement elm = elements()->at(i);
int cp_offset = trace->cp_offset() + elm.cp_offset() + backward_offset;
if (elm.text_type() == TextElement::ATOM) {
if (SkipPass(pass, elm.atom()->ignore_case())) continue;
Vector<const uc16> quarks = elm.atom()->data();
for (int j = preloaded ? 0 : quarks.length() - 1; j >= 0; j--) {
if (first_element_checked && i == 0 && j == 0) continue;
@ -3254,9 +3238,7 @@ int TextNode::Length() {
return elm.cp_offset() + elm.length();
}
bool TextNode::SkipPass(int int_pass, bool ignore_case) {
TextEmitPassType pass = static_cast<TextEmitPassType>(int_pass);
bool TextNode::SkipPass(TextEmitPassType pass, bool ignore_case) {
if (ignore_case) {
return pass == SIMPLE_CHARACTER_MATCH;
} else {
@ -3264,32 +3246,33 @@ bool TextNode::SkipPass(int int_pass, bool ignore_case) {
}
}
TextNode* TextNode::CreateForCharacterRanges(Zone* zone,
ZoneList<CharacterRange>* ranges,
bool read_backward,
RegExpNode* on_success) {
RegExpNode* on_success,
JSRegExp::Flags flags) {
DCHECK_NOT_NULL(ranges);
ZoneList<TextElement>* elms = new (zone) ZoneList<TextElement>(1, zone);
elms->Add(TextElement::CharClass(new (zone) RegExpCharacterClass(ranges)),
zone);
elms->Add(
TextElement::CharClass(new (zone) RegExpCharacterClass(ranges, flags)),
zone);
return new (zone) TextNode(elms, read_backward, on_success);
}
TextNode* TextNode::CreateForSurrogatePair(Zone* zone, CharacterRange lead,
CharacterRange trail,
bool read_backward,
RegExpNode* on_success) {
RegExpNode* on_success,
JSRegExp::Flags flags) {
ZoneList<CharacterRange>* lead_ranges = CharacterRange::List(zone, lead);
ZoneList<CharacterRange>* trail_ranges = CharacterRange::List(zone, trail);
ZoneList<TextElement>* elms = new (zone) ZoneList<TextElement>(2, zone);
elms->Add(
TextElement::CharClass(new (zone) RegExpCharacterClass(lead_ranges)),
zone);
elms->Add(
TextElement::CharClass(new (zone) RegExpCharacterClass(trail_ranges)),
zone);
elms->Add(TextElement::CharClass(
new (zone) RegExpCharacterClass(lead_ranges, flags)),
zone);
elms->Add(TextElement::CharClass(
new (zone) RegExpCharacterClass(trail_ranges, flags)),
zone);
return new (zone) TextNode(elms, read_backward, on_success);
}
@ -3323,27 +3306,15 @@ void TextNode::Emit(RegExpCompiler* compiler, Trace* trace) {
// check that now.
if (trace->characters_preloaded() == 1) {
for (int pass = kFirstRealPass; pass <= kLastPass; pass++) {
if (!SkipPass(pass, compiler->ignore_case())) {
TextEmitPass(compiler,
static_cast<TextEmitPassType>(pass),
true,
trace,
false,
&bound_checked_to);
}
TextEmitPass(compiler, static_cast<TextEmitPassType>(pass), true, trace,
false, &bound_checked_to);
}
first_elt_done = true;
}
for (int pass = kFirstRealPass; pass <= kLastPass; pass++) {
if (!SkipPass(pass, compiler->ignore_case())) {
TextEmitPass(compiler,
static_cast<TextEmitPassType>(pass),
false,
trace,
first_elt_done,
&bound_checked_to);
}
TextEmitPass(compiler, static_cast<TextEmitPassType>(pass), false, trace,
first_elt_done, &bound_checked_to);
}
Trace successor_trace(*trace);
@ -3386,11 +3357,15 @@ void TextNode::MakeCaseIndependent(Isolate* isolate, bool is_one_byte) {
TextElement elm = elements()->at(i);
if (elm.text_type() == TextElement::CHAR_CLASS) {
RegExpCharacterClass* cc = elm.char_class();
// None of the standard character classes is different in the case
// independent case and it slows us down if we don't know that.
if (cc->is_standard(zone())) continue;
ZoneList<CharacterRange>* ranges = cc->ranges(zone());
CharacterRange::AddCaseEquivalents(isolate, zone(), ranges, is_one_byte);
if (IgnoreCase(cc->flags()) &&
!NeedsUnicodeCaseEquivalents(cc->flags())) {
// None of the standard character classes is different in the case
// independent case and it slows us down if we don't know that.
if (cc->is_standard(zone())) continue;
ZoneList<CharacterRange>* ranges = cc->ranges(zone());
CharacterRange::AddCaseEquivalents(isolate, zone(), ranges,
is_one_byte);
}
}
}
}
@ -4353,9 +4328,9 @@ void BackReferenceNode::Emit(RegExpCompiler* compiler, Trace* trace) {
RecursionCheck rc(compiler);
DCHECK_EQ(start_reg_ + 1, end_reg_);
if (compiler->ignore_case()) {
if (IgnoreCase(flags_)) {
assembler->CheckNotBackReferenceIgnoreCase(
start_reg_, read_backward(), compiler->unicode(), trace->backtrack());
start_reg_, read_backward(), IsUnicode(flags_), trace->backtrack());
} else {
assembler->CheckNotBackReference(start_reg_, read_backward(),
trace->backtrack());
@ -4364,7 +4339,7 @@ void BackReferenceNode::Emit(RegExpCompiler* compiler, Trace* trace) {
if (read_backward()) trace->set_at_start(Trace::UNKNOWN);
// Check that the back reference does not end inside a surrogate pair.
if (compiler->unicode() && !compiler->one_byte()) {
if (IsUnicode(flags_) && !compiler->one_byte()) {
assembler->CheckNotInSurrogatePair(trace->cp_offset(), trace->backtrack());
}
on_success()->Emit(compiler, trace);
@ -4887,24 +4862,24 @@ void UnicodeRangeSplitter::Call(uc32 from, DispatchTable::Entry entry) {
(*target)->Add(CharacterRange::Range(entry.from(), entry.to()), zone_);
}
void AddBmpCharacters(RegExpCompiler* compiler, ChoiceNode* result,
RegExpNode* on_success, UnicodeRangeSplitter* splitter) {
ZoneList<CharacterRange>* bmp = splitter->bmp();
if (bmp == nullptr) return;
JSRegExp::Flags default_flags = JSRegExp::Flags();
result->AddAlternative(GuardedAlternative(TextNode::CreateForCharacterRanges(
compiler->zone(), bmp, compiler->read_backward(), on_success)));
compiler->zone(), bmp, compiler->read_backward(), on_success,
default_flags)));
}
void AddNonBmpSurrogatePairs(RegExpCompiler* compiler, ChoiceNode* result,
RegExpNode* on_success,
UnicodeRangeSplitter* splitter) {
ZoneList<CharacterRange>* non_bmp = splitter->non_bmp();
if (non_bmp == nullptr) return;
DCHECK(compiler->unicode());
DCHECK(!compiler->one_byte());
Zone* zone = compiler->zone();
JSRegExp::Flags default_flags = JSRegExp::Flags();
CharacterRange::Canonicalize(non_bmp);
for (int i = 0; i < non_bmp->length(); i++) {
// Match surrogate pair.
@ -4924,7 +4899,7 @@ void AddNonBmpSurrogatePairs(RegExpCompiler* compiler, ChoiceNode* result,
GuardedAlternative(TextNode::CreateForSurrogatePair(
zone, CharacterRange::Singleton(from_l),
CharacterRange::Range(from_t, to_t), compiler->read_backward(),
on_success)));
on_success, default_flags)));
} else {
if (from_t != kTrailSurrogateStart) {
// Add [from_l][from_t-\udfff]
@ -4932,7 +4907,7 @@ void AddNonBmpSurrogatePairs(RegExpCompiler* compiler, ChoiceNode* result,
GuardedAlternative(TextNode::CreateForSurrogatePair(
zone, CharacterRange::Singleton(from_l),
CharacterRange::Range(from_t, kTrailSurrogateEnd),
compiler->read_backward(), on_success)));
compiler->read_backward(), on_success, default_flags)));
from_l++;
}
if (to_t != kTrailSurrogateEnd) {
@ -4941,7 +4916,7 @@ void AddNonBmpSurrogatePairs(RegExpCompiler* compiler, ChoiceNode* result,
GuardedAlternative(TextNode::CreateForSurrogatePair(
zone, CharacterRange::Singleton(to_l),
CharacterRange::Range(kTrailSurrogateStart, to_t),
compiler->read_backward(), on_success)));
compiler->read_backward(), on_success, default_flags)));
to_l--;
}
if (from_l <= to_l) {
@ -4950,49 +4925,47 @@ void AddNonBmpSurrogatePairs(RegExpCompiler* compiler, ChoiceNode* result,
GuardedAlternative(TextNode::CreateForSurrogatePair(
zone, CharacterRange::Range(from_l, to_l),
CharacterRange::Range(kTrailSurrogateStart, kTrailSurrogateEnd),
compiler->read_backward(), on_success)));
compiler->read_backward(), on_success, default_flags)));
}
}
}
}
RegExpNode* NegativeLookaroundAgainstReadDirectionAndMatch(
RegExpCompiler* compiler, ZoneList<CharacterRange>* lookbehind,
ZoneList<CharacterRange>* match, RegExpNode* on_success,
bool read_backward) {
ZoneList<CharacterRange>* match, RegExpNode* on_success, bool read_backward,
JSRegExp::Flags flags) {
Zone* zone = compiler->zone();
RegExpNode* match_node = TextNode::CreateForCharacterRanges(
zone, match, read_backward, on_success);
zone, match, read_backward, on_success, flags);
int stack_register = compiler->UnicodeLookaroundStackRegister();
int position_register = compiler->UnicodeLookaroundPositionRegister();
RegExpLookaround::Builder lookaround(false, match_node, stack_register,
position_register);
RegExpNode* negative_match = TextNode::CreateForCharacterRanges(
zone, lookbehind, !read_backward, lookaround.on_match_success());
zone, lookbehind, !read_backward, lookaround.on_match_success(), flags);
return lookaround.ForMatch(negative_match);
}
RegExpNode* MatchAndNegativeLookaroundInReadDirection(
RegExpCompiler* compiler, ZoneList<CharacterRange>* match,
ZoneList<CharacterRange>* lookahead, RegExpNode* on_success,
bool read_backward) {
bool read_backward, JSRegExp::Flags flags) {
Zone* zone = compiler->zone();
int stack_register = compiler->UnicodeLookaroundStackRegister();
int position_register = compiler->UnicodeLookaroundPositionRegister();
RegExpLookaround::Builder lookaround(false, on_success, stack_register,
position_register);
RegExpNode* negative_match = TextNode::CreateForCharacterRanges(
zone, lookahead, read_backward, lookaround.on_match_success());
zone, lookahead, read_backward, lookaround.on_match_success(), flags);
return TextNode::CreateForCharacterRanges(
zone, match, read_backward, lookaround.ForMatch(negative_match));
zone, match, read_backward, lookaround.ForMatch(negative_match), flags);
}
void AddLoneLeadSurrogates(RegExpCompiler* compiler, ChoiceNode* result,
RegExpNode* on_success,
UnicodeRangeSplitter* splitter) {
JSRegExp::Flags default_flags = JSRegExp::Flags();
ZoneList<CharacterRange>* lead_surrogates = splitter->lead_surrogates();
if (lead_surrogates == nullptr) return;
Zone* zone = compiler->zone();
@ -5005,20 +4978,22 @@ void AddLoneLeadSurrogates(RegExpCompiler* compiler, ChoiceNode* result,
// Reading backward. Assert that reading forward, there is no trail
// surrogate, and then backward match the lead surrogate.
match = NegativeLookaroundAgainstReadDirectionAndMatch(
compiler, trail_surrogates, lead_surrogates, on_success, true);
compiler, trail_surrogates, lead_surrogates, on_success, true,
default_flags);
} else {
// Reading forward. Forward match the lead surrogate and assert that
// no trail surrogate follows.
match = MatchAndNegativeLookaroundInReadDirection(
compiler, lead_surrogates, trail_surrogates, on_success, false);
compiler, lead_surrogates, trail_surrogates, on_success, false,
default_flags);
}
result->AddAlternative(GuardedAlternative(match));
}
void AddLoneTrailSurrogates(RegExpCompiler* compiler, ChoiceNode* result,
RegExpNode* on_success,
UnicodeRangeSplitter* splitter) {
JSRegExp::Flags default_flags = JSRegExp::Flags();
ZoneList<CharacterRange>* trail_surrogates = splitter->trail_surrogates();
if (trail_surrogates == nullptr) return;
Zone* zone = compiler->zone();
@ -5031,12 +5006,14 @@ void AddLoneTrailSurrogates(RegExpCompiler* compiler, ChoiceNode* result,
// Reading backward. Backward match the trail surrogate and assert that no
// lead surrogate precedes it.
match = MatchAndNegativeLookaroundInReadDirection(
compiler, trail_surrogates, lead_surrogates, on_success, true);
compiler, trail_surrogates, lead_surrogates, on_success, true,
default_flags);
} else {
// Reading forward. Assert that reading backward, there is no lead
// surrogate, and then forward match the trail surrogate.
match = NegativeLookaroundAgainstReadDirectionAndMatch(
compiler, lead_surrogates, trail_surrogates, on_success, false);
compiler, lead_surrogates, trail_surrogates, on_success, false,
default_flags);
}
result->AddAlternative(GuardedAlternative(match));
}
@ -5052,7 +5029,9 @@ RegExpNode* UnanchoredAdvance(RegExpCompiler* compiler,
// the associated trail surrogate.
ZoneList<CharacterRange>* range = CharacterRange::List(
zone, CharacterRange::Range(0, String::kMaxUtf16CodeUnit));
return TextNode::CreateForCharacterRanges(zone, range, false, on_success);
JSRegExp::Flags default_flags = JSRegExp::Flags();
return TextNode::CreateForCharacterRanges(zone, range, false, on_success,
default_flags);
}
void AddUnicodeCaseEquivalents(ZoneList<CharacterRange>* ranges, Zone* zone) {
@ -5093,10 +5072,10 @@ RegExpNode* RegExpCharacterClass::ToNode(RegExpCompiler* compiler,
set_.Canonicalize();
Zone* zone = compiler->zone();
ZoneList<CharacterRange>* ranges = this->ranges(zone);
if (compiler->needs_unicode_case_equivalents()) {
if (NeedsUnicodeCaseEquivalents(flags_)) {
AddUnicodeCaseEquivalents(ranges, zone);
}
if (compiler->unicode() && !compiler->one_byte() &&
if (IsUnicode(flags_) && !compiler->one_byte() &&
!contains_split_surrogate()) {
if (is_negated()) {
ZoneList<CharacterRange>* negated =
@ -5105,9 +5084,10 @@ RegExpNode* RegExpCharacterClass::ToNode(RegExpCompiler* compiler,
ranges = negated;
}
if (ranges->length() == 0) {
JSRegExp::Flags default_flags = JSRegExp::Flags();
ranges->Add(CharacterRange::Everything(), zone);
RegExpCharacterClass* fail =
new (zone) RegExpCharacterClass(ranges, NEGATED);
new (zone) RegExpCharacterClass(ranges, default_flags, NEGATED);
return new (zone) TextNode(fail, compiler->read_backward(), on_success);
}
if (standard_type() == '*') {
@ -5182,10 +5162,12 @@ bool RegExpDisjunction::SortConsecutiveAtoms(RegExpCompiler* compiler) {
// i is length or it is the index of an atom.
if (i == length) break;
int first_atom = i;
JSRegExp::Flags flags = alternatives->at(i)->AsAtom()->flags();
i++;
while (i < length) {
RegExpTree* alternative = alternatives->at(i);
if (!alternative->IsAtom()) break;
if (alternative->AsAtom()->flags() != flags) break;
i++;
}
// Sort atoms to get ones with common prefixes together.
@ -5197,7 +5179,7 @@ bool RegExpDisjunction::SortConsecutiveAtoms(RegExpCompiler* compiler) {
DCHECK_LT(first_atom, alternatives->length());
DCHECK_LE(i, alternatives->length());
DCHECK_LE(first_atom, i);
if (compiler->ignore_case()) {
if (IgnoreCase(flags)) {
unibrow::Mapping<unibrow::Ecma262Canonicalize>* canonicalize =
compiler->isolate()->regexp_macro_assembler_canonicalize();
auto compare_closure =
@ -5229,7 +5211,8 @@ void RegExpDisjunction::RationalizeConsecutiveAtoms(RegExpCompiler* compiler) {
i++;
continue;
}
RegExpAtom* atom = alternative->AsAtom();
RegExpAtom* const atom = alternative->AsAtom();
JSRegExp::Flags flags = atom->flags();
unibrow::uchar common_prefix = atom->data().at(0);
int first_with_prefix = i;
int prefix_length = atom->length();
@ -5237,10 +5220,11 @@ void RegExpDisjunction::RationalizeConsecutiveAtoms(RegExpCompiler* compiler) {
while (i < length) {
alternative = alternatives->at(i);
if (!alternative->IsAtom()) break;
atom = alternative->AsAtom();
RegExpAtom* const atom = alternative->AsAtom();
if (atom->flags() != flags) break;
unibrow::uchar new_prefix = atom->data().at(0);
if (new_prefix != common_prefix) {
if (!compiler->ignore_case()) break;
if (!IgnoreCase(flags)) break;
unibrow::Mapping<unibrow::Ecma262Canonicalize>* canonicalize =
compiler->isolate()->regexp_macro_assembler_canonicalize();
new_prefix = Canonical(canonicalize, new_prefix);
@ -5257,7 +5241,7 @@ void RegExpDisjunction::RationalizeConsecutiveAtoms(RegExpCompiler* compiler) {
// common prefix if the terms were similar or presorted in the input.
// Find out how long the common prefix is.
int run_length = i - first_with_prefix;
atom = alternatives->at(first_with_prefix)->AsAtom();
RegExpAtom* const atom = alternatives->at(first_with_prefix)->AsAtom();
for (int j = 1; j < run_length && prefix_length > 1; j++) {
RegExpAtom* old_atom =
alternatives->at(j + first_with_prefix)->AsAtom();
@ -5268,8 +5252,8 @@ void RegExpDisjunction::RationalizeConsecutiveAtoms(RegExpCompiler* compiler) {
}
}
}
RegExpAtom* prefix =
new (zone) RegExpAtom(atom->data().SubVector(0, prefix_length));
RegExpAtom* prefix = new (zone)
RegExpAtom(atom->data().SubVector(0, prefix_length), flags);
ZoneList<RegExpTree*>* pair = new (zone) ZoneList<RegExpTree*>(2, zone);
pair->Add(prefix, zone);
ZoneList<RegExpTree*>* suffixes =
@ -5282,7 +5266,8 @@ void RegExpDisjunction::RationalizeConsecutiveAtoms(RegExpCompiler* compiler) {
suffixes->Add(new (zone) RegExpEmpty(), zone);
} else {
RegExpTree* suffix = new (zone) RegExpAtom(
old_atom->data().SubVector(prefix_length, old_atom->length()));
old_atom->data().SubVector(prefix_length, old_atom->length()),
flags);
suffixes->Add(suffix, zone);
}
}
@ -5305,7 +5290,6 @@ void RegExpDisjunction::FixSingleCharacterDisjunctions(
Zone* zone = compiler->zone();
ZoneList<RegExpTree*>* alternatives = this->alternatives();
int length = alternatives->length();
const bool unicode = compiler->unicode();
int write_posn = 0;
int i = 0;
@ -5316,24 +5300,28 @@ void RegExpDisjunction::FixSingleCharacterDisjunctions(
i++;
continue;
}
RegExpAtom* atom = alternative->AsAtom();
RegExpAtom* const atom = alternative->AsAtom();
if (atom->length() != 1) {
alternatives->at(write_posn++) = alternatives->at(i);
i++;
continue;
}
DCHECK_IMPLIES(unicode,
JSRegExp::Flags flags = atom->flags();
DCHECK_IMPLIES(IsUnicode(flags),
!unibrow::Utf16::IsLeadSurrogate(atom->data().at(0)));
bool contains_trail_surrogate =
unibrow::Utf16::IsTrailSurrogate(atom->data().at(0));
int first_in_run = i;
i++;
// Find a run of single-character atom alternatives that have identical
// flags (case independence and unicode-ness).
while (i < length) {
alternative = alternatives->at(i);
if (!alternative->IsAtom()) break;
atom = alternative->AsAtom();
RegExpAtom* const atom = alternative->AsAtom();
if (atom->length() != 1) break;
DCHECK_IMPLIES(unicode,
if (atom->flags() != flags) break;
DCHECK_IMPLIES(IsUnicode(flags),
!unibrow::Utf16::IsLeadSurrogate(atom->data().at(0)));
contains_trail_surrogate |=
unibrow::Utf16::IsTrailSurrogate(atom->data().at(0));
@ -5349,12 +5337,12 @@ void RegExpDisjunction::FixSingleCharacterDisjunctions(
DCHECK_EQ(old_atom->length(), 1);
ranges->Add(CharacterRange::Singleton(old_atom->data().at(0)), zone);
}
RegExpCharacterClass::Flags flags;
if (unicode && contains_trail_surrogate) {
flags = RegExpCharacterClass::CONTAINS_SPLIT_SURROGATE;
RegExpCharacterClass::CharacterClassFlags character_class_flags;
if (IsUnicode(flags) && contains_trail_surrogate) {
character_class_flags = RegExpCharacterClass::CONTAINS_SPLIT_SURROGATE;
}
alternatives->at(write_posn++) =
new (zone) RegExpCharacterClass(ranges, flags);
new (zone) RegExpCharacterClass(ranges, flags, character_class_flags);
} else {
// Just copy any trivial alternatives.
for (int j = first_in_run; j < i; j++) {
@ -5586,8 +5574,9 @@ namespace {
// \B to (?<=\w)(?=\w)|(?<=\W)(?=\W)
RegExpNode* BoundaryAssertionAsLookaround(RegExpCompiler* compiler,
RegExpNode* on_success,
RegExpAssertion::AssertionType type) {
DCHECK(compiler->needs_unicode_case_equivalents());
RegExpAssertion::AssertionType type,
JSRegExp::Flags flags) {
DCHECK(NeedsUnicodeCaseEquivalents(flags));
Zone* zone = compiler->zone();
ZoneList<CharacterRange>* word_range =
new (zone) ZoneList<CharacterRange>(2, zone);
@ -5605,13 +5594,13 @@ RegExpNode* BoundaryAssertionAsLookaround(RegExpCompiler* compiler,
RegExpLookaround::Builder lookbehind(lookbehind_for_word, on_success,
stack_register, position_register);
RegExpNode* backward = TextNode::CreateForCharacterRanges(
zone, word_range, true, lookbehind.on_match_success());
zone, word_range, true, lookbehind.on_match_success(), flags);
// Look to the right.
RegExpLookaround::Builder lookahead(lookahead_for_word,
lookbehind.ForMatch(backward),
stack_register, position_register);
RegExpNode* forward = TextNode::CreateForCharacterRanges(
zone, word_range, false, lookahead.on_match_success());
zone, word_range, false, lookahead.on_match_success(), flags);
result->AddAlternative(GuardedAlternative(lookahead.ForMatch(forward)));
}
return result;
@ -5629,13 +5618,14 @@ RegExpNode* RegExpAssertion::ToNode(RegExpCompiler* compiler,
case START_OF_INPUT:
return AssertionNode::AtStart(on_success);
case BOUNDARY:
return compiler->needs_unicode_case_equivalents()
? BoundaryAssertionAsLookaround(compiler, on_success, BOUNDARY)
return NeedsUnicodeCaseEquivalents(flags_)
? BoundaryAssertionAsLookaround(compiler, on_success, BOUNDARY,
flags_)
: AssertionNode::AtBoundary(on_success);
case NON_BOUNDARY:
return compiler->needs_unicode_case_equivalents()
return NeedsUnicodeCaseEquivalents(flags_)
? BoundaryAssertionAsLookaround(compiler, on_success,
NON_BOUNDARY)
NON_BOUNDARY, flags_)
: AssertionNode::AtNonBoundary(on_success);
case END_OF_INPUT:
return AssertionNode::AtEnd(on_success);
@ -5651,7 +5641,9 @@ RegExpNode* RegExpAssertion::ToNode(RegExpCompiler* compiler,
ZoneList<CharacterRange>* newline_ranges =
new(zone) ZoneList<CharacterRange>(3, zone);
CharacterRange::AddClassEscape('n', newline_ranges, false, zone);
RegExpCharacterClass* newline_atom = new (zone) RegExpCharacterClass('n');
JSRegExp::Flags default_flags = JSRegExp::Flags();
RegExpCharacterClass* newline_atom =
new (zone) RegExpCharacterClass('n', default_flags);
TextNode* newline_matcher = new (zone) TextNode(
newline_atom, false, ActionNode::PositiveSubmatchSuccess(
stack_pointer_register, position_register,
@ -5681,7 +5673,7 @@ RegExpNode* RegExpBackReference::ToNode(RegExpCompiler* compiler,
RegExpNode* on_success) {
return new (compiler->zone())
BackReferenceNode(RegExpCapture::StartRegister(index()),
RegExpCapture::EndRegister(index()),
RegExpCapture::EndRegister(index()), flags_,
compiler->read_backward(), on_success);
}
@ -6337,9 +6329,7 @@ void TextNode::CalculateOffsets() {
void Analysis::VisitText(TextNode* that) {
if (ignore_case()) {
that->MakeCaseIndependent(isolate(), is_one_byte_);
}
that->MakeCaseIndependent(isolate(), is_one_byte_);
EnsureAnalyzed(that->on_success());
if (!has_failed()) {
that->CalculateOffsets();
@ -6450,7 +6440,7 @@ void TextNode::FillInBMInfo(Isolate* isolate, int initial_offset, int budget,
return;
}
uc16 character = atom->data()[j];
if (bm->compiler()->ignore_case()) {
if (IgnoreCase(atom->flags())) {
unibrow::uchar chars[unibrow::Ecma262UnCanonicalize::kMaxWidth];
int length = GetCaseIndependentLetters(
isolate, character, bm->max_char() == String::kMaxOneByteCharCode,
@ -6602,9 +6592,9 @@ void DispatchTableConstructor::VisitAction(ActionNode* that) {
target->Accept(this);
}
RegExpNode* OptionallyStepBackToLeadSurrogate(RegExpCompiler* compiler,
RegExpNode* on_success) {
RegExpNode* on_success,
JSRegExp::Flags flags) {
// If the regexp matching starts within a surrogate pair, step back
// to the lead surrogate and start matching from there.
DCHECK(!compiler->read_backward());
@ -6619,11 +6609,11 @@ RegExpNode* OptionallyStepBackToLeadSurrogate(RegExpCompiler* compiler,
int stack_register = compiler->UnicodeLookaroundStackRegister();
int position_register = compiler->UnicodeLookaroundPositionRegister();
RegExpNode* step_back = TextNode::CreateForCharacterRanges(
zone, lead_surrogates, true, on_success);
zone, lead_surrogates, true, on_success, flags);
RegExpLookaround::Builder builder(true, step_back, stack_register,
position_register);
RegExpNode* match_trail = TextNode::CreateForCharacterRanges(
zone, trail_surrogates, false, builder.on_match_success());
zone, trail_surrogates, false, builder.on_match_success(), flags);
optional_step_back->AddAlternative(
GuardedAlternative(builder.ForMatch(match_trail)));
@ -6640,12 +6630,10 @@ RegExpEngine::CompilationResult RegExpEngine::Compile(
if ((data->capture_count + 1) * 2 - 1 > RegExpMacroAssembler::kMaxRegister) {
return IrregexpRegExpTooBig(isolate);
}
bool ignore_case = flags & JSRegExp::kIgnoreCase;
bool is_sticky = flags & JSRegExp::kSticky;
bool is_global = flags & JSRegExp::kGlobal;
bool is_unicode = flags & JSRegExp::kUnicode;
RegExpCompiler compiler(isolate, zone, data->capture_count, flags,
is_one_byte);
bool is_sticky = IsSticky(flags);
bool is_global = IsGlobal(flags);
bool is_unicode = IsUnicode(flags);
RegExpCompiler compiler(isolate, zone, data->capture_count, is_one_byte);
if (compiler.optimize()) compiler.set_optimize(!TooMuchRegExpCode(pattern));
@ -6673,9 +6661,11 @@ RegExpEngine::CompilationResult RegExpEngine::Compile(
if (!is_start_anchored && !is_sticky) {
// Add a .*? at the beginning, outside the body capture, unless
// this expression is anchored at the beginning or sticky.
JSRegExp::Flags default_flags = JSRegExp::Flags();
RegExpNode* loop_node = RegExpQuantifier::ToNode(
0, RegExpTree::kInfinity, false, new (zone) RegExpCharacterClass('*'),
&compiler, captured_body, data->contains_anchor);
0, RegExpTree::kInfinity, false,
new (zone) RegExpCharacterClass('*', default_flags), &compiler,
captured_body, data->contains_anchor);
if (data->contains_anchor) {
// Unroll loop once, to take care of the case that might start
@ -6683,26 +6673,27 @@ RegExpEngine::CompilationResult RegExpEngine::Compile(
ChoiceNode* first_step_node = new(zone) ChoiceNode(2, zone);
first_step_node->AddAlternative(GuardedAlternative(captured_body));
first_step_node->AddAlternative(GuardedAlternative(new (zone) TextNode(
new (zone) RegExpCharacterClass('*'), false, loop_node)));
new (zone) RegExpCharacterClass('*', default_flags), false,
loop_node)));
node = first_step_node;
} else {
node = loop_node;
}
}
if (is_one_byte) {
node = node->FilterOneByte(RegExpCompiler::kMaxRecursion, ignore_case);
node = node->FilterOneByte(RegExpCompiler::kMaxRecursion);
// Do it again to propagate the new nodes to places where they were not
// put because they had not been calculated yet.
if (node != nullptr) {
node = node->FilterOneByte(RegExpCompiler::kMaxRecursion, ignore_case);
node = node->FilterOneByte(RegExpCompiler::kMaxRecursion);
}
} else if (compiler.unicode() && (is_global || is_sticky)) {
node = OptionallyStepBackToLeadSurrogate(&compiler, node);
} else if (is_unicode && (is_global || is_sticky)) {
node = OptionallyStepBackToLeadSurrogate(&compiler, node, flags);
}
if (node == nullptr) node = new (zone) EndNode(EndNode::BACKTRACK, zone);
data->node = node;
Analysis analysis(isolate, flags, is_one_byte);
Analysis analysis(isolate, is_one_byte);
analysis.EnsureAnalyzed(node);
if (analysis.has_failed()) {
const char* error_message = analysis.error_message();

71
src/regexp/jsregexp.h
View File

@ -21,6 +21,36 @@ class RegExpNode;
class RegExpTree;
class BoyerMooreLookahead;
inline bool IgnoreCase(JSRegExp::Flags flags) {
return (flags & JSRegExp::kIgnoreCase) != 0;
}
inline bool IsUnicode(JSRegExp::Flags flags) {
return (flags & JSRegExp::kUnicode) != 0;
}
inline bool IsSticky(JSRegExp::Flags flags) {
return (flags & JSRegExp::kSticky) != 0;
}
inline bool IsGlobal(JSRegExp::Flags flags) {
return (flags & JSRegExp::kGlobal) != 0;
}
inline bool DotAll(JSRegExp::Flags flags) {
return (flags & JSRegExp::kDotAll) != 0;
}
inline bool Multiline(JSRegExp::Flags flags) {
return (flags & JSRegExp::kMultiline) != 0;
}
inline bool NeedsUnicodeCaseEquivalents(JSRegExp::Flags flags) {
// Both unicode and ignore_case flags are set. We need to use ICU to find
// the closure over case equivalents.
return IsUnicode(flags) && IgnoreCase(flags);
}
class RegExpImpl {
public:
// Whether V8 is compiled with native regexp support or not.
@ -495,9 +525,7 @@ class RegExpNode: public ZoneObject {
// If we know that the input is one-byte then there are some nodes that can
// never match. This method returns a node that can be substituted for
// itself, or nullptr if the node can never match.
virtual RegExpNode* FilterOneByte(int depth, bool ignore_case) {
return this;
}
virtual RegExpNode* FilterOneByte(int depth) { return this; }
// Helper for FilterOneByte.
RegExpNode* replacement() {
DCHECK(info()->replacement_calculated);
@ -569,7 +597,7 @@ class SeqRegExpNode: public RegExpNode {
: RegExpNode(on_success->zone()), on_success_(on_success) { }
RegExpNode* on_success() { return on_success_; }
void set_on_success(RegExpNode* node) { on_success_ = node; }
virtual RegExpNode* FilterOneByte(int depth, bool ignore_case);
virtual RegExpNode* FilterOneByte(int depth);
virtual void FillInBMInfo(Isolate* isolate, int offset, int budget,
BoyerMooreLookahead* bm, bool not_at_start) {
on_success_->FillInBMInfo(isolate, offset, budget - 1, bm, not_at_start);
@ -577,7 +605,7 @@ class SeqRegExpNode: public RegExpNode {
}
protected:
RegExpNode* FilterSuccessor(int depth, bool ignore_case);
RegExpNode* FilterSuccessor(int depth);
private:
RegExpNode* on_success_;
@ -682,13 +710,15 @@ class TextNode: public SeqRegExpNode {
static TextNode* CreateForCharacterRanges(Zone* zone,
ZoneList<CharacterRange>* ranges,
bool read_backward,
RegExpNode* on_success);
RegExpNode* on_success,
JSRegExp::Flags flags);
// Create TextNode for a surrogate pair with a range given for the
// lead and the trail surrogate each.
static TextNode* CreateForSurrogatePair(Zone* zone, CharacterRange lead,
CharacterRange trail,
bool read_backward,
RegExpNode* on_success);
RegExpNode* on_success,
JSRegExp::Flags flags);
virtual void Accept(NodeVisitor* visitor);
virtual void Emit(RegExpCompiler* compiler, Trace* trace);
virtual int EatsAtLeast(int still_to_find, int budget, bool not_at_start);
@ -705,7 +735,7 @@ class TextNode: public SeqRegExpNode {
virtual void FillInBMInfo(Isolate* isolate, int offset, int budget,
BoyerMooreLookahead* bm, bool not_at_start);
void CalculateOffsets();
virtual RegExpNode* FilterOneByte(int depth, bool ignore_case);
virtual RegExpNode* FilterOneByte(int depth);
private:
enum TextEmitPassType {
@ -715,7 +745,7 @@ class TextNode: public SeqRegExpNode {
CASE_CHARACTER_MATCH, // Case-independent single character check.
CHARACTER_CLASS_MATCH // Character class.
};
static bool SkipPass(int pass, bool ignore_case);
static bool SkipPass(TextEmitPassType pass, bool ignore_case);
static const int kFirstRealPass = SIMPLE_CHARACTER_MATCH;
static const int kLastPass = CHARACTER_CLASS_MATCH;
void TextEmitPass(RegExpCompiler* compiler,
@ -779,11 +809,12 @@ class AssertionNode: public SeqRegExpNode {
class BackReferenceNode: public SeqRegExpNode {
public:
BackReferenceNode(int start_reg, int end_reg, bool read_backward,
RegExpNode* on_success)
BackReferenceNode(int start_reg, int end_reg, JSRegExp::Flags flags,
bool read_backward, RegExpNode* on_success)
: SeqRegExpNode(on_success),
start_reg_(start_reg),
end_reg_(end_reg),
flags_(flags),
read_backward_(read_backward) {}
virtual void Accept(NodeVisitor* visitor);
int start_register() { return start_reg_; }
@ -805,6 +836,7 @@ class BackReferenceNode: public SeqRegExpNode {
private:
int start_reg_;
int end_reg_;
JSRegExp::Flags flags_;
bool read_backward_;
};
@ -929,7 +961,7 @@ class ChoiceNode: public RegExpNode {
virtual bool try_to_emit_quick_check_for_alternative(bool is_first) {
return true;
}
virtual RegExpNode* FilterOneByte(int depth, bool ignore_case);
virtual RegExpNode* FilterOneByte(int depth);
virtual bool read_backward() { return false; }
protected:
@ -1001,7 +1033,7 @@ class NegativeLookaroundChoiceNode : public ChoiceNode {
virtual bool try_to_emit_quick_check_for_alternative(bool is_first) {
return !is_first;
}
virtual RegExpNode* FilterOneByte(int depth, bool ignore_case);
virtual RegExpNode* FilterOneByte(int depth);
};
@ -1028,7 +1060,7 @@ class LoopChoiceNode: public ChoiceNode {
bool body_can_be_zero_length() { return body_can_be_zero_length_; }
virtual bool read_backward() { return read_backward_; }
virtual void Accept(NodeVisitor* visitor);
virtual RegExpNode* FilterOneByte(int depth, bool ignore_case);
virtual RegExpNode* FilterOneByte(int depth);
private:
// AddAlternative is made private for loop nodes because alternatives
@ -1435,11 +1467,8 @@ FOR_EACH_NODE_TYPE(DECLARE_VISIT)
// +-------+ +------------+
class Analysis: public NodeVisitor {
public:
Analysis(Isolate* isolate, JSRegExp::Flags flags, bool is_one_byte)
: isolate_(isolate),
flags_(flags),
is_one_byte_(is_one_byte),
error_message_(nullptr) {}
Analysis(Isolate* isolate, bool is_one_byte)
: isolate_(isolate), is_one_byte_(is_one_byte), error_message_(nullptr) {}
void EnsureAnalyzed(RegExpNode* node);
#define DECLARE_VISIT(Type) \
@ -1459,12 +1488,8 @@ FOR_EACH_NODE_TYPE(DECLARE_VISIT)
Isolate* isolate() const { return isolate_; }
bool ignore_case() const { return (flags_ & JSRegExp::kIgnoreCase) != 0; }
bool unicode() const { return (flags_ & JSRegExp::kUnicode) != 0; }
private:
Isolate* isolate_;
JSRegExp::Flags flags_;
bool is_one_byte_;
const char* error_message_;

49
src/regexp/regexp-ast.h
View File

@ -6,6 +6,7 @@
#define V8_REGEXP_REGEXP_AST_H_
#include "src/objects.h"
#include "src/objects/js-regexp.h"
#include "src/objects/string.h"
#include "src/utils.h"
#include "src/zone/zone-containers.h"
@ -144,7 +145,7 @@ class CharacterSet final BASE_EMBEDDED {
explicit CharacterSet(ZoneList<CharacterRange>* ranges)
: ranges_(ranges), standard_set_type_(0) {}
ZoneList<CharacterRange>* ranges(Zone* zone);
uc16 standard_set_type() { return standard_set_type_; }
uc16 standard_set_type() const { return standard_set_type_; }
void set_standard_set_type(uc16 special_set_type) {
standard_set_type_ = special_set_type;
}
@ -274,7 +275,8 @@ class RegExpAssertion final : public RegExpTree {
BOUNDARY,
NON_BOUNDARY
};
explicit RegExpAssertion(AssertionType type) : assertion_type_(type) {}
RegExpAssertion(AssertionType type, JSRegExp::Flags flags)
: assertion_type_(type), flags_(flags) {}
void* Accept(RegExpVisitor* visitor, void* data) override;
RegExpNode* ToNode(RegExpCompiler* compiler, RegExpNode* on_success) override;
RegExpAssertion* AsAssertion() override;
@ -286,7 +288,8 @@ class RegExpAssertion final : public RegExpTree {
AssertionType assertion_type() { return assertion_type_; }
private:
AssertionType assertion_type_;
const AssertionType assertion_type_;
const JSRegExp::Flags flags_;
};
@ -300,12 +303,18 @@ class RegExpCharacterClass final : public RegExpTree {
NEGATED = 1 << 0,
CONTAINS_SPLIT_SURROGATE = 1 << 1,
};
typedef base::Flags<Flag> Flags;
typedef base::Flags<Flag> CharacterClassFlags;
explicit RegExpCharacterClass(ZoneList<CharacterRange>* ranges,
Flags flags = Flags())
: set_(ranges), flags_(flags) {}
explicit RegExpCharacterClass(uc16 type) : set_(type), flags_(0) {}
RegExpCharacterClass(
ZoneList<CharacterRange>* ranges, JSRegExp::Flags flags,
CharacterClassFlags character_class_flags = CharacterClassFlags())
: set_(ranges),
flags_(flags),
character_class_flags_(character_class_flags) {}
RegExpCharacterClass(uc16 type, JSRegExp::Flags flags)
: set_(type),
flags_(flags),
character_class_flags_(CharacterClassFlags()) {}
void* Accept(RegExpVisitor* visitor, void* data) override;
RegExpNode* ToNode(RegExpCompiler* compiler, RegExpNode* on_success) override;
RegExpCharacterClass* AsCharacterClass() override;
@ -332,22 +341,25 @@ class RegExpCharacterClass final : public RegExpTree {
// D : non-ASCII digit
// . : non-newline
// * : All characters, for advancing unanchored regexp
uc16 standard_type() { return set_.standard_set_type(); }
uc16 standard_type() const { return set_.standard_set_type(); }
ZoneList<CharacterRange>* ranges(Zone* zone) { return set_.ranges(zone); }
bool is_negated() const { return (flags_ & NEGATED) != 0; }
bool is_negated() const { return (character_class_flags_ & NEGATED) != 0; }
JSRegExp::Flags flags() const { return flags_; }
bool contains_split_surrogate() const {
return (flags_ & CONTAINS_SPLIT_SURROGATE) != 0;
return (character_class_flags_ & CONTAINS_SPLIT_SURROGATE) != 0;
}
private:
CharacterSet set_;
const Flags flags_;
const JSRegExp::Flags flags_;
const CharacterClassFlags character_class_flags_;
};
class RegExpAtom final : public RegExpTree {
public:
explicit RegExpAtom(Vector<const uc16> data) : data_(data) {}
explicit RegExpAtom(Vector<const uc16> data, JSRegExp::Flags flags)
: data_(data), flags_(flags) {}
void* Accept(RegExpVisitor* visitor, void* data) override;
RegExpNode* ToNode(RegExpCompiler* compiler, RegExpNode* on_success) override;
RegExpAtom* AsAtom() override;
@ -358,9 +370,12 @@ class RegExpAtom final : public RegExpTree {
void AppendToText(RegExpText* text, Zone* zone) override;
Vector<const uc16> data() { return data_; }
int length() { return data_.length(); }
JSRegExp::Flags flags() const { return flags_; }
bool ignore_case() const { return (flags_ & JSRegExp::kIgnoreCase) != 0; }
private:
Vector<const uc16> data_;
const JSRegExp::Flags flags_;
};
@ -532,9 +547,10 @@ class RegExpLookaround final : public RegExpTree {
class RegExpBackReference final : public RegExpTree {
public:
RegExpBackReference() : capture_(nullptr), name_(nullptr) {}
explicit RegExpBackReference(RegExpCapture* capture)
: capture_(capture), name_(nullptr) {}
explicit RegExpBackReference(JSRegExp::Flags flags)
: capture_(nullptr), name_(nullptr), flags_(flags) {}
RegExpBackReference(RegExpCapture* capture, JSRegExp::Flags flags)
: capture_(capture), name_(nullptr), flags_(flags) {}
void* Accept(RegExpVisitor* visitor, void* data) override;
RegExpNode* ToNode(RegExpCompiler* compiler, RegExpNode* on_success) override;
RegExpBackReference* AsBackReference() override;
@ -552,6 +568,7 @@ class RegExpBackReference final : public RegExpTree {
private:
RegExpCapture* capture_;
const ZoneVector<uc16>* name_;
const JSRegExp::Flags flags_;
};

281
src/regexp/regexp-parser.cc
View File

@ -31,10 +31,7 @@ RegExpParser::RegExpParser(FlatStringReader* in, Handle<String>* error,
named_back_references_(nullptr),
in_(in),
current_(kEndMarker),
dotall_(flags & JSRegExp::kDotAll),
ignore_case_(flags & JSRegExp::kIgnoreCase),
multiline_(flags & JSRegExp::kMultiline),
unicode_(flags & JSRegExp::kUnicode),
top_level_flags_(flags),
next_pos_(0),
captures_started_(0),
capture_count_(0),
@ -44,7 +41,6 @@ RegExpParser::RegExpParser(FlatStringReader* in, Handle<String>* error,
is_scanned_for_captures_(false),
has_named_captures_(false),
failed_(false) {
DCHECK_IMPLIES(dotall(), FLAG_harmony_regexp_dotall);
Advance();
}
@ -183,7 +179,7 @@ RegExpTree* RegExpParser::ParsePattern() {
RegExpTree* RegExpParser::ParseDisjunction() {
// Used to store current state while parsing subexpressions.
RegExpParserState initial_state(nullptr, INITIAL, RegExpLookaround::LOOKAHEAD,
0, nullptr, ignore_case(), unicode(), zone());
0, nullptr, top_level_flags_, zone());
RegExpParserState* state = &initial_state;
// Cache the builder in a local variable for quick access.
RegExpBuilder* builder = initial_state.builder();
@ -253,12 +249,12 @@ RegExpTree* RegExpParser::ParseDisjunction() {
return ReportError(CStrVector("Nothing to repeat"));
case '^': {
Advance();
if (multiline()) {
builder->AddAssertion(
new (zone()) RegExpAssertion(RegExpAssertion::START_OF_LINE));
if (builder->multiline()) {
builder->AddAssertion(new (zone()) RegExpAssertion(
RegExpAssertion::START_OF_LINE, builder->flags()));
} else {
builder->AddAssertion(
new (zone()) RegExpAssertion(RegExpAssertion::START_OF_INPUT));
builder->AddAssertion(new (zone()) RegExpAssertion(
RegExpAssertion::START_OF_INPUT, builder->flags()));
set_contains_anchor();
}
continue;
@ -266,9 +262,10 @@ RegExpTree* RegExpParser::ParseDisjunction() {
case '$': {
Advance();
RegExpAssertion::AssertionType assertion_type =
multiline() ? RegExpAssertion::END_OF_LINE
: RegExpAssertion::END_OF_INPUT;
builder->AddAssertion(new (zone()) RegExpAssertion(assertion_type));
builder->multiline() ? RegExpAssertion::END_OF_LINE
: RegExpAssertion::END_OF_INPUT;
builder->AddAssertion(
new (zone()) RegExpAssertion(assertion_type, builder->flags()));
continue;
}
case '.': {
@ -276,7 +273,7 @@ RegExpTree* RegExpParser::ParseDisjunction() {
ZoneList<CharacterRange>* ranges =
new (zone()) ZoneList<CharacterRange>(2, zone());
if (dotall()) {
if (builder->dotall()) {
// Everything.
DCHECK(FLAG_harmony_regexp_dotall);
CharacterRange::AddClassEscape('*', ranges, false, zone());
@ -285,78 +282,18 @@ RegExpTree* RegExpParser::ParseDisjunction() {
CharacterRange::AddClassEscape('.', ranges, false, zone());
}
RegExpCharacterClass* cc = new (zone()) RegExpCharacterClass(ranges);
RegExpCharacterClass* cc =
new (zone()) RegExpCharacterClass(ranges, builder->flags());
builder->AddCharacterClass(cc);
break;
}
case '(': {
SubexpressionType subexpr_type = CAPTURE;
RegExpLookaround::Type lookaround_type = state->lookaround_type();
bool is_named_capture = false;
Advance();
if (current() == '?') {
switch (Next()) {
case ':':
subexpr_type = GROUPING;
Advance(2);
break;
case '=':
lookaround_type = RegExpLookaround::LOOKAHEAD;
subexpr_type = POSITIVE_LOOKAROUND;
Advance(2);
break;
case '!':
lookaround_type = RegExpLookaround::LOOKAHEAD;
subexpr_type = NEGATIVE_LOOKAROUND;
Advance(2);
break;
case '<':
Advance();
if (FLAG_harmony_regexp_lookbehind) {
if (Next() == '=') {
subexpr_type = POSITIVE_LOOKAROUND;
lookaround_type = RegExpLookaround::LOOKBEHIND;
Advance(2);
break;
} else if (Next() == '!') {
subexpr_type = NEGATIVE_LOOKAROUND;
lookaround_type = RegExpLookaround::LOOKBEHIND;
Advance(2);
break;
}
}
if (FLAG_harmony_regexp_named_captures) {
has_named_captures_ = true;
is_named_capture = true;
Advance();
break;
}
// Fall through.
default:
return ReportError(CStrVector("Invalid group"));
}
}
const ZoneVector<uc16>* capture_name = nullptr;
if (subexpr_type == CAPTURE) {
if (captures_started_ >= kMaxCaptures) {
return ReportError(CStrVector("Too many captures"));
}
captures_started_++;
if (is_named_capture) {
capture_name = ParseCaptureGroupName(CHECK_FAILED);
}
}
// Store current state and begin new disjunction parsing.
state = new (zone()) RegExpParserState(
state, subexpr_type, lookaround_type, captures_started_,
capture_name, ignore_case(), unicode(), zone());
state = ParseOpenParenthesis(state CHECK_FAILED);
builder = state->builder();
continue;
}
case '[': {
RegExpTree* cc = ParseCharacterClass(CHECK_FAILED);
RegExpTree* cc = ParseCharacterClass(builder CHECK_FAILED);
builder->AddCharacterClass(cc->AsCharacterClass());
break;
}
@ -368,13 +305,13 @@ RegExpTree* RegExpParser::ParseDisjunction() {
return ReportError(CStrVector("\\ at end of pattern"));
case 'b':
Advance(2);
builder->AddAssertion(
new (zone()) RegExpAssertion(RegExpAssertion::BOUNDARY));
builder->AddAssertion(new (zone()) RegExpAssertion(
RegExpAssertion::BOUNDARY, builder->flags()));
continue;
case 'B':
Advance(2);
builder->AddAssertion(
new (zone()) RegExpAssertion(RegExpAssertion::NON_BOUNDARY));
builder->AddAssertion(new (zone()) RegExpAssertion(
RegExpAssertion::NON_BOUNDARY, builder->flags()));
continue;
// AtomEscape ::
// CharacterClassEscape
@ -391,10 +328,10 @@ RegExpTree* RegExpParser::ParseDisjunction() {
Advance(2);
ZoneList<CharacterRange>* ranges =
new (zone()) ZoneList<CharacterRange>(2, zone());
CharacterRange::AddClassEscape(c, ranges,
unicode() && ignore_case(), zone());
CharacterRange::AddClassEscape(
c, ranges, unicode() && builder->ignore_case(), zone());
RegExpCharacterClass* cc =
new (zone()) RegExpCharacterClass(ranges);
new (zone()) RegExpCharacterClass(ranges, builder->flags());
builder->AddCharacterClass(cc);
break;
}
@ -410,7 +347,7 @@ RegExpTree* RegExpParser::ParseDisjunction() {
return ReportError(CStrVector("Invalid property name"));
}
RegExpCharacterClass* cc =
new (zone()) RegExpCharacterClass(ranges);
new (zone()) RegExpCharacterClass(ranges, builder->flags());
builder->AddCharacterClass(cc);
} else {
// With /u, no identity escapes except for syntax characters
@ -443,7 +380,8 @@ RegExpTree* RegExpParser::ParseDisjunction() {
builder->AddEmpty();
} else {
RegExpCapture* capture = GetCapture(index);
RegExpTree* atom = new (zone()) RegExpBackReference(capture);
RegExpTree* atom =
new (zone()) RegExpBackReference(capture, builder->flags());
builder->AddAtom(atom);
}
break;
@ -638,6 +576,143 @@ RegExpTree* RegExpParser::ParseDisjunction() {
}
}
RegExpParser::RegExpParserState* RegExpParser::ParseOpenParenthesis(
RegExpParserState* state) {
RegExpLookaround::Type lookaround_type = state->lookaround_type();
bool is_named_capture = false;
JSRegExp::Flags switch_on = JSRegExp::kNone;
JSRegExp::Flags switch_off = JSRegExp::kNone;
const ZoneVector<uc16>* capture_name = nullptr;
SubexpressionType subexpr_type = CAPTURE;
Advance();
if (current() == '?') {
switch (Next()) {
case ':':
Advance(2);
subexpr_type = GROUPING;
break;
case '=':
Advance(2);
lookaround_type = RegExpLookaround::LOOKAHEAD;
subexpr_type = POSITIVE_LOOKAROUND;
break;
case '!':
Advance(2);
lookaround_type = RegExpLookaround::LOOKAHEAD;
subexpr_type = NEGATIVE_LOOKAROUND;
break;
case '-':
case 'i':
case 's':
case 'm': {
if (!FLAG_regexp_mode_modifiers ||
(Next() == 's' && !FLAG_harmony_regexp_dotall)) {
ReportError(CStrVector("Invalid group"));
return nullptr;
}
Advance();
bool flags_sense = true; // Switching on flags.
while (subexpr_type != GROUPING) {
switch (current()) {
case '-':
if (!flags_sense) {
ReportError(CStrVector("Multiple dashes in flag group"));
return nullptr;
}
flags_sense = false;
Advance();
continue;
case 's':
if (!FLAG_harmony_regexp_dotall) {
ReportError(CStrVector("Invalid group"));
return nullptr;
}
// Fall through.
case 'i':
case 'm': {
JSRegExp::Flags bit = JSRegExp::kUnicode;
if (current() == 'i') bit = JSRegExp::kIgnoreCase;
if (current() == 'm') bit = JSRegExp::kMultiline;
if (current() == 's') bit = JSRegExp::kDotAll;
if (((switch_on | switch_off) & bit) != 0) {
ReportError(CStrVector("Repeated flag in flag group"));
return nullptr;
}
if (flags_sense) {
switch_on |= bit;
} else {
switch_off |= bit;
}
Advance();
continue;
}
case ')': {
Advance();
state->builder()
->FlushText(); // Flush pending text using old flags.
// These (?i)-style flag switches don't put us in a subexpression
// at all, they just modify the flags in the rest of the current
// subexpression.
JSRegExp::Flags flags =
(state->builder()->flags() | switch_on) & ~switch_off;
state->builder()->set_flags(flags);
return state;
}
case ':':
Advance();
subexpr_type = GROUPING; // Will break us out of the outer loop.
continue;
default:
ReportError(CStrVector("Invalid flag group"));
return nullptr;
}
}
break;
}
case '<':
Advance();
if (FLAG_harmony_regexp_lookbehind) {
if (Next() == '=') {
Advance(2);
lookaround_type = RegExpLookaround::LOOKBEHIND;
subexpr_type = POSITIVE_LOOKAROUND;
break;
} else if (Next() == '!') {
Advance(2);
lookaround_type = RegExpLookaround::LOOKBEHIND;
subexpr_type = NEGATIVE_LOOKAROUND;
break;
}
}
if (FLAG_harmony_regexp_named_captures) {
is_named_capture = true;
has_named_captures_ = true;
Advance();
break;
}
// Fall through.
default:
ReportError(CStrVector("Invalid group"));
return nullptr;
}
}
if (subexpr_type == CAPTURE) {
if (captures_started_ >= kMaxCaptures) {
ReportError(CStrVector("Too many captures"));
return nullptr;
}
captures_started_++;
if (is_named_capture) {
capture_name = ParseCaptureGroupName(CHECK_FAILED);
}
}
JSRegExp::Flags flags = (state->builder()->flags() | switch_on) & ~switch_off;
// Store current state and begin new disjunction parsing.
return new (zone())
RegExpParserState(state, subexpr_type, lookaround_type, captures_started_,
capture_name, flags, zone());
}
#ifdef DEBUG
// Currently only used in an DCHECK.
@ -855,7 +930,8 @@ bool RegExpParser::ParseNamedBackReference(RegExpBuilder* builder,
if (state->IsInsideCaptureGroup(name)) {
builder->AddEmpty();
} else {
RegExpBackReference* atom = new (zone()) RegExpBackReference();
RegExpBackReference* atom =
new (zone()) RegExpBackReference(builder->flags());
atom->set_name(name);
builder->AddAtom(atom);
@ -1525,7 +1601,7 @@ void RegExpParser::ParseClassEscape(ZoneList<CharacterRange>* ranges,
}
}
RegExpTree* RegExpParser::ParseCharacterClass() {
RegExpTree* RegExpParser::ParseCharacterClass(const RegExpBuilder* builder) {
static const char* kUnterminated = "Unterminated character class";
static const char* kRangeInvalid = "Invalid character class";
static const char* kRangeOutOfOrder = "Range out of order in character class";
@ -1539,7 +1615,7 @@ RegExpTree* RegExpParser::ParseCharacterClass() {
}
ZoneList<CharacterRange>* ranges =
new (zone()) ZoneList<CharacterRange>(2, zone());
bool add_unicode_case_equivalents = unicode() && ignore_case();
bool add_unicode_case_equivalents = unicode() && builder->ignore_case();
while (has_more() && current() != ']') {
uc32 char_1, char_2;
bool is_class_1, is_class_2;
@ -1586,9 +1662,10 @@ RegExpTree* RegExpParser::ParseCharacterClass() {
ranges->Add(CharacterRange::Everything(), zone());
is_negated = !is_negated;
}
RegExpCharacterClass::Flags flags;
if (is_negated) flags = RegExpCharacterClass::NEGATED;
return new (zone()) RegExpCharacterClass(ranges, flags);
RegExpCharacterClass::CharacterClassFlags character_class_flags;
if (is_negated) character_class_flags = RegExpCharacterClass::NEGATED;
return new (zone())
RegExpCharacterClass(ranges, builder->flags(), character_class_flags);
}
@ -1622,11 +1699,10 @@ bool RegExpParser::ParseRegExp(Isolate* isolate, Zone* zone,
return !parser.failed();
}
RegExpBuilder::RegExpBuilder(Zone* zone, bool ignore_case, bool unicode)
RegExpBuilder::RegExpBuilder(Zone* zone, JSRegExp::Flags flags)
: zone_(zone),
pending_empty_(false),
ignore_case_(ignore_case),
unicode_(unicode),
flags_(flags),
characters_(nullptr),
pending_surrogate_(kNoPendingSurrogate),
terms_(),
@ -1662,7 +1738,7 @@ void RegExpBuilder::AddTrailSurrogate(uc16 trail_surrogate) {
surrogate_pair.Add(lead_surrogate, zone());
surrogate_pair.Add(trail_surrogate, zone());
RegExpAtom* atom =
new (zone()) RegExpAtom(surrogate_pair.ToConstVector());
new (zone()) RegExpAtom(surrogate_pair.ToConstVector(), flags_);
AddAtom(atom);
}
} else {
@ -1686,7 +1762,8 @@ void RegExpBuilder::FlushCharacters() {
FlushPendingSurrogate();
pending_empty_ = false;
if (characters_ != nullptr) {
RegExpTree* atom = new (zone()) RegExpAtom(characters_->ToConstVector());
RegExpTree* atom =
new (zone()) RegExpAtom(characters_->ToConstVector(), flags_);
characters_ = nullptr;
text_.Add(atom, zone());
LAST(ADD_ATOM);
@ -1762,7 +1839,7 @@ void RegExpBuilder::AddCharacterClass(RegExpCharacterClass* cc) {
void RegExpBuilder::AddCharacterClassForDesugaring(uc32 c) {
AddTerm(new (zone()) RegExpCharacterClass(
CharacterRange::List(zone(), CharacterRange::Singleton(c))));
CharacterRange::List(zone(), CharacterRange::Singleton(c)), flags_));
}
@ -1880,11 +1957,11 @@ bool RegExpBuilder::AddQuantifierToAtom(
int num_chars = char_vector.length();
if (num_chars > 1) {
Vector<const uc16> prefix = char_vector.SubVector(0, num_chars - 1);
text_.Add(new (zone()) RegExpAtom(prefix), zone());
text_.Add(new (zone()) RegExpAtom(prefix, flags_), zone());
char_vector = char_vector.SubVector(num_chars - 1, num_chars);
}
characters_ = nullptr;
atom = new (zone()) RegExpAtom(char_vector);
atom = new (zone()) RegExpAtom(char_vector, flags_);
FlushText();
} else if (text_.length() > 0) {
DCHECK(last_added_ == ADD_ATOM);

65
src/regexp/regexp-parser.h
View File

@ -99,7 +99,7 @@ class BufferedZoneList {
// Accumulates RegExp atoms and assertions into lists of terms and alternatives.
class RegExpBuilder : public ZoneObject {
public:
RegExpBuilder(Zone* zone, bool ignore_case, bool unicode);
RegExpBuilder(Zone* zone, JSRegExp::Flags flags);
void AddCharacter(uc16 character);
void AddUnicodeCharacter(uc32 character);
void AddEscapedUnicodeCharacter(uc32 character);
@ -114,7 +114,14 @@ class RegExpBuilder : public ZoneObject {
void NewAlternative(); // '|'
bool AddQuantifierToAtom(int min, int max,
RegExpQuantifier::QuantifierType type);
void FlushText();
RegExpTree* ToRegExp();
JSRegExp::Flags flags() const { return flags_; }
void set_flags(JSRegExp::Flags flags) { flags_ = flags; }
bool ignore_case() const { return (flags_ & JSRegExp::kIgnoreCase) != 0; }
bool multiline() const { return (flags_ & JSRegExp::kMultiline) != 0; }
bool dotall() const { return (flags_ & JSRegExp::kDotAll) != 0; }
private:
static const uc16 kNoPendingSurrogate = 0;
@ -122,18 +129,15 @@ class RegExpBuilder : public ZoneObject {
void AddTrailSurrogate(uc16 trail_surrogate);
void FlushPendingSurrogate();
void FlushCharacters();
void FlushText();
void FlushTerms();
bool NeedsDesugaringForUnicode(RegExpCharacterClass* cc);
bool NeedsDesugaringForIgnoreCase(uc32 c);
Zone* zone() const { return zone_; }
bool ignore_case() const { return ignore_case_; }
bool unicode() const { return unicode_; }
bool unicode() const { return (flags_ & JSRegExp::kUnicode) != 0; }
Zone* zone_;
bool pending_empty_;
bool ignore_case_;
bool unicode_;
JSRegExp::Flags flags_;
ZoneList<uc16>* characters_;
uc16 pending_surrogate_;
BufferedZoneList<RegExpTree, 2> terms_;
@ -159,7 +163,6 @@ class RegExpParser BASE_EMBEDDED {
RegExpTree* ParsePattern();
RegExpTree* ParseDisjunction();
RegExpTree* ParseGroup();
RegExpTree* ParseCharacterClass();
// Parses a {...,...} quantifier and stores the range in the given
// out parameters.
@ -175,6 +178,7 @@ class RegExpParser BASE_EMBEDDED {
bool ParseUnicodeEscape(uc32* value);
bool ParseUnlimitedLengthHexNumber(int max_value, uc32* value);
bool ParsePropertyClass(ZoneList<CharacterRange>* result, bool negate);
RegExpTree* ParseCharacterClass(const RegExpBuilder* state);
uc32 ParseOctalLiteral();
@ -205,10 +209,9 @@ class RegExpParser BASE_EMBEDDED {
int captures_started() { return captures_started_; }
int position() { return next_pos_ - 1; }
bool failed() { return failed_; }
bool dotall() const { return dotall_; }
bool ignore_case() const { return ignore_case_; }
bool multiline() const { return multiline_; }
bool unicode() const { return unicode_; }
// The Unicode flag can't be changed using in-regexp syntax, so it's OK to
// just read the initial flag value here.
bool unicode() const { return (top_level_flags_ & JSRegExp::kUnicode) != 0; }
static bool IsSyntaxCharacterOrSlash(uc32 c);
@ -226,34 +229,35 @@ class RegExpParser BASE_EMBEDDED {
class RegExpParserState : public ZoneObject {
public:
// Push a state on the stack.
RegExpParserState(RegExpParserState* previous_state,
SubexpressionType group_type,
RegExpLookaround::Type lookaround_type,
int disjunction_capture_index,
const ZoneVector<uc16>* capture_name, bool ignore_case,
bool unicode, Zone* zone)
const ZoneVector<uc16>* capture_name,
JSRegExp::Flags flags, Zone* zone)
: previous_state_(previous_state),
builder_(new (zone) RegExpBuilder(zone, ignore_case, unicode)),
builder_(new (zone) RegExpBuilder(zone, flags)),
group_type_(group_type),
lookaround_type_(lookaround_type),
disjunction_capture_index_(disjunction_capture_index),
capture_name_(capture_name) {}
// Parser state of containing expression, if any.
RegExpParserState* previous_state() { return previous_state_; }
RegExpParserState* previous_state() const { return previous_state_; }
bool IsSubexpression() { return previous_state_ != nullptr; }
// RegExpBuilder building this regexp's AST.
RegExpBuilder* builder() { return builder_; }
RegExpBuilder* builder() const { return builder_; }
// Type of regexp being parsed (parenthesized group or entire regexp).
SubexpressionType group_type() { return group_type_; }
SubexpressionType group_type() const { return group_type_; }
// Lookahead or Lookbehind.
RegExpLookaround::Type lookaround_type() { return lookaround_type_; }
RegExpLookaround::Type lookaround_type() const { return lookaround_type_; }
// Index in captures array of first capture in this sub-expression, if any.
// Also the capture index of this sub-expression itself, if group_type
// is CAPTURE.
int capture_index() { return disjunction_capture_index_; }
int capture_index() const { return disjunction_capture_index_; }
// The name of the current sub-expression, if group_type is CAPTURE. Only
// used for named captures.
const ZoneVector<uc16>* capture_name() { return capture_name_; }
const ZoneVector<uc16>* capture_name() const { return capture_name_; }
bool IsNamedCapture() const { return capture_name_ != nullptr; }
@ -264,17 +268,17 @@ class RegExpParser BASE_EMBEDDED {
private:
// Linked list implementation of stack of states.
RegExpParserState* previous_state_;
RegExpParserState* const previous_state_;
// Builder for the stored disjunction.
RegExpBuilder* builder_;
RegExpBuilder* const builder_;
// Stored disjunction type (capture, look-ahead or grouping), if any.
SubexpressionType group_type_;
const SubexpressionType group_type_;
// Stored read direction.
RegExpLookaround::Type lookaround_type_;
const RegExpLookaround::Type lookaround_type_;
// Stored disjunction's capture index (if any).
int disjunction_capture_index_;
const int disjunction_capture_index_;
// Stored capture name (if any).
const ZoneVector<uc16>* capture_name_;
const ZoneVector<uc16>* const capture_name_;
};
// Return the 1-indexed RegExpCapture object, allocate if necessary.
@ -291,6 +295,7 @@ class RegExpParser BASE_EMBEDDED {
bool ParseNamedBackReference(RegExpBuilder* builder,
RegExpParserState* state);
RegExpParserState* ParseOpenParenthesis(RegExpParserState* state);
// After the initial parsing pass, patch corresponding RegExpCapture objects
// into all RegExpBackReferences. This is done after initial parsing in order
@ -323,10 +328,10 @@ class RegExpParser BASE_EMBEDDED {
ZoneList<RegExpBackReference*>* named_back_references_;
FlatStringReader* in_;
uc32 current_;
bool dotall_;
bool ignore_case_;
bool multiline_;
bool unicode_;
// These are the flags specified outside the regexp syntax ie after the
// terminating '/' or in the second argument to the constructor. The current
// flags are stored on the RegExpBuilder.
JSRegExp::Flags top_level_flags_;
int next_pos_;
int captures_started_;
int capture_count_; // Only valid after we have scanned for captures.

2
test/mjsunit/mjsunit.status
View File

@ -158,6 +158,8 @@
'es6/unicode-regexp-ignore-case-noi18n': [FAIL, ['no_i18n == True', PASS]],
'regress/regress-5036': [PASS, ['no_i18n == True', FAIL]],
'es7/regexp-ui-word': [PASS, ['no_i18n == True', FAIL]],
'regexp-modifiers-i18n': [PASS, ['no_i18n == True', FAIL]],
'regexp-modifiers-autogenerated-i18n': [PASS, ['no_i18n == True', FAIL]],
# desugaring regexp property class relies on ICU.
'harmony/regexp-property-*': [PASS, ['no_i18n == True', FAIL]],

81
test/mjsunit/regexp-modifiers-autogenerated-i18n.js Normal file
View File

@ -0,0 +1,81 @@
// Copyright 2017 the V8 project authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
// Flags: --regexp-mode-modifiers --harmony-regexp-property
// These regexps are just grepped out of the other tests we already have
// and the syntax changed from out-of-line i flag to inline i flag.
// These tests won't all run on the noi18n build of V8.
assertTrue(/(?i)\u00e5/u.test("\u00c5"));
assertTrue(/(?i)\u00e5/u.test("\u00e5"));
assertTrue(/(?i)\u00c5/u.test("\u00e5"));
assertTrue(/(?i)\u00c5/u.test("\u00c5"));
assertTrue(/(?i)\u212b/u.test("\u212b"));
assertFalse(/(?i)\u00df/u.test("SS"));
assertFalse(/(?i)\u1f8d/u.test("\u1f05\u03b9"));
assertTrue(/(?i)\u1f6b/u.test("\u1f63"));
assertTrue(/(?i)\u00e5/u.test("\u212b"));
assertTrue(/(?i)\u00e5/u.test("\u00c5"));
assertTrue(/(?i)\u00e5/u.test("\u00e5"));
assertTrue(/(?i)\u00e5/u.test("\u212b"));
assertTrue(/(?i)\u00c5/u.test("\u00e5"));
assertTrue(/(?i)\u00c5/u.test("\u212b"));
assertTrue(/(?i)\u00c5/u.test("\u00c5"));
assertTrue(/(?i)\u212b/u.test("\u00c5"));
assertTrue(/(?i)\u212b/u.test("\u00e5"));
assertTrue(/(?i)\u212b/u.test("\u212b"));
assertTrue(/(?i)\u{10400}/u.test("\u{10428}"));
assertTrue(/(?i)\ud801\udc00/u.test("\u{10428}"));
assertTrue(/(?i)[\u{10428}]/u.test("\u{10400}"));
assertTrue(/(?i)[\ud801\udc28]/u.test("\u{10400}"));
assertFalse(/(?i)\u00df/u.test("SS"));
assertFalse(/(?i)\u1f8d/u.test("\u1f05\u03b9"));
assertTrue(/(?i)\u1f8d/u.test("\u1f85"));
assertTrue(/(?i)\u1f6b/u.test("\u1f63"));
assertTrue(/(?i)\u00e5\u00e5\u00e5/u.test("\u212b\u00e5\u00c5"));
assertTrue(/(?i)AB\u{10400}/u.test("ab\u{10428}"));
assertTrue(/(?i)\w/u.test('\u017F'));
assertTrue(/(?i)\w/u.test('\u212A'));
assertFalse(/(?i)\W/u.test('\u017F'));
assertFalse(/(?i)\W/u.test('\u212A'));
assertFalse(/(?i)\W/u.test('s'));
assertFalse(/(?i)\W/u.test('S'));
assertFalse(/(?i)\W/u.test('K'));
assertFalse(/(?i)\W/u.test('k'));
assertTrue(/(?i)[\w]/u.test('\u017F'));
assertTrue(/(?i)[\w]/u.test('\u212A'));
assertFalse(/(?i)[\W]/u.test('\u017F'));
assertFalse(/(?i)[\W]/u.test('\u212A'));
assertFalse(/(?i)[\W]/u.test('s'));
assertFalse(/(?i)[\W]/u.test('S'));
assertFalse(/(?i)[\W]/u.test('K'));
assertFalse(/(?i)[\W]/u.test('k'));
assertTrue(/(?i)\b/u.test('\u017F'));
assertTrue(/(?i)\b/u.test('\u212A'));
assertTrue(/(?i)\b/u.test('s'));
assertTrue(/(?i)\b/u.test('S'));
assertFalse(/(?i)\B/u.test('\u017F'));
assertFalse(/(?i)\B/u.test('\u212A'));
assertFalse(/(?i)\B/u.test('s'));
assertFalse(/(?i)\B/u.test('S'));
assertFalse(/(?i)\B/u.test('K'));
assertFalse(/(?i)\B/u.test('k'));
assertTrue(/(?i)\p{Ll}/u.test("a"));
assertTrue(/(?i)\p{Ll}/u.test("\u{118D4}"));
assertTrue(/(?i)\p{Ll}/u.test("A"));
assertTrue(/(?i)\p{Ll}/u.test("\u{118B4}"));
assertTrue(/(?i)\P{Ll}/u.test("a"));
assertTrue(/(?i)\P{Ll}/u.test("\u{118D4}"));
assertTrue(/(?i)\P{Ll}/u.test("A"));
assertTrue(/(?i)\P{Ll}/u.test("\u{118B4}"));
assertTrue(/(?i)\p{Lu}/u.test("a"));
assertTrue(/(?i)\p{Lu}/u.test("\u{118D4}"));
assertTrue(/(?i)\p{Lu}/u.test("A"));
assertTrue(/(?i)\p{Lu}/u.test("\u{118B4}"));
assertTrue(/(?i)\P{Lu}/u.test("a"));
assertTrue(/(?i)\P{Lu}/u.test("\u{118D4}"));
assertTrue(/(?i)\P{Lu}/u.test("A"));
assertTrue(/(?i)\P{Lu}/u.test("\u{118B4}"));

74
test/mjsunit/regexp-modifiers-autogenerated.js Normal file
View File

@ -0,0 +1,74 @@
// Copyright 2017 the V8 project authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
// Flags: --regexp-mode-modifiers --harmony-regexp-property
// These regexps are just grepped out of the other tests we already have
// and the syntax changed from out-of-line i flag to inline i flag.
assertFalse(/(?i)x(...)\1/.test("x\u03a3\u03c2\u03c3\u03c2\u03c3"));
assertTrue(/(?i)\u03a3((?:))\1\1x/.test("\u03c2x"), "backref-UC16-empty");
assertTrue(/(?i)x(?:...|(...))\1x/.test("x\u03a3\u03c2\u03c3x"));
assertTrue(/(?i)x(?:...|(...))\1x/.test("x\u03c2\u03c3\u039b\u03a3\u03c2\u03bbx"));
assertFalse(/(?i)\xc1/.test('fooA'), "quickcheck-uc16-pattern-ascii-subject");
assertFalse(/(?i)x(...)\1/.test("xaaaaa"), "backref-ASCII-short");
assertTrue(/(?i)x((?:))\1\1x/.test("xx"), "backref-ASCII-empty");
assertTrue(/(?i)x(?:...|(...))\1x/.test("xabcx"), "backref-ASCII-uncaptured");
assertTrue(/(?i)x(?:...|(...))\1x/.test("xabcABCx"), "backref-ASCII-backtrack");
assertFalse(/(?i)f/.test('b'));
assertFalse(/(?i)[abc]f/.test('x'));
assertFalse(/(?i)[abc]f/.test('xa'));
assertFalse(/(?i)[abc]</.test('x'));
assertFalse(/(?i)[abc]</.test('xa'));
assertFalse(/(?i)f[abc]/.test('x'));
assertFalse(/(?i)f[abc]/.test('xa'));
assertFalse(/(?i)<[abc]/.test('x'));
assertFalse(/(?i)<[abc]/.test('xa'));
assertFalse(/(?i)[\u00e5]/.test("\u212b"));
assertFalse(/(?i)[\u212b]/.test("\u00e5\u1234"));
assertFalse(/(?i)[\u212b]/.test("\u00e5"));
assertFalse(/(?i)\u{10400}/.test("\u{10428}"));
assertFalse(/(?i)[\u00e5]/.test("\u212b"));
assertFalse(/(?i)[\u212b]/.test("\u00e5\u1234"));
assertFalse(/(?i)[\u212b]/.test("\u00e5"));
assertFalse(/(?i)\u{10400}/.test("\u{10428}"));
assertTrue(/(?i)[@-A]/.test("a"));
assertTrue(/(?i)[@-A]/.test("A"));
assertTrue(/(?i)[@-A]/.test("@"));
assertFalse(/(?i)[¿-À]/.test('¾'));
assertTrue(/(?i)[¿-À]/.test('¿'));
assertTrue(/(?i)[¿-À]/.test('À'));
assertTrue(/(?i)[¿-À]/.test('à'));
assertFalse(/(?i)[¿-À]/.test('á'));
assertFalse(/(?i)[¿-À]/.test('Á'));
assertFalse(/(?i)[¿-À]/.test('Á'));
assertFalse(/(?i)[Ö-×]/.test('Õ'));
assertTrue(/(?i)[Ö-×]/.test('Ö'));
assertTrue(/(?i)[Ö-×]/.test('ö'));
assertTrue(/(?i)[Ö-×]/.test('×'));
assertFalse(/(?i)[Ö-×]/.test('Ø'));
assertTrue(/(?i)(a[\u1000A])+/.test('aa'));
assertTrue(/(?i)\u0178/.test('\u00ff'));
assertTrue(/(?i)\u039c/.test('\u00b5'));
assertTrue(/(?i)\u039c/.test('\u03bc'));
assertTrue(/(?i)\u00b5/.test('\u03bc'));
assertTrue(/(?i)[\u039b-\u039d]/.test('\u00b5'));
assertFalse(/(?i)[^\u039b-\u039d]/.test('\u00b5'));
assertTrue(/(?m)^bar/.test("bar"));
assertTrue(/(?m)^bar/.test("bar\nfoo"));
assertTrue(/(?m)^bar/.test("foo\nbar"));
assertTrue(/(?m)bar$/.test("bar"));
assertTrue(/(?m)bar$/.test("bar\nfoo"));
assertTrue(/(?m)bar$/.test("foo\nbar"));
assertFalse(/(?m)^bxr/.test("bar"));
assertFalse(/(?m)^bxr/.test("bar\nfoo"));
assertFalse(/(?m)^bxr/.test("foo\nbar"));
assertFalse(/(?m)bxr$/.test("bar"));
assertFalse(/(?m)bxr$/.test("bar\nfoo"));
assertFalse(/(?m)bxr$/.test("foo\nbar"));
assertTrue(/(?m)^.*$/.test("\n"));
assertTrue(/(?m)^([()]|.)*$/.test("()\n()"));
assertTrue(/(?m)^([()]|.)*$/.test("()\n"));
assertTrue(/(?m)^[()]*$/.test("()\n."));

27
test/mjsunit/regexp-modifiers-dotall.js Normal file
View File

@ -0,0 +1,27 @@
// Copyright 2017 the V8 project authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
// Flags: --regexp-mode-modifiers --harmony-regexp-dotall
// S flag switches dotall mode on and off. Combine with i flag changes to test
// the parser.
test(/.(?s).(?i-s).a(?-i)a/);
test(/.(?s:.)(?i:.a)a/);
test(/.(?s).(?i-s).a(?-i)a/u);
test(/.(?s:.)(?i:.a)a/u);
// m flag makes no difference
test(/.(?sm).(?i-s).a(?-i)a/);
test(/.(?s:.)(?i:.a)a/);
test(/.(?sm).(?im-s).a(?m-i)a/u);
test(/.(?s:.)(?i:.a)a/u);
function test(re) {
assertTrue(re.test("...aa"));
assertTrue(re.test(".\n.aa"));
assertTrue(re.test(".\n.Aa"));
assertFalse(re.test("\n\n.Aa"));
assertFalse(re.test(".\n\nAa"));
assertFalse(re.test(".\n.AA"));
}

138
test/mjsunit/regexp-modifiers-i18n.js Normal file
View File

@ -0,0 +1,138 @@
// Copyright 2017 the V8 project authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
// Flags: --regexp-mode-modifiers
// These tests won't all run on the noi18n build of V8.
aa(/(a)(?i)\1/u);
aa(/([az])(?i)\1/u);
function aa(re) {
assertTrue(re.test("aa"));
assertTrue(re.test("aA"));
assertFalse(re.test("Aa"));
assertFalse(re.test("AA"));
}
aai(/(a)(?-i)\1/iu);
aai(/([az])(?-i)\1/iu);
function aai(re) {
assertTrue(re.test("aa"));
assertFalse(re.test("aA"));
assertFalse(re.test("Aa"));
assertTrue(re.test("AA"));
}
abcd(/a(b(?i)c)d/u);
abcd(/[aw]([bx](?i)[cy])[dz]/u);
function abcd(re) {
assertTrue(re.test("abcd"));
assertFalse(re.test("abcD"));
assertTrue(re.test("abCd"));
assertFalse(re.test("abCD"));
assertFalse(re.test("aBcd"));
assertFalse(re.test("aBcD"));
assertFalse(re.test("aBCd"));
assertFalse(re.test("aBCD"));
assertFalse(re.test("Abcd"));
assertFalse(re.test("AbcD"));
assertFalse(re.test("AbCd"));
assertFalse(re.test("AbCD"));
assertFalse(re.test("ABcd"));
assertFalse(re.test("ABcD"));
assertFalse(re.test("ABCd"));
assertFalse(re.test("ABCD"));
}
abcdei(/a(b(?-i)c)d/iu);
abcdei(/[aw]([bx](?-i)[cy])[dz]/iu);
function abcdei(re) {
assertTrue(re.test("abcd"));
assertTrue(re.test("abcD"));
assertFalse(re.test("abCd"));
assertFalse(re.test("abCD"));
assertTrue(re.test("aBcd"));
assertTrue(re.test("aBcD"));
assertFalse(re.test("aBCd"));
assertFalse(re.test("aBCD"));
assertTrue(re.test("Abcd"));
assertTrue(re.test("AbcD"));
assertFalse(re.test("AbCd"));
assertFalse(re.test("AbCD"));
assertTrue(re.test("ABcd"));
assertTrue(re.test("ABcD"));
assertFalse(re.test("ABCd"));
assertFalse(re.test("ABCD"));
}
abc(/a(?i:b)c/u);
abc(/[ax](?i:[by])[cz]/u);
function abc(re) {
assertTrue(re.test("abc"));
assertFalse(re.test("abC"));
assertTrue(re.test("aBc"));
assertFalse(re.test("aBC"));
assertFalse(re.test("Abc"));
assertFalse(re.test("AbC"));
assertFalse(re.test("ABc"));
assertFalse(re.test("ABC"));
}
abci(/a(?-i:b)c/iu);
abci(/[ax](?-i:[by])[cz]/iu);
function abci(re) {
assertTrue(re.test("abc"));
assertTrue(re.test("abC"));
assertFalse(re.test("aBc"));
assertFalse(re.test("aBC"));
assertTrue(re.test("Abc"));
assertTrue(re.test("AbC"));
assertFalse(re.test("ABc"));
assertFalse(re.test("ABC"));
}
// The following tests are taken from test/mjsunit/es7/regexp-ui-word.js but
// using inline syntax instead of the global /i flag.
assertTrue(/(?i)\w/u.test('\u017F'));
assertTrue(/(?i)\w/u.test('\u212A'));
assertFalse(/(?i)\W/u.test('\u017F'));
assertFalse(/(?i)\W/u.test('\u212A'));
assertFalse(/(?i)\W/u.test('s'));
assertFalse(/(?i)\W/u.test('S'));
assertFalse(/(?i)\W/u.test('K'));
assertFalse(/(?i)\W/u.test('k'));
assertTrue(/(?i)[\w]/u.test('\u017F'));
assertTrue(/(?i)[\w]/u.test('\u212A'));
assertFalse(/(?i)[\W]/u.test('\u017F'));
assertFalse(/(?i)[\W]/u.test('\u212A'));
assertFalse(/(?i)[\W]/u.test('s'));
assertFalse(/(?i)[\W]/u.test('S'));
assertFalse(/(?i)[\W]/u.test('K'));
assertFalse(/(?i)[\W]/u.test('k'));
assertTrue(/(?i)\b/u.test('\u017F'));
assertFalse(/(?i:)\b/u.test('\u017F'));
assertTrue(/(?i)\b/u.test('\u212A'));
assertFalse(/(?i:)\b/u.test('\u212A'));
assertTrue(/(?i)\b/u.test('s'));
assertTrue(/(?i)\b/u.test('S'));
assertFalse(/(?i)\B/u.test('\u017F'));
assertFalse(/(?i)\B/u.test('\u212A'));
assertFalse(/(?i)\B/u.test('s'));
assertFalse(/(?i)\B/u.test('S'));
assertFalse(/(?i)\B/u.test('K'));
assertFalse(/(?i)\B/u.test('k'));
assertEquals(["abcd\u017F", "\u017F"], /a.*?(.)(?i)\b/u.exec('abcd\u017F cd'));
assertEquals(["abcd\u212A", "\u212A"], /a.*?(.)(?i)\b/u.exec('abcd\u212A cd'));
assertEquals(["a\u017F", "\u017F"], /a.*?(?i:\B)(.)/u.exec('a\u017F '));
assertEquals(["a\u212A", "\u212A"], /a.*?(?i:\B)(.)/u.exec('a\u212A '));

146
test/mjsunit/regexp-modifiers.js Normal file
View File

@ -0,0 +1,146 @@
// Copyright 2017 the V8 project authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
// Flags: --regexp-mode-modifiers
aa(/(a)(?i)\1/);
aa(/([az])(?i)\1/);
function aa(re) {
assertTrue(re.test("aa"));
assertTrue(re.test("aA"));
assertFalse(re.test("Aa"));
assertFalse(re.test("AA"));
}
aai(/(a)(?-i)\1/i);
aai(/([az])(?-i)\1/i);
function aai(re) {
assertTrue(re.test("aa"));
assertFalse(re.test("aA"));
assertFalse(re.test("Aa"));
assertTrue(re.test("AA"));
}
abcd(/a(b(?i)c)d/);
abcd(/[aw]([bx](?i)[cy])[dz]/);
function abcd(re) {
assertTrue(re.test("abcd"));
assertFalse(re.test("abcD"));
assertTrue(re.test("abCd"));
assertFalse(re.test("abCD"));
assertFalse(re.test("aBcd"));
assertFalse(re.test("aBcD"));
assertFalse(re.test("aBCd"));
assertFalse(re.test("aBCD"));
assertFalse(re.test("Abcd"));
assertFalse(re.test("AbcD"));
assertFalse(re.test("AbCd"));
assertFalse(re.test("AbCD"));
assertFalse(re.test("ABcd"));
assertFalse(re.test("ABcD"));
assertFalse(re.test("ABCd"));
assertFalse(re.test("ABCD"));
}
abcdei(/a(b(?-i)c)d/i);
abcdei(/[aw]([bx](?-i)[cy])[dz]/i);
function abcdei(re) {
assertTrue(re.test("abcd"));
assertTrue(re.test("abcD"));
assertFalse(re.test("abCd"));
assertFalse(re.test("abCD"));
assertTrue(re.test("aBcd"));
assertTrue(re.test("aBcD"));
assertFalse(re.test("aBCd"));
assertFalse(re.test("aBCD"));
assertTrue(re.test("Abcd"));
assertTrue(re.test("AbcD"));
assertFalse(re.test("AbCd"));
assertFalse(re.test("AbCD"));
assertTrue(re.test("ABcd"));
assertTrue(re.test("ABcD"));
assertFalse(re.test("ABCd"));
assertFalse(re.test("ABCD"));
}
abc(/a(?i:b)c/);
abc(/[ax](?i:[by])[cz]/);
function abc(re) {
assertTrue(re.test("abc"));
assertFalse(re.test("abC"));
assertTrue(re.test("aBc"));
assertFalse(re.test("aBC"));
assertFalse(re.test("Abc"));
assertFalse(re.test("AbC"));
assertFalse(re.test("ABc"));
assertFalse(re.test("ABC"));
}
abci(/a(?-i:b)c/i);
abci(/[ax](?-i:[by])[cz]/i);
function abci(re) {
assertTrue(re.test("abc"));
assertTrue(re.test("abC"));
assertFalse(re.test("aBc"));
assertFalse(re.test("aBC"));
assertTrue(re.test("Abc"));
assertTrue(re.test("AbC"));
assertFalse(re.test("ABc"));
assertFalse(re.test("ABC"));
}
assertThrows(() => new RegExp("foo(?i:"));
assertThrows(() => new RegExp("foo(?--i)"));
assertThrows(() => new RegExp("foo(?i-i)"));
assertThrows(() => new RegExp("foo(?m:"));
assertThrows(() => new RegExp("foo(?--m)"));
assertThrows(() => new RegExp("foo(?m-m)"));
var re = /^\s(?m)^.$\s(?-m)$/;
assertTrue(re.test("\n.\n"));
assertFalse(re.test(" .\n"));
assertFalse(re.test("\n. "));
assertFalse(re.test(" . "));
assertFalse(re.test("_\n.\n"));
assertFalse(re.test("\n.\n_"));
assertFalse(re.test("_\n.\n_"));
assertEquals(["abcd", "d"], /a.*?(.)(?i)\b/.exec('abcd\u017F cd'));
assertEquals(["abcd", "d"], /a.*?(.)(?i)\b/.exec('abcd\u212A cd'));
assertEquals(["a\u017F ", " "], /a.*?(?i)\B(.)/.exec('a\u017F '));
assertEquals(["a\u212A ", " "], /a.*?(?i)\B(.)/.exec('a\u212A '));
// Nested flags.
var res = [
/^a(?i:b(?-i:c(?i:d)e)f)g$/,
/^a(?i:b(?-i)c(?i)d(?-i)e(?i)f)g$/,
/^(?-i:a(?i:b(?-i:c(?i:d)e)f)g)$/i,
/^(?-i:a(?i:b(?-i)c(?i)d(?-i)e(?i)f)g)$/i,
];
for (var idx = 0; idx < res.length; idx++) {
var re = res[idx];
for (var i = 0; i < 128; i++) {
var s = (i & 1) ? "A" : "a";
s += (i & 2) ? "B" : "b";
s += (i & 4) ? "C" : "c";
s += (i & 8) ? "D" : "d";
s += (i & 16) ? "E" : "e";
s += (i & 32) ? "F" : "f";
s += (i & 64) ? "G" : "g";
if ((i & (1 | 4 | 16 | 64)) != 0) {
assertFalse(re.test(s), s);
} else {
assertTrue(re.test(s), s);
}
}
}