diff options
Diffstat (limited to 'js/src/irregexp/imported/regexp-parser.cc')
| -rw-r--r-- | js/src/irregexp/imported/regexp-parser.cc | 393 |
1 files changed, 263 insertions, 130 deletions
diff --git a/js/src/irregexp/imported/regexp-parser.cc b/js/src/irregexp/imported/regexp-parser.cc index ea2a6c6d7a..965fc567b7 100644 --- a/js/src/irregexp/imported/regexp-parser.cc +++ b/js/src/irregexp/imported/regexp-parser.cc @@ -13,7 +13,7 @@ #include "unicode/unistr.h" #include "unicode/usetiter.h" #include "unicode/utf16.h" // For U16_NEXT -#endif // V8_INTL_SUPPORT +#endif // V8_INTL_SUPPORT namespace v8 { namespace internal { @@ -67,8 +67,7 @@ class RegExpTextBuilder { bool ignore_case() const { return IsIgnoreCase(flags_); } bool IsUnicodeMode() const { // Either /v or /u enable UnicodeMode - // TODO(v8:11935): Change permalink once proposal is in stage 4. - // https://arai-a.github.io/ecma262-compare/snapshot.html?pr=2418#sec-parsepattern + // https://tc39.es/ecma262/#sec-parsepattern return IsUnicode(flags_) || IsUnicodeSets(flags_); } Zone* zone() const { return zone_; } @@ -264,7 +263,7 @@ RegExpTree* RegExpTextBuilder::PopLastAtom() { characters_ = nullptr; atom = zone()->New<RegExpAtom>(char_vector); return atom; - } else if (text_.size() > 0) { + } else if (!text_.empty()) { atom = text_.back(); text_.pop_back(); return atom; @@ -315,8 +314,7 @@ class RegExpBuilder { void FlushTerms(); bool IsUnicodeMode() const { // Either /v or /u enable UnicodeMode - // TODO(v8:11935): Change permalink once proposal is in stage 4. - // https://arai-a.github.io/ecma262-compare/snapshot.html?pr=2418#sec-parsepattern + // https://tc39.es/ecma262/#sec-parsepattern return IsUnicode(flags_) || IsUnicodeSets(flags_); } Zone* zone() const { return zone_; } @@ -354,7 +352,12 @@ class RegExpParserState : public ZoneObject { group_type_(group_type), lookaround_type_(lookaround_type), disjunction_capture_index_(disjunction_capture_index), - capture_name_(capture_name) {} + capture_name_(capture_name) { + if (previous_state != nullptr) { + non_participating_capture_group_interval_ = + previous_state->non_participating_capture_group_interval(); + } + } // Parser state of containing expression, if any. RegExpParserState* previous_state() const { return previous_state_; } bool IsSubexpression() { return previous_state_ != nullptr; } @@ -371,6 +374,9 @@ class RegExpParserState : public ZoneObject { // The name of the current sub-expression, if group_type is CAPTURE. Only // used for named captures. const ZoneVector<base::uc16>* capture_name() const { return capture_name_; } + std::pair<int, int> non_participating_capture_group_interval() const { + return non_participating_capture_group_interval_; + } bool IsNamedCapture() const { return capture_name_ != nullptr; } @@ -398,6 +404,18 @@ class RegExpParserState : public ZoneObject { return false; } + void NewAlternative(int captures_started) { + if (non_participating_capture_group_interval().second != 0) { + // Extend the non-participating interval. + non_participating_capture_group_interval_.second = captures_started; + } else { + // Create new non-participating interval from the start of the current + // enclosing group to all captures created within that group so far. + non_participating_capture_group_interval_ = + std::make_pair(capture_index(), captures_started); + } + } + private: // Linked list implementation of stack of states. RegExpParserState* const previous_state_; @@ -411,6 +429,11 @@ class RegExpParserState : public ZoneObject { const int disjunction_capture_index_; // Stored capture name (if any). const ZoneVector<base::uc16>* const capture_name_; + // Interval of (named) capture indices ]from, to] that are not participating + // in the current state (i.e. they cannot match). + // Capture indices are not participating if they were created in a different + // alternative. + std::pair<int, int> non_participating_capture_group_interval_; }; template <class CharT> @@ -463,17 +486,22 @@ class RegExpParserImpl final { RegExpTree* ParseClassSetOperand(const RegExpBuilder* builder, ClassSetOperandType* type_out, ZoneList<CharacterRange>* ranges, - CharacterClassStrings* strings); + CharacterClassStrings* strings, + base::uc32* character); base::uc32 ParseClassSetCharacter(); // Parses and returns a single escaped character. base::uc32 ParseCharacterEscape(InClassEscapeState in_class_escape_state, bool* is_escaped_unicode_character); + void AddMaybeSimpleCaseFoldedRange(ZoneList<CharacterRange>* ranges, + CharacterRange new_range); + RegExpTree* ParseClassUnion(const RegExpBuilder* builder, bool is_negated, RegExpTree* first_operand, ClassSetOperandType first_operand_type, ZoneList<CharacterRange>* ranges, - CharacterClassStrings* strings); + CharacterClassStrings* strings, + base::uc32 first_character); RegExpTree* ParseClassIntersection(const RegExpBuilder* builder, bool is_negated, RegExpTree* first_operand, ClassSetOperandType first_operand_type); @@ -504,11 +532,10 @@ class RegExpParserImpl final { int captures_started() const { return captures_started_; } int position() const { return next_pos_ - 1; } bool failed() const { return failed_; } - RegExpFlags flags() const { return top_level_flags_; } + RegExpFlags flags() const { return flags_; } bool IsUnicodeMode() const { // Either /v or /u enable UnicodeMode - // TODO(v8:11935): Change permalink once proposal is in stage 4. - // https://arai-a.github.io/ecma262-compare/snapshot.html?pr=2418#sec-parsepattern + // https://tc39.es/ecma262/#sec-parsepattern return IsUnicode(flags()) || IsUnicodeSets(flags()) || force_unicode_; } bool unicode_sets() const { return IsUnicodeSets(flags()); } @@ -528,7 +555,7 @@ class RegExpParserImpl final { // Creates a new named capture at the specified index. Must be called exactly // once for each named capture. Fails if a capture with the same name is // encountered. - bool CreateNamedCaptureAtIndex(const ZoneVector<base::uc16>* name, int index); + bool CreateNamedCaptureAtIndex(const RegExpParserState* state, int index); // Parses the name of a capture group (?<name>pattern). The name must adhere // to IdentifierName in the ECMAScript standard. @@ -543,7 +570,7 @@ class RegExpParserImpl final { // to avoid complicating cases in which references comes before the capture. void PatchNamedBackReferences(); - ZoneVector<RegExpCapture*>* GetNamedCaptures() const; + ZoneVector<RegExpCapture*>* GetNamedCaptures(); // Returns true iff the pattern contains named captures. May call // ScanForCaptures to look ahead at the remaining pattern. @@ -593,16 +620,20 @@ class RegExpParserImpl final { RegExpError error_ = RegExpError::kNone; int error_pos_ = 0; ZoneList<RegExpCapture*>* captures_; - ZoneSet<RegExpCapture*, RegExpCaptureNameLess>* named_captures_; + // Maps capture names to a list of capture indices with this name. + ZoneMap<RegExpCapture*, ZoneList<int>*, RegExpCaptureNameLess>* + named_captures_; ZoneList<RegExpBackReference*>* named_back_references_; + ZoneList<CharacterRange>* temp_ranges_; const CharT* const input_; const int input_length_; base::uc32 current_; - const RegExpFlags top_level_flags_; + RegExpFlags flags_; bool force_unicode_ = false; // Force parser to act as if unicode were set. int next_pos_; int captures_started_; int capture_count_; // Only valid after we have scanned for captures. + int lookaround_count_; // Only valid after we have scanned for lookbehinds. bool has_more_; bool simple_; bool contains_anchor_; @@ -625,10 +656,11 @@ RegExpParserImpl<CharT>::RegExpParserImpl( input_(input), input_length_(input_length), current_(kEndMarker), - top_level_flags_(flags), + flags_(flags), next_pos_(0), captures_started_(0), capture_count_(0), + lookaround_count_(0), has_more_(true), simple_(false), contains_anchor_(false), @@ -909,21 +941,21 @@ RegExpTree* RegExpParserImpl<CharT>::ParseDisjunction() { // Build result of subexpression. if (group_type == CAPTURE) { if (state->IsNamedCapture()) { - CreateNamedCaptureAtIndex(state->capture_name(), - capture_index CHECK_FAILED); + CreateNamedCaptureAtIndex(state, capture_index CHECK_FAILED); } RegExpCapture* capture = GetCapture(capture_index); capture->set_body(body); body = capture; } else if (group_type == GROUPING) { - body = zone()->template New<RegExpGroup>(body); + body = zone()->template New<RegExpGroup>(body, builder->flags()); } else { DCHECK(group_type == POSITIVE_LOOKAROUND || group_type == NEGATIVE_LOOKAROUND); bool is_positive = (group_type == POSITIVE_LOOKAROUND); body = zone()->template New<RegExpLookaround>( body, is_positive, end_capture_index - capture_index, - capture_index, state->lookaround_type()); + capture_index, state->lookaround_type(), lookaround_count_); + lookaround_count_++; } // Restore previous state. @@ -937,6 +969,7 @@ RegExpTree* RegExpParserImpl<CharT>::ParseDisjunction() { } case '|': { Advance(); + state->NewAlternative(captures_started()); builder->NewAlternative(); continue; } @@ -984,6 +1017,7 @@ RegExpTree* RegExpParserImpl<CharT>::ParseDisjunction() { case '(': { state = ParseOpenParenthesis(state CHECK_FAILED); builder = state->builder(); + flags_ = builder->flags(); continue; } case '[': { @@ -1037,8 +1071,8 @@ RegExpTree* RegExpParserImpl<CharT>::ParseDisjunction() { builder->AddEmpty(); } else { RegExpCapture* capture = GetCapture(index); - RegExpTree* atom = zone()->template New<RegExpBackReference>( - capture, builder->flags()); + RegExpTree* atom = + zone()->template New<RegExpBackReference>(capture, zone()); builder->AddAtom(atom); } break; @@ -1246,43 +1280,91 @@ RegExpParserState* RegExpParserImpl<CharT>::ParseOpenParenthesis( bool is_named_capture = false; const ZoneVector<base::uc16>* capture_name = nullptr; SubexpressionType subexpr_type = CAPTURE; + RegExpFlags flags = state->builder()->flags(); + bool parsing_modifiers = false; + bool modifiers_polarity = true; + RegExpFlags modifiers; 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 '<': - Advance(); - if (Next() == '=') { + do { + switch (Next()) { + case '-': + if (!v8_flags.js_regexp_modifiers) { + ReportError(RegExpError::kInvalidGroup); + return nullptr; + } + Advance(); + parsing_modifiers = true; + if (modifiers_polarity == false) { + ReportError(RegExpError::kMultipleFlagDashes); + return nullptr; + } + modifiers_polarity = false; + break; + case 'm': + case 'i': + case 's': { + if (!v8_flags.js_regexp_modifiers) { + ReportError(RegExpError::kInvalidGroup); + return nullptr; + } + Advance(); + parsing_modifiers = true; + RegExpFlag flag = TryRegExpFlagFromChar(current()).value(); + if ((modifiers & flag) != 0) { + ReportError(RegExpError::kRepeatedFlag); + return nullptr; + } + modifiers |= flag; + flags.set(flag, modifiers_polarity); + break; + } + case ':': + Advance(2); + parsing_modifiers = false; + subexpr_type = GROUPING; + break; + case '=': Advance(2); - lookaround_type = RegExpLookaround::LOOKBEHIND; + parsing_modifiers = false; + lookaround_type = RegExpLookaround::LOOKAHEAD; subexpr_type = POSITIVE_LOOKAROUND; break; - } else if (Next() == '!') { + case '!': Advance(2); - lookaround_type = RegExpLookaround::LOOKBEHIND; + parsing_modifiers = false; + lookaround_type = RegExpLookaround::LOOKAHEAD; subexpr_type = NEGATIVE_LOOKAROUND; break; - } - is_named_capture = true; - has_named_captures_ = true; - Advance(); - break; - default: - ReportError(RegExpError::kInvalidGroup); - return nullptr; + case '<': + Advance(); + parsing_modifiers = false; + 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; + } + is_named_capture = true; + has_named_captures_ = true; + Advance(); + break; + default: + ReportError(RegExpError::kInvalidGroup); + return nullptr; + } + } while (parsing_modifiers); + } + if (modifiers_polarity == false) { + // We encountered a dash. + if (modifiers == 0) { + ReportError(RegExpError::kInvalidFlagGroup); + return nullptr; } } if (subexpr_type == CAPTURE) { @@ -1299,7 +1381,7 @@ RegExpParserState* RegExpParserImpl<CharT>::ParseOpenParenthesis( // Store current state and begin new disjunction parsing. return zone()->template New<RegExpParserState>( state, subexpr_type, lookaround_type, captures_started_, capture_name, - state->builder()->flags(), zone()); + flags, zone()); } // In order to know whether an escape is a backreference or not we have to scan @@ -1511,7 +1593,10 @@ const ZoneVector<base::uc16>* RegExpParserImpl<CharT>::ParseCaptureGroupName() { template <class CharT> bool RegExpParserImpl<CharT>::CreateNamedCaptureAtIndex( - const ZoneVector<base::uc16>* name, int index) { + const RegExpParserState* state, int index) { + const ZoneVector<base::uc16>* name = state->capture_name(); + const std::pair<int, int> non_participating_capture_group_interval = + state->non_participating_capture_group_interval(); DCHECK(0 < index && index <= captures_started_); DCHECK_NOT_NULL(name); @@ -1521,21 +1606,33 @@ bool RegExpParserImpl<CharT>::CreateNamedCaptureAtIndex( capture->set_name(name); if (named_captures_ == nullptr) { - named_captures_ = - zone_->template New<ZoneSet<RegExpCapture*, RegExpCaptureNameLess>>( - zone()); + named_captures_ = zone_->template New< + ZoneMap<RegExpCapture*, ZoneList<int>*, RegExpCaptureNameLess>>(zone()); } else { // Check for duplicates and bail if we find any. - const auto& named_capture_it = named_captures_->find(capture); if (named_capture_it != named_captures_->end()) { - ReportError(RegExpError::kDuplicateCaptureGroupName); - return false; + if (v8_flags.js_regexp_duplicate_named_groups) { + ZoneList<int>* named_capture_indices = named_capture_it->second; + DCHECK_NOT_NULL(named_capture_indices); + DCHECK(!named_capture_indices->is_empty()); + for (int named_index : *named_capture_indices) { + if (named_index < non_participating_capture_group_interval.first || + named_index > non_participating_capture_group_interval.second) { + ReportError(RegExpError::kDuplicateCaptureGroupName); + return false; + } + } + } else { + ReportError(RegExpError::kDuplicateCaptureGroupName); + return false; + } } } - named_captures_->emplace(capture); - + auto entry = named_captures_->try_emplace( + capture, zone()->template New<ZoneList<int>>(1, zone())); + entry.first->second->Add(index, zone()); return true; } @@ -1558,7 +1655,7 @@ bool RegExpParserImpl<CharT>::ParseNamedBackReference( builder->AddEmpty(); } else { RegExpBackReference* atom = - zone()->template New<RegExpBackReference>(builder->flags()); + zone()->template New<RegExpBackReference>(zone()); atom->set_name(name); builder->AddAtom(atom); @@ -1595,16 +1692,17 @@ void RegExpParserImpl<CharT>::PatchNamedBackReferences() { DCHECK_NULL(search_capture->name()); search_capture->set_name(ref->name()); - int index = -1; const auto& capture_it = named_captures_->find(search_capture); - if (capture_it != named_captures_->end()) { - index = (*capture_it)->index(); - } else { + if (capture_it == named_captures_->end()) { ReportError(RegExpError::kInvalidNamedCaptureReference); return; } - ref->set_capture(GetCapture(index)); + DCHECK_IMPLIES(!v8_flags.js_regexp_duplicate_named_groups, + capture_it->second->length() == 1); + for (int index : *capture_it->second) { + ref->add_capture(GetCapture(index), zone()); + } } } @@ -1627,13 +1725,22 @@ RegExpCapture* RegExpParserImpl<CharT>::GetCapture(int index) { } template <class CharT> -ZoneVector<RegExpCapture*>* RegExpParserImpl<CharT>::GetNamedCaptures() const { - if (named_captures_ == nullptr || named_captures_->empty()) { +ZoneVector<RegExpCapture*>* RegExpParserImpl<CharT>::GetNamedCaptures() { + if (named_captures_ == nullptr) { return nullptr; } + DCHECK(!named_captures_->empty()); - return zone()->template New<ZoneVector<RegExpCapture*>>( - named_captures_->begin(), named_captures_->end(), zone()); + ZoneVector<RegExpCapture*>* flattened_named_captures = + zone()->template New<ZoneVector<RegExpCapture*>>(zone()); + for (auto capture : *named_captures_) { + DCHECK_IMPLIES(!v8_flags.js_regexp_duplicate_named_groups, + capture.second->length() == 1); + for (int index : *capture.second) { + flattened_named_captures->push_back(GetCapture(index)); + } + } + return flattened_named_captures; } template <class CharT> @@ -1890,7 +1997,7 @@ bool LookupPropertyValueName(UProperty property, ExtractStringsFromUnicodeSet(set, result_strings, flags, zone); } const bool needs_case_folding = IsUnicodeSets(flags) && IsIgnoreCase(flags); - if (needs_case_folding) CharacterRange::UnicodeSimpleCloseOver(set); + if (needs_case_folding) set.closeOver(USET_SIMPLE_CASE_INSENSITIVE); set.removeAllStrings(); if (negate) set.complement(); for (int i = 0; i < set.getRangeCount(); i++) { @@ -2096,13 +2203,22 @@ bool RegExpParserImpl<CharT>::AddPropertyClassRange( if (!IsSupportedBinaryProperty(property, unicode_sets())) return false; if (!IsExactPropertyAlias(name, property)) return false; // Negation of properties with strings is not allowed. - // TODO(v8:11935): Change permalink once proposal is in stage 4. // See - // https://arai-a.github.io/ecma262-compare/snapshot.html?pr=2418#sec-static-semantics-maycontainstrings + // https://tc39.es/ecma262/#sec-static-semantics-maycontainstrings if (negate && IsBinaryPropertyOfStrings(property)) return false; - return LookupPropertyValueName(property, negate ? "N" : "Y", false, - add_to_ranges, add_to_strings, flags(), - zone()); + if (unicode_sets()) { + // In /v mode we can't simple lookup the "false" binary property values, + // as the spec requires us to perform case folding before calculating the + // complement. + // See https://tc39.es/ecma262/#sec-compiletocharset + // UnicodePropertyValueExpression :: LoneUnicodePropertyNameOrValue + return LookupPropertyValueName(property, "Y", negate, add_to_ranges, + add_to_strings, flags(), zone()); + } else { + return LookupPropertyValueName(property, negate ? "N" : "Y", false, + add_to_ranges, add_to_strings, flags(), + zone()); + } } else { // Both property name and value name are specified. Attempt to interpret // the property name as enumerated property. @@ -2325,8 +2441,7 @@ base::uc32 RegExpParserImpl<CharT>::ParseCharacterEscape( return c; } -// TODO(v8:11935): Change permalink once proposal is in stage 4. -// https://arai-a.github.io/ecma262-compare/snapshot.html?pr=2418#prod-ClassRanges +// https://tc39.es/ecma262/#prod-ClassRanges template <class CharT> RegExpTree* RegExpParserImpl<CharT>::ParseClassRanges( ZoneList<CharacterRange>* ranges, bool add_unicode_case_equivalents) { @@ -2475,8 +2590,7 @@ void AddClassString(ZoneList<base::uc32>* normalized_string, } // namespace -// TODO(v8:11935): Change permalink once proposal is in stage 4. -// https://arai-a.github.io/ecma262-compare/snapshot.html?pr=2418#prod-ClassStringDisjunction +// https://tc39.es/ecma262/#prod-ClassStringDisjunction template <class CharT> RegExpTree* RegExpParserImpl<CharT>::ParseClassStringDisjunction( ZoneList<CharacterRange>* ranges, CharacterClassStrings* strings) { @@ -2526,8 +2640,7 @@ RegExpTree* RegExpParserImpl<CharT>::ParseClassStringDisjunction( return nullptr; } -// TODO(v8:11935): Change permalink once proposal is in stage 4. -// https://arai-a.github.io/ecma262-compare/snapshot.html?pr=2418#prod-ClassSetOperand +// https://tc39.es/ecma262/#prod-ClassSetOperand // Tree returned based on type_out: // * kNestedClass: RegExpClassSetExpression // * For all other types: RegExpClassSetOperand @@ -2538,12 +2651,13 @@ RegExpTree* RegExpParserImpl<CharT>::ParseClassSetOperand( zone()->template New<ZoneList<CharacterRange>>(1, zone()); CharacterClassStrings* strings = zone()->template New<CharacterClassStrings>(zone()); - RegExpTree* tree = - ParseClassSetOperand(builder, type_out, ranges, strings CHECK_FAILED); + base::uc32 character; + RegExpTree* tree = ParseClassSetOperand(builder, type_out, ranges, strings, + &character CHECK_FAILED); DCHECK_IMPLIES(*type_out != ClassSetOperandType::kNestedClass, tree == nullptr); DCHECK_IMPLIES(*type_out == ClassSetOperandType::kClassSetCharacter, - ranges->length() == 1); + ranges->is_empty()); DCHECK_IMPLIES(*type_out == ClassSetOperandType::kClassSetCharacter, strings->empty()); DCHECK_IMPLIES(*type_out == ClassSetOperandType::kNestedClass, @@ -2558,21 +2672,27 @@ RegExpTree* RegExpParserImpl<CharT>::ParseClassSetOperand( // CharacterClassEscape includes \p{}, which can contain ranges, strings or // both and \P{}, which could contain nothing (i.e. \P{Any}). if (tree == nullptr) { + if (*type_out == ClassSetOperandType::kClassSetCharacter) { + AddMaybeSimpleCaseFoldedRange(ranges, + CharacterRange::Singleton(character)); + } tree = zone()->template New<RegExpClassSetOperand>(ranges, strings); } return tree; } -// TODO(v8:11935): Change permalink once proposal is in stage 4. -// https://arai-a.github.io/ecma262-compare/snapshot.html?pr=2418#prod-ClassSetOperand -// Based on |type_out| either a tree is returned or ranges/strings modified. -// If a tree is returned, ranges/strings are not modified. -// If |type_out| is kNestedClass, a tree of type RegExpClassSetExpression is -// returned. For all other types, ranges is modified and nullptr is returned. +// https://tc39.es/ecma262/#prod-ClassSetOperand +// Based on |type_out| either a tree is returned or +// |ranges|/|strings|/|character| modified. If a tree is returned, +// ranges/strings are not modified. If |type_out| is kNestedClass, a tree of +// type RegExpClassSetExpression is returned. If | type_out| is +// kClassSetCharacter, |character| is set and nullptr returned. For all other +// types, |ranges|/|strings|/|character| is modified and nullptr is returned. template <class CharT> RegExpTree* RegExpParserImpl<CharT>::ParseClassSetOperand( const RegExpBuilder* builder, ClassSetOperandType* type_out, - ZoneList<CharacterRange>* ranges, CharacterClassStrings* strings) { + ZoneList<CharacterRange>* ranges, CharacterClassStrings* strings, + base::uc32* character) { DCHECK(unicode_sets()); base::uc32 c = current(); if (c == '\\') { @@ -2599,7 +2719,7 @@ RegExpTree* RegExpParserImpl<CharT>::ParseClassSetOperand( *type_out = ClassSetOperandType::kClassSetCharacter; c = ParseClassSetCharacter(CHECK_FAILED); - ranges->Add(CharacterRange::Singleton(c), zone()); + *character = c; return nullptr; } @@ -2653,13 +2773,28 @@ bool MayContainStrings(ClassSetOperandType type, RegExpTree* operand) { } // namespace -// TODO(v8:11935): Change permalink once proposal is in stage 4. -// https://arai-a.github.io/ecma262-compare/snapshot.html?pr=2418#prod-ClassUnion +template <class CharT> +void RegExpParserImpl<CharT>::AddMaybeSimpleCaseFoldedRange( + ZoneList<CharacterRange>* ranges, CharacterRange new_range) { + DCHECK(unicode_sets()); + if (ignore_case()) { + ZoneList<CharacterRange>* new_ranges = + zone()->template New<ZoneList<CharacterRange>>(2, zone()); + new_ranges->Add(new_range, zone()); + CharacterRange::AddUnicodeCaseEquivalents(new_ranges, zone()); + ranges->AddAll(*new_ranges, zone()); + } else { + ranges->Add(new_range, zone()); + } + CharacterRange::Canonicalize(ranges); +} + +// https://tc39.es/ecma262/#prod-ClassUnion template <class CharT> RegExpTree* RegExpParserImpl<CharT>::ParseClassUnion( const RegExpBuilder* builder, bool is_negated, RegExpTree* first_operand, ClassSetOperandType first_operand_type, ZoneList<CharacterRange>* ranges, - CharacterClassStrings* strings) { + CharacterClassStrings* strings, base::uc32 character) { DCHECK(unicode_sets()); ZoneList<RegExpTree*>* operands = zone()->template New<ZoneList<RegExpTree*>>(2, zone()); @@ -2673,7 +2808,6 @@ RegExpTree* RegExpParserImpl<CharT>::ParseClassUnion( operands->Add(first_operand, zone()); } ClassSetOperandType last_type = first_operand_type; - const bool needs_case_folding = ignore_case(); while (has_more() && current() != ']') { if (current() == '-') { // Mix of ClassSetRange and ClassSubtraction is not allowed. @@ -2690,42 +2824,36 @@ RegExpTree* RegExpParserImpl<CharT>::ParseClassUnion( // represent a character range. // In case one of them is not a ClassSetCharacter, it is a syntax error, // as '-' can not be used unescaped within a class with /v. - // TODO(v8:11935): Change permalink once proposal is in stage 4. // See - // https://arai-a.github.io/ecma262-compare/snapshot.html?pr=2418#prod-ClassSetRange + // https://tc39.es/ecma262/#prod-ClassSetRange if (last_type != ClassSetOperandType::kClassSetCharacter) { return ReportError(RegExpError::kInvalidCharacterClass); } - ParseClassSetOperand(builder, &last_type, ranges, strings CHECK_FAILED); + base::uc32 from = character; + ParseClassSetOperand(builder, &last_type, ranges, strings, + &character CHECK_FAILED); if (last_type != ClassSetOperandType::kClassSetCharacter) { return ReportError(RegExpError::kInvalidCharacterClass); } - // Remove the last two singleton characters added to ranges, and combine - // them into a range. - auto rhs_ranges = ranges->RemoveLast(); - auto lhs_ranges = ranges->RemoveLast(); - DCHECK(lhs_ranges.IsSingleton()); - DCHECK(rhs_ranges.IsSingleton()); - base::uc32 from = lhs_ranges.from(); - base::uc32 to = rhs_ranges.from(); - if (from > to) { + if (from > character) { return ReportError(RegExpError::kOutOfOrderCharacterClass); } - ranges->Add(CharacterRange::Range(from, to), zone()); + AddMaybeSimpleCaseFoldedRange(ranges, + CharacterRange::Range(from, character)); last_type = ClassSetOperandType::kClassSetRange; } else { DCHECK_NE(current(), '-'); - RegExpTree* operand = ParseClassSetOperand(builder, &last_type, ranges, - strings CHECK_FAILED); + if (last_type == ClassSetOperandType::kClassSetCharacter) { + AddMaybeSimpleCaseFoldedRange(ranges, + CharacterRange::Singleton(character)); + } + RegExpTree* operand = ParseClassSetOperand( + builder, &last_type, ranges, strings, &character CHECK_FAILED); if (operand != nullptr) { may_contain_strings |= MayContainStrings(last_type, operand); // Add the range we started building as operand and reset the current // range. if (!ranges->is_empty() || !strings->empty()) { - if (needs_case_folding) { - CharacterRange::Canonicalize(ranges); - CharacterRange::AddUnicodeCaseEquivalents(ranges, zone()); - } may_contain_strings |= !strings->empty(); operands->Add( zone()->template New<RegExpClassSetOperand>(ranges, strings), @@ -2742,12 +2870,12 @@ RegExpTree* RegExpParserImpl<CharT>::ParseClassUnion( return ReportError(RegExpError::kUnterminatedCharacterClass); } + if (last_type == ClassSetOperandType::kClassSetCharacter) { + AddMaybeSimpleCaseFoldedRange(ranges, CharacterRange::Singleton(character)); + } + // Add the range we started building as operand. if (!ranges->is_empty() || !strings->empty()) { - if (needs_case_folding) { - CharacterRange::Canonicalize(ranges); - CharacterRange::AddUnicodeCaseEquivalents(ranges, zone()); - } may_contain_strings |= !strings->empty(); operands->Add(zone()->template New<RegExpClassSetOperand>(ranges, strings), zone()); @@ -2773,8 +2901,7 @@ RegExpTree* RegExpParserImpl<CharT>::ParseClassUnion( may_contain_strings, operands); } -// TODO(v8:11935): Change permalink once proposal is in stage 4. -// https://arai-a.github.io/ecma262-compare/snapshot.html?pr=2418#prod-ClassIntersection +// https://tc39.es/ecma262/#prod-ClassIntersection template <class CharT> RegExpTree* RegExpParserImpl<CharT>::ParseClassIntersection( const RegExpBuilder* builder, bool is_negated, RegExpTree* first_operand, @@ -2815,8 +2942,7 @@ RegExpTree* RegExpParserImpl<CharT>::ParseClassIntersection( may_contain_strings, operands); } -// TODO(v8:11935): Change permalink once proposal is in stage 4. -// https://arai-a.github.io/ecma262-compare/snapshot.html?pr=2418#prod-ClassSubtraction +// https://tc39.es/ecma262/#prod-ClassSubtraction template <class CharT> RegExpTree* RegExpParserImpl<CharT>::ParseClassSubtraction( const RegExpBuilder* builder, bool is_negated, RegExpTree* first_operand, @@ -2891,12 +3017,16 @@ RegExpTree* RegExpParserImpl<CharT>::ParseCharacterClass( ClassSetOperandType operand_type; CharacterClassStrings* strings = zone()->template New<CharacterClassStrings>(zone()); - RegExpTree* operand = ParseClassSetOperand(builder, &operand_type, ranges, - strings CHECK_FAILED); + base::uc32 character; + RegExpTree* operand = ParseClassSetOperand( + builder, &operand_type, ranges, strings, &character CHECK_FAILED); switch (current()) { case '-': if (Next() == '-') { if (operand == nullptr) { + if (operand_type == ClassSetOperandType::kClassSetCharacter) { + ranges->Add(CharacterRange::Singleton(character), zone()); + } operand = zone()->template New<RegExpClassSetOperand>(ranges, strings); } @@ -2908,6 +3038,9 @@ RegExpTree* RegExpParserImpl<CharT>::ParseCharacterClass( case '&': if (Next() == '&') { if (operand == nullptr) { + if (operand_type == ClassSetOperandType::kClassSetCharacter) { + ranges->Add(CharacterRange::Singleton(character), zone()); + } operand = zone()->template New<RegExpClassSetOperand>(ranges, strings); } @@ -2916,7 +3049,7 @@ RegExpTree* RegExpParserImpl<CharT>::ParseCharacterClass( } } return ParseClassUnion(builder, is_negated, operand, operand_type, ranges, - strings); + strings, character); } } @@ -3047,7 +3180,7 @@ bool RegExpBuilder::AddQuantifierToAtom( RegExpTree* atom = text_builder().PopLastAtom(); if (atom != nullptr) { FlushText(); - } else if (terms_.size() > 0) { + } else if (!terms_.empty()) { atom = terms_.back(); terms_.pop_back(); if (atom->IsLookaround()) { |