blob: eb2bb1c432a7eac7f3e7f3115966ddd383d065ca [file] [log] [blame]
// Copyright 2019 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.
#include "src/objects/js-regexp.h"
#include "src/objects/js-array-inl.h"
#include "src/objects/js-regexp-inl.h"
#include "src/regexp/regexp.h"
namespace v8 {
namespace internal {
MaybeHandle<JSArray> JSRegExpResult::GetAndCacheIndices(
Isolate* isolate, Handle<JSRegExpResult> regexp_result) {
// Check for cached indices. We do a slow lookup and set of
// the cached_indices_or_match_info and names fields just in
// case they have been migrated to dictionaries.
Handle<Object> indices_or_regexp(
GetProperty(
isolate, regexp_result,
isolate->factory()->regexp_result_cached_indices_or_regexp_symbol())
.ToHandleChecked());
if (indices_or_regexp->IsJSRegExp()) {
// Build and cache indices for next lookup.
// TODO(joshualitt): Instead of caching the indices, we could call
// ReconfigureToDataProperty on 'indices' setting its value to this
// newly created array. However, care would have to be taken to ensure
// a new map is not created each time.
// Grab regexp, its last_index, and the original subject string from the
// result and the re-execute the regexp to generate a new MatchInfo.
Handle<JSRegExp> regexp(JSRegExp::cast(*indices_or_regexp), isolate);
Handle<Object> input_object(
GetProperty(isolate, regexp_result,
isolate->factory()->regexp_result_regexp_input_symbol())
.ToHandleChecked());
Handle<String> subject(String::cast(*input_object), isolate);
Handle<Object> last_index_object(
GetProperty(
isolate, regexp_result,
isolate->factory()->regexp_result_regexp_last_index_symbol())
.ToHandleChecked());
int capture_count = regexp->CaptureCount();
Handle<RegExpMatchInfo> match_info =
RegExpMatchInfo::New(isolate, capture_count);
int last_index = Smi::ToInt(*last_index_object);
Handle<Object> result;
ASSIGN_RETURN_ON_EXCEPTION(
isolate, result,
RegExp::Exec(isolate, regexp, subject, last_index, match_info),
JSArray);
DCHECK_EQ(*result, *match_info);
Handle<Object> maybe_names(
GetProperty(isolate, regexp_result,
isolate->factory()->regexp_result_names_symbol())
.ToHandleChecked());
indices_or_regexp =
JSRegExpResultIndices::BuildIndices(isolate, match_info, maybe_names);
// Cache the result and clear the names array, last_index and subject.
SetProperty(
isolate, regexp_result,
isolate->factory()->regexp_result_cached_indices_or_regexp_symbol(),
indices_or_regexp)
.ToHandleChecked();
SetProperty(isolate, regexp_result,
isolate->factory()->regexp_result_names_symbol(),
isolate->factory()->undefined_value())
.ToHandleChecked();
SetProperty(isolate, regexp_result,
isolate->factory()->regexp_result_regexp_last_index_symbol(),
isolate->factory()->undefined_value())
.ToHandleChecked();
SetProperty(isolate, regexp_result,
isolate->factory()->regexp_result_regexp_input_symbol(),
isolate->factory()->undefined_value())
.ToHandleChecked();
}
return Handle<JSArray>::cast(indices_or_regexp);
}
Handle<JSRegExpResultIndices> JSRegExpResultIndices::BuildIndices(
Isolate* isolate, Handle<RegExpMatchInfo> match_info,
Handle<Object> maybe_names) {
Handle<JSRegExpResultIndices> indices(Handle<JSRegExpResultIndices>::cast(
isolate->factory()->NewJSObjectFromMap(
isolate->regexp_result_indices_map())));
// Initialize indices length to avoid having a partially initialized object
// should GC be triggered by creating a NewFixedArray.
indices->set_length(Smi::zero());
// Build indices array from RegExpMatchInfo.
int num_indices = match_info->NumberOfCaptureRegisters();
int num_results = num_indices >> 1;
Handle<FixedArray> indices_array =
isolate->factory()->NewFixedArray(num_results);
JSArray::SetContent(indices, indices_array);
for (int i = 0; i < num_results; i++) {
int base_offset = i * 2;
int start_offset = match_info->Capture(base_offset);
int end_offset = match_info->Capture(base_offset + 1);
// Any unmatched captures are set to undefined, otherwise we set them to a
// subarray of the indices.
if (start_offset == -1) {
indices_array->set(i, ReadOnlyRoots(isolate).undefined_value());
} else {
Handle<FixedArray> indices_sub_array(
isolate->factory()->NewFixedArray(2));
indices_sub_array->set(0, Smi::FromInt(start_offset));
indices_sub_array->set(1, Smi::FromInt(end_offset));
Handle<JSArray> indices_sub_jsarray =
isolate->factory()->NewJSArrayWithElements(indices_sub_array,
PACKED_SMI_ELEMENTS, 2);
indices_array->set(i, *indices_sub_jsarray);
}
}
// If there are no capture groups, set the groups property to undefined.
FieldIndex groups_index = FieldIndex::ForDescriptor(
indices->map(), InternalIndex(kGroupsDescriptorIndex));
if (maybe_names->IsUndefined(isolate)) {
indices->RawFastPropertyAtPut(groups_index,
ReadOnlyRoots(isolate).undefined_value());
return indices;
}
// Create a groups property which returns a dictionary of named captures to
// their corresponding capture indices.
Handle<FixedArray> names(Handle<FixedArray>::cast(maybe_names));
int num_names = names->length() >> 1;
Handle<NameDictionary> group_names = NameDictionary::New(isolate, num_names);
for (int i = 0; i < num_names; i++) {
int base_offset = i * 2;
int name_offset = base_offset;
int index_offset = base_offset + 1;
Handle<String> name(String::cast(names->get(name_offset)), isolate);
Handle<Smi> smi_index(Smi::cast(names->get(index_offset)), isolate);
Handle<Object> capture_indices(indices_array->get(smi_index->value()),
isolate);
if (!capture_indices->IsUndefined(isolate)) {
capture_indices = Handle<JSArray>::cast(capture_indices);
}
group_names = NameDictionary::Add(
isolate, group_names, name, capture_indices, PropertyDetails::Empty());
}
// Convert group_names to a JSObject and store at the groups property of the
// result indices.
Handle<FixedArrayBase> elements = isolate->factory()->empty_fixed_array();
Handle<HeapObject> null =
Handle<HeapObject>::cast(isolate->factory()->null_value());
Handle<JSObject> js_group_names =
isolate->factory()->NewSlowJSObjectWithPropertiesAndElements(
null, group_names, elements);
indices->RawFastPropertyAtPut(groups_index, *js_group_names);
return indices;
}
uint32_t JSRegExp::BacktrackLimit() const {
CHECK_EQ(TypeTag(), IRREGEXP);
return static_cast<uint32_t>(Smi::ToInt(DataAt(kIrregexpBacktrackLimit)));
}
// static
JSRegExp::Flags JSRegExp::FlagsFromString(Isolate* isolate,
Handle<String> flags, bool* success) {
int length = flags->length();
if (length == 0) {
*success = true;
return JSRegExp::kNone;
}
// A longer flags string cannot be valid.
if (length > JSRegExp::kFlagCount) return JSRegExp::Flags(0);
JSRegExp::Flags value(0);
if (flags->IsSeqOneByteString()) {
DisallowHeapAllocation no_gc;
SeqOneByteString seq_flags = SeqOneByteString::cast(*flags);
for (int i = 0; i < length; i++) {
base::Optional<JSRegExp::Flag> maybe_flag =
JSRegExp::FlagFromChar(seq_flags.Get(i));
if (!maybe_flag.has_value()) return JSRegExp::Flags(0);
JSRegExp::Flag flag = *maybe_flag;
// Duplicate flag.
if (value & flag) return JSRegExp::Flags(0);
value |= flag;
}
} else {
flags = String::Flatten(isolate, flags);
DisallowHeapAllocation no_gc;
String::FlatContent flags_content = flags->GetFlatContent(no_gc);
for (int i = 0; i < length; i++) {
base::Optional<JSRegExp::Flag> maybe_flag =
JSRegExp::FlagFromChar(flags_content.Get(i));
if (!maybe_flag.has_value()) return JSRegExp::Flags(0);
JSRegExp::Flag flag = *maybe_flag;
// Duplicate flag.
if (value & flag) return JSRegExp::Flags(0);
value |= flag;
}
}
*success = true;
return value;
}
// static
MaybeHandle<JSRegExp> JSRegExp::New(Isolate* isolate, Handle<String> pattern,
Flags flags, uint32_t backtrack_limit) {
Handle<JSFunction> constructor = isolate->regexp_function();
Handle<JSRegExp> regexp =
Handle<JSRegExp>::cast(isolate->factory()->NewJSObject(constructor));
return JSRegExp::Initialize(regexp, pattern, flags, backtrack_limit);
}
// static
Handle<JSRegExp> JSRegExp::Copy(Handle<JSRegExp> regexp) {
Isolate* const isolate = regexp->GetIsolate();
return Handle<JSRegExp>::cast(isolate->factory()->CopyJSObject(regexp));
}
Object JSRegExp::Code(bool is_latin1) const {
DCHECK_EQ(TypeTag(), JSRegExp::IRREGEXP);
return DataAt(code_index(is_latin1));
}
Object JSRegExp::Bytecode(bool is_latin1) const {
DCHECK_EQ(TypeTag(), JSRegExp::IRREGEXP);
return DataAt(bytecode_index(is_latin1));
}
bool JSRegExp::ShouldProduceBytecode() {
return FLAG_regexp_interpret_all ||
(FLAG_regexp_tier_up && !MarkedForTierUp());
}
// Only irregexps are subject to tier-up.
bool JSRegExp::CanTierUp() {
return FLAG_regexp_tier_up && TypeTag() == JSRegExp::IRREGEXP;
}
// An irregexp is considered to be marked for tier up if the tier-up ticks
// value reaches zero.
bool JSRegExp::MarkedForTierUp() {
DCHECK(data().IsFixedArray());
if (!CanTierUp()) {
return false;
}
return Smi::ToInt(DataAt(kIrregexpTicksUntilTierUpIndex)) == 0;
}
void JSRegExp::ResetLastTierUpTick() {
DCHECK(FLAG_regexp_tier_up);
DCHECK_EQ(TypeTag(), JSRegExp::IRREGEXP);
int tier_up_ticks = Smi::ToInt(DataAt(kIrregexpTicksUntilTierUpIndex)) + 1;
FixedArray::cast(data()).set(JSRegExp::kIrregexpTicksUntilTierUpIndex,
Smi::FromInt(tier_up_ticks));
}
void JSRegExp::TierUpTick() {
DCHECK(FLAG_regexp_tier_up);
DCHECK_EQ(TypeTag(), JSRegExp::IRREGEXP);
int tier_up_ticks = Smi::ToInt(DataAt(kIrregexpTicksUntilTierUpIndex));
if (tier_up_ticks == 0) {
return;
}
FixedArray::cast(data()).set(JSRegExp::kIrregexpTicksUntilTierUpIndex,
Smi::FromInt(tier_up_ticks - 1));
}
void JSRegExp::MarkTierUpForNextExec() {
DCHECK(FLAG_regexp_tier_up);
DCHECK_EQ(TypeTag(), JSRegExp::IRREGEXP);
FixedArray::cast(data()).set(JSRegExp::kIrregexpTicksUntilTierUpIndex,
Smi::zero());
}
// static
MaybeHandle<JSRegExp> JSRegExp::Initialize(Handle<JSRegExp> regexp,
Handle<String> source,
Handle<String> flags_string) {
Isolate* isolate = regexp->GetIsolate();
bool success = false;
Flags flags = JSRegExp::FlagsFromString(isolate, flags_string, &success);
if (!success) {
THROW_NEW_ERROR(
isolate,
NewSyntaxError(MessageTemplate::kInvalidRegExpFlags, flags_string),
JSRegExp);
}
return Initialize(regexp, source, flags);
}
namespace {
bool IsLineTerminator(int c) {
// Expected to return true for '\n', '\r', 0x2028, and 0x2029.
return unibrow::IsLineTerminator(static_cast<unibrow::uchar>(c));
}
// TODO(jgruber): Consider merging CountAdditionalEscapeChars and
// WriteEscapedRegExpSource into a single function to deduplicate dispatch logic
// and move related code closer to each other.
template <typename Char>
int CountAdditionalEscapeChars(Handle<String> source, bool* needs_escapes_out) {
DisallowHeapAllocation no_gc;
int escapes = 0;
bool needs_escapes = false;
bool in_char_class = false;
Vector<const Char> src = source->GetCharVector<Char>(no_gc);
for (int i = 0; i < src.length(); i++) {
const Char c = src[i];
if (c == '\\') {
if (i + 1 < src.length() && IsLineTerminator(src[i + 1])) {
// This '\' is ignored since the next character itself will be escaped.
escapes--;
} else {
// Escape. Skip next character, which will be copied verbatim;
i++;
}
} else if (c == '/' && !in_char_class) {
// Not escaped forward-slash needs escape.
needs_escapes = true;
escapes++;
} else if (c == '[') {
in_char_class = true;
} else if (c == ']') {
in_char_class = false;
} else if (c == '\n') {
needs_escapes = true;
escapes++;
} else if (c == '\r') {
needs_escapes = true;
escapes++;
} else if (static_cast<int>(c) == 0x2028) {
needs_escapes = true;
escapes += std::strlen("\\u2028") - 1;
} else if (static_cast<int>(c) == 0x2029) {
needs_escapes = true;
escapes += std::strlen("\\u2029") - 1;
} else {
DCHECK(!IsLineTerminator(c));
}
}
DCHECK(!in_char_class);
DCHECK_GE(escapes, 0);
DCHECK_IMPLIES(escapes != 0, needs_escapes);
*needs_escapes_out = needs_escapes;
return escapes;
}
template <typename Char>
void WriteStringToCharVector(Vector<Char> v, int* d, const char* string) {
int s = 0;
while (string[s] != '\0') v[(*d)++] = string[s++];
}
template <typename Char, typename StringType>
Handle<StringType> WriteEscapedRegExpSource(Handle<String> source,
Handle<StringType> result) {
DisallowHeapAllocation no_gc;
Vector<const Char> src = source->GetCharVector<Char>(no_gc);
Vector<Char> dst(result->GetChars(no_gc), result->length());
int s = 0;
int d = 0;
bool in_char_class = false;
while (s < src.length()) {
const Char c = src[s];
if (c == '\\') {
if (s + 1 < src.length() && IsLineTerminator(src[s + 1])) {
// This '\' is ignored since the next character itself will be escaped.
s++;
continue;
} else {
// Escape. Copy this and next character.
dst[d++] = src[s++];
}
if (s == src.length()) break;
} else if (c == '/' && !in_char_class) {
// Not escaped forward-slash needs escape.
dst[d++] = '\\';
} else if (c == '[') {
in_char_class = true;
} else if (c == ']') {
in_char_class = false;
} else if (c == '\n') {
WriteStringToCharVector(dst, &d, "\\n");
s++;
continue;
} else if (c == '\r') {
WriteStringToCharVector(dst, &d, "\\r");
s++;
continue;
} else if (static_cast<int>(c) == 0x2028) {
WriteStringToCharVector(dst, &d, "\\u2028");
s++;
continue;
} else if (static_cast<int>(c) == 0x2029) {
WriteStringToCharVector(dst, &d, "\\u2029");
s++;
continue;
} else {
DCHECK(!IsLineTerminator(c));
}
dst[d++] = src[s++];
}
DCHECK_EQ(result->length(), d);
DCHECK(!in_char_class);
return result;
}
MaybeHandle<String> EscapeRegExpSource(Isolate* isolate,
Handle<String> source) {
DCHECK(source->IsFlat());
if (source->length() == 0) return isolate->factory()->query_colon_string();
bool one_byte = String::IsOneByteRepresentationUnderneath(*source);
bool needs_escapes = false;
int additional_escape_chars =
one_byte ? CountAdditionalEscapeChars<uint8_t>(source, &needs_escapes)
: CountAdditionalEscapeChars<uc16>(source, &needs_escapes);
if (!needs_escapes) return source;
int length = source->length() + additional_escape_chars;
if (one_byte) {
Handle<SeqOneByteString> result;
ASSIGN_RETURN_ON_EXCEPTION(isolate, result,
isolate->factory()->NewRawOneByteString(length),
String);
return WriteEscapedRegExpSource<uint8_t>(source, result);
} else {
Handle<SeqTwoByteString> result;
ASSIGN_RETURN_ON_EXCEPTION(isolate, result,
isolate->factory()->NewRawTwoByteString(length),
String);
return WriteEscapedRegExpSource<uc16>(source, result);
}
}
} // namespace
// static
MaybeHandle<JSRegExp> JSRegExp::Initialize(Handle<JSRegExp> regexp,
Handle<String> source, Flags flags,
uint32_t backtrack_limit) {
Isolate* isolate = regexp->GetIsolate();
Factory* factory = isolate->factory();
// If source is the empty string we set it to "(?:)" instead as
// suggested by ECMA-262, 5th, section 15.10.4.1.
if (source->length() == 0) source = factory->query_colon_string();
source = String::Flatten(isolate, source);
RETURN_ON_EXCEPTION(
isolate, RegExp::Compile(isolate, regexp, source, flags, backtrack_limit),
JSRegExp);
Handle<String> escaped_source;
ASSIGN_RETURN_ON_EXCEPTION(isolate, escaped_source,
EscapeRegExpSource(isolate, source), JSRegExp);
regexp->set_source(*escaped_source);
regexp->set_flags(Smi::FromInt(flags));
Map map = regexp->map();
Object constructor = map.GetConstructor();
if (constructor.IsJSFunction() &&
JSFunction::cast(constructor).initial_map() == map) {
// If we still have the original map, set in-object properties directly.
regexp->InObjectPropertyAtPut(JSRegExp::kLastIndexFieldIndex, Smi::zero(),
SKIP_WRITE_BARRIER);
} else {
// Map has changed, so use generic, but slower, method.
RETURN_ON_EXCEPTION(
isolate,
Object::SetProperty(isolate, regexp, factory->lastIndex_string(),
Handle<Smi>(Smi::zero(), isolate)),
JSRegExp);
}
return regexp;
}
} // namespace internal
} // namespace v8