1 // Copyright 2011 the V8 project authors. All rights reserved.
2 // Use of this source code is governed by a BSD-style license that can be
3 // found in the LICENSE file.
4
5 #ifndef V8_JSON_PARSER_H_
6 #define V8_JSON_PARSER_H_
7
8 #include "src/v8.h"
9
10 #include "src/char-predicates-inl.h"
11 #include "src/conversions.h"
12 #include "src/heap/spaces-inl.h"
13 #include "src/messages.h"
14 #include "src/token.h"
15
16 namespace v8 {
17 namespace internal {
18
19 // A simple json parser.
20 template <bool seq_one_byte>
21 class JsonParser BASE_EMBEDDED {
22 public:
Parse(Handle<String> source)23 MUST_USE_RESULT static MaybeHandle<Object> Parse(Handle<String> source) {
24 return JsonParser(source).ParseJson();
25 }
26
27 static const int kEndOfString = -1;
28
29 private:
JsonParser(Handle<String> source)30 explicit JsonParser(Handle<String> source)
31 : source_(source),
32 source_length_(source->length()),
33 isolate_(source->map()->GetHeap()->isolate()),
34 factory_(isolate_->factory()),
35 zone_(isolate_),
36 object_constructor_(isolate_->native_context()->object_function(),
37 isolate_),
38 position_(-1) {
39 source_ = String::Flatten(source_);
40 pretenure_ = (source_length_ >= kPretenureTreshold) ? TENURED : NOT_TENURED;
41
42 // Optimized fast case where we only have Latin1 characters.
43 if (seq_one_byte) {
44 seq_source_ = Handle<SeqOneByteString>::cast(source_);
45 }
46 }
47
48 // Parse a string containing a single JSON value.
49 MaybeHandle<Object> ParseJson();
50
Advance()51 inline void Advance() {
52 position_++;
53 if (position_ >= source_length_) {
54 c0_ = kEndOfString;
55 } else if (seq_one_byte) {
56 c0_ = seq_source_->SeqOneByteStringGet(position_);
57 } else {
58 c0_ = source_->Get(position_);
59 }
60 }
61
62 // The JSON lexical grammar is specified in the ECMAScript 5 standard,
63 // section 15.12.1.1. The only allowed whitespace characters between tokens
64 // are tab, carriage-return, newline and space.
65
AdvanceSkipWhitespace()66 inline void AdvanceSkipWhitespace() {
67 do {
68 Advance();
69 } while (c0_ == ' ' || c0_ == '\t' || c0_ == '\n' || c0_ == '\r');
70 }
71
SkipWhitespace()72 inline void SkipWhitespace() {
73 while (c0_ == ' ' || c0_ == '\t' || c0_ == '\n' || c0_ == '\r') {
74 Advance();
75 }
76 }
77
AdvanceGetChar()78 inline uc32 AdvanceGetChar() {
79 Advance();
80 return c0_;
81 }
82
83 // Checks that current charater is c.
84 // If so, then consume c and skip whitespace.
MatchSkipWhiteSpace(uc32 c)85 inline bool MatchSkipWhiteSpace(uc32 c) {
86 if (c0_ == c) {
87 AdvanceSkipWhitespace();
88 return true;
89 }
90 return false;
91 }
92
93 // A JSON string (production JSONString) is subset of valid JavaScript string
94 // literals. The string must only be double-quoted (not single-quoted), and
95 // the only allowed backslash-escapes are ", /, \, b, f, n, r, t and
96 // four-digit hex escapes (uXXXX). Any other use of backslashes is invalid.
ParseJsonString()97 Handle<String> ParseJsonString() {
98 return ScanJsonString<false>();
99 }
100
ParseJsonString(Handle<String> expected)101 bool ParseJsonString(Handle<String> expected) {
102 int length = expected->length();
103 if (source_->length() - position_ - 1 > length) {
104 DisallowHeapAllocation no_gc;
105 String::FlatContent content = expected->GetFlatContent();
106 if (content.IsOneByte()) {
107 DCHECK_EQ('"', c0_);
108 const uint8_t* input_chars = seq_source_->GetChars() + position_ + 1;
109 const uint8_t* expected_chars = content.ToOneByteVector().start();
110 for (int i = 0; i < length; i++) {
111 uint8_t c0 = input_chars[i];
112 if (c0 != expected_chars[i] ||
113 c0 == '"' || c0 < 0x20 || c0 == '\\') {
114 return false;
115 }
116 }
117 if (input_chars[length] == '"') {
118 position_ = position_ + length + 1;
119 AdvanceSkipWhitespace();
120 return true;
121 }
122 }
123 }
124 return false;
125 }
126
ParseJsonInternalizedString()127 Handle<String> ParseJsonInternalizedString() {
128 return ScanJsonString<true>();
129 }
130
131 template <bool is_internalized>
132 Handle<String> ScanJsonString();
133 // Creates a new string and copies prefix[start..end] into the beginning
134 // of it. Then scans the rest of the string, adding characters after the
135 // prefix. Called by ScanJsonString when reaching a '\' or non-Latin1 char.
136 template <typename StringType, typename SinkChar>
137 Handle<String> SlowScanJsonString(Handle<String> prefix, int start, int end);
138
139 // A JSON number (production JSONNumber) is a subset of the valid JavaScript
140 // decimal number literals.
141 // It includes an optional minus sign, must have at least one
142 // digit before and after a decimal point, may not have prefixed zeros (unless
143 // the integer part is zero), and may include an exponent part (e.g., "e-10").
144 // Hexadecimal and octal numbers are not allowed.
145 Handle<Object> ParseJsonNumber();
146
147 // Parse a single JSON value from input (grammar production JSONValue).
148 // A JSON value is either a (double-quoted) string literal, a number literal,
149 // one of "true", "false", or "null", or an object or array literal.
150 Handle<Object> ParseJsonValue();
151
152 // Parse a JSON object literal (grammar production JSONObject).
153 // An object literal is a squiggly-braced and comma separated sequence
154 // (possibly empty) of key/value pairs, where the key is a JSON string
155 // literal, the value is a JSON value, and the two are separated by a colon.
156 // A JSON array doesn't allow numbers and identifiers as keys, like a
157 // JavaScript array.
158 Handle<Object> ParseJsonObject();
159
160 // Parses a JSON array literal (grammar production JSONArray). An array
161 // literal is a square-bracketed and comma separated sequence (possibly empty)
162 // of JSON values.
163 // A JSON array doesn't allow leaving out values from the sequence, nor does
164 // it allow a terminal comma, like a JavaScript array does.
165 Handle<Object> ParseJsonArray();
166
167
168 // Mark that a parsing error has happened at the current token, and
169 // return a null handle. Primarily for readability.
ReportUnexpectedCharacter()170 inline Handle<Object> ReportUnexpectedCharacter() {
171 return Handle<Object>::null();
172 }
173
isolate()174 inline Isolate* isolate() { return isolate_; }
factory()175 inline Factory* factory() { return factory_; }
object_constructor()176 inline Handle<JSFunction> object_constructor() { return object_constructor_; }
177
178 static const int kInitialSpecialStringLength = 1024;
179 static const int kPretenureTreshold = 100 * 1024;
180
181
182 private:
zone()183 Zone* zone() { return &zone_; }
184
185 void CommitStateToJsonObject(Handle<JSObject> json_object, Handle<Map> map,
186 ZoneList<Handle<Object> >* properties);
187
188 Handle<String> source_;
189 int source_length_;
190 Handle<SeqOneByteString> seq_source_;
191
192 PretenureFlag pretenure_;
193 Isolate* isolate_;
194 Factory* factory_;
195 Zone zone_;
196 Handle<JSFunction> object_constructor_;
197 uc32 c0_;
198 int position_;
199 };
200
201 template <bool seq_one_byte>
ParseJson()202 MaybeHandle<Object> JsonParser<seq_one_byte>::ParseJson() {
203 // Advance to the first character (possibly EOS)
204 AdvanceSkipWhitespace();
205 Handle<Object> result = ParseJsonValue();
206 if (result.is_null() || c0_ != kEndOfString) {
207 // Some exception (for example stack overflow) is already pending.
208 if (isolate_->has_pending_exception()) return Handle<Object>::null();
209
210 // Parse failed. Current character is the unexpected token.
211 const char* message;
212 Factory* factory = this->factory();
213 Handle<JSArray> array;
214
215 switch (c0_) {
216 case kEndOfString:
217 message = "unexpected_eos";
218 array = factory->NewJSArray(0);
219 break;
220 case '-':
221 case '0':
222 case '1':
223 case '2':
224 case '3':
225 case '4':
226 case '5':
227 case '6':
228 case '7':
229 case '8':
230 case '9':
231 message = "unexpected_token_number";
232 array = factory->NewJSArray(0);
233 break;
234 case '"':
235 message = "unexpected_token_string";
236 array = factory->NewJSArray(0);
237 break;
238 default:
239 message = "unexpected_token";
240 Handle<Object> name = factory->LookupSingleCharacterStringFromCode(c0_);
241 Handle<FixedArray> element = factory->NewFixedArray(1);
242 element->set(0, *name);
243 array = factory->NewJSArrayWithElements(element);
244 break;
245 }
246
247 MessageLocation location(factory->NewScript(source_),
248 position_,
249 position_ + 1);
250 Handle<Object> error;
251 ASSIGN_RETURN_ON_EXCEPTION(isolate(), error,
252 factory->NewSyntaxError(message, array), Object);
253 return isolate()->template Throw<Object>(error, &location);
254 }
255 return result;
256 }
257
258
259 // Parse any JSON value.
260 template <bool seq_one_byte>
ParseJsonValue()261 Handle<Object> JsonParser<seq_one_byte>::ParseJsonValue() {
262 StackLimitCheck stack_check(isolate_);
263 if (stack_check.HasOverflowed()) {
264 isolate_->StackOverflow();
265 return Handle<Object>::null();
266 }
267
268 if (c0_ == '"') return ParseJsonString();
269 if ((c0_ >= '0' && c0_ <= '9') || c0_ == '-') return ParseJsonNumber();
270 if (c0_ == '{') return ParseJsonObject();
271 if (c0_ == '[') return ParseJsonArray();
272 if (c0_ == 'f') {
273 if (AdvanceGetChar() == 'a' && AdvanceGetChar() == 'l' &&
274 AdvanceGetChar() == 's' && AdvanceGetChar() == 'e') {
275 AdvanceSkipWhitespace();
276 return factory()->false_value();
277 }
278 return ReportUnexpectedCharacter();
279 }
280 if (c0_ == 't') {
281 if (AdvanceGetChar() == 'r' && AdvanceGetChar() == 'u' &&
282 AdvanceGetChar() == 'e') {
283 AdvanceSkipWhitespace();
284 return factory()->true_value();
285 }
286 return ReportUnexpectedCharacter();
287 }
288 if (c0_ == 'n') {
289 if (AdvanceGetChar() == 'u' && AdvanceGetChar() == 'l' &&
290 AdvanceGetChar() == 'l') {
291 AdvanceSkipWhitespace();
292 return factory()->null_value();
293 }
294 return ReportUnexpectedCharacter();
295 }
296 return ReportUnexpectedCharacter();
297 }
298
299
300 // Parse a JSON object. Position must be right at '{'.
301 template <bool seq_one_byte>
ParseJsonObject()302 Handle<Object> JsonParser<seq_one_byte>::ParseJsonObject() {
303 HandleScope scope(isolate());
304 Handle<JSObject> json_object =
305 factory()->NewJSObject(object_constructor(), pretenure_);
306 Handle<Map> map(json_object->map());
307 ZoneList<Handle<Object> > properties(8, zone());
308 DCHECK_EQ(c0_, '{');
309
310 bool transitioning = true;
311
312 AdvanceSkipWhitespace();
313 if (c0_ != '}') {
314 do {
315 if (c0_ != '"') return ReportUnexpectedCharacter();
316
317 int start_position = position_;
318 Advance();
319
320 uint32_t index = 0;
321 if (c0_ >= '0' && c0_ <= '9') {
322 // Maybe an array index, try to parse it.
323 if (c0_ == '0') {
324 // With a leading zero, the string has to be "0" only to be an index.
325 Advance();
326 } else {
327 do {
328 int d = c0_ - '0';
329 if (index > 429496729U - ((d > 5) ? 1 : 0)) break;
330 index = (index * 10) + d;
331 Advance();
332 } while (c0_ >= '0' && c0_ <= '9');
333 }
334
335 if (c0_ == '"') {
336 // Successfully parsed index, parse and store element.
337 AdvanceSkipWhitespace();
338
339 if (c0_ != ':') return ReportUnexpectedCharacter();
340 AdvanceSkipWhitespace();
341 Handle<Object> value = ParseJsonValue();
342 if (value.is_null()) return ReportUnexpectedCharacter();
343
344 JSObject::SetOwnElement(json_object, index, value, SLOPPY).Assert();
345 continue;
346 }
347 // Not an index, fallback to the slow path.
348 }
349
350 position_ = start_position;
351 #ifdef DEBUG
352 c0_ = '"';
353 #endif
354
355 Handle<String> key;
356 Handle<Object> value;
357
358 // Try to follow existing transitions as long as possible. Once we stop
359 // transitioning, no transition can be found anymore.
360 if (transitioning) {
361 // First check whether there is a single expected transition. If so, try
362 // to parse it first.
363 bool follow_expected = false;
364 Handle<Map> target;
365 if (seq_one_byte) {
366 key = Map::ExpectedTransitionKey(map);
367 follow_expected = !key.is_null() && ParseJsonString(key);
368 }
369 // If the expected transition hits, follow it.
370 if (follow_expected) {
371 target = Map::ExpectedTransitionTarget(map);
372 } else {
373 // If the expected transition failed, parse an internalized string and
374 // try to find a matching transition.
375 key = ParseJsonInternalizedString();
376 if (key.is_null()) return ReportUnexpectedCharacter();
377
378 target = Map::FindTransitionToField(map, key);
379 // If a transition was found, follow it and continue.
380 transitioning = !target.is_null();
381 }
382 if (c0_ != ':') return ReportUnexpectedCharacter();
383
384 AdvanceSkipWhitespace();
385 value = ParseJsonValue();
386 if (value.is_null()) return ReportUnexpectedCharacter();
387
388 if (transitioning) {
389 int descriptor = map->NumberOfOwnDescriptors();
390 PropertyDetails details =
391 target->instance_descriptors()->GetDetails(descriptor);
392 Representation expected_representation = details.representation();
393
394 if (value->FitsRepresentation(expected_representation)) {
395 if (expected_representation.IsDouble()) {
396 value = Object::NewStorageFor(isolate(), value,
397 expected_representation);
398 } else if (expected_representation.IsHeapObject() &&
399 !target->instance_descriptors()->GetFieldType(
400 descriptor)->NowContains(value)) {
401 Handle<HeapType> value_type(value->OptimalType(
402 isolate(), expected_representation));
403 Map::GeneralizeFieldType(target, descriptor, value_type);
404 }
405 DCHECK(target->instance_descriptors()->GetFieldType(
406 descriptor)->NowContains(value));
407 properties.Add(value, zone());
408 map = target;
409 continue;
410 } else {
411 transitioning = false;
412 }
413 }
414
415 // Commit the intermediate state to the object and stop transitioning.
416 CommitStateToJsonObject(json_object, map, &properties);
417 } else {
418 key = ParseJsonInternalizedString();
419 if (key.is_null() || c0_ != ':') return ReportUnexpectedCharacter();
420
421 AdvanceSkipWhitespace();
422 value = ParseJsonValue();
423 if (value.is_null()) return ReportUnexpectedCharacter();
424 }
425
426 Runtime::DefineObjectProperty(json_object, key, value, NONE).Check();
427 } while (MatchSkipWhiteSpace(','));
428 if (c0_ != '}') {
429 return ReportUnexpectedCharacter();
430 }
431
432 // If we transitioned until the very end, transition the map now.
433 if (transitioning) {
434 CommitStateToJsonObject(json_object, map, &properties);
435 }
436 }
437 AdvanceSkipWhitespace();
438 return scope.CloseAndEscape(json_object);
439 }
440
441
442 template <bool seq_one_byte>
CommitStateToJsonObject(Handle<JSObject> json_object,Handle<Map> map,ZoneList<Handle<Object>> * properties)443 void JsonParser<seq_one_byte>::CommitStateToJsonObject(
444 Handle<JSObject> json_object, Handle<Map> map,
445 ZoneList<Handle<Object> >* properties) {
446 JSObject::AllocateStorageForMap(json_object, map);
447 DCHECK(!json_object->map()->is_dictionary_map());
448
449 DisallowHeapAllocation no_gc;
450 Factory* factory = isolate()->factory();
451 // If the |json_object|'s map is exactly the same as |map| then the
452 // |properties| values correspond to the |map| and nothing more has to be
453 // done. But if the |json_object|'s map is different then we have to
454 // iterate descriptors to ensure that properties still correspond to the
455 // map.
456 bool slow_case = json_object->map() != *map;
457 DescriptorArray* descriptors = NULL;
458
459 int length = properties->length();
460 if (slow_case) {
461 descriptors = json_object->map()->instance_descriptors();
462 DCHECK(json_object->map()->NumberOfOwnDescriptors() == length);
463 }
464 for (int i = 0; i < length; i++) {
465 Handle<Object> value = (*properties)[i];
466 if (slow_case && value->IsMutableHeapNumber() &&
467 !descriptors->GetDetails(i).representation().IsDouble()) {
468 // Turn mutable heap numbers into immutable if the field representation
469 // is not double.
470 HeapNumber::cast(*value)->set_map(*factory->heap_number_map());
471 }
472 FieldIndex index = FieldIndex::ForPropertyIndex(*map, i);
473 json_object->FastPropertyAtPut(index, *value);
474 }
475 }
476
477
478 // Parse a JSON array. Position must be right at '['.
479 template <bool seq_one_byte>
ParseJsonArray()480 Handle<Object> JsonParser<seq_one_byte>::ParseJsonArray() {
481 HandleScope scope(isolate());
482 ZoneList<Handle<Object> > elements(4, zone());
483 DCHECK_EQ(c0_, '[');
484
485 AdvanceSkipWhitespace();
486 if (c0_ != ']') {
487 do {
488 Handle<Object> element = ParseJsonValue();
489 if (element.is_null()) return ReportUnexpectedCharacter();
490 elements.Add(element, zone());
491 } while (MatchSkipWhiteSpace(','));
492 if (c0_ != ']') {
493 return ReportUnexpectedCharacter();
494 }
495 }
496 AdvanceSkipWhitespace();
497 // Allocate a fixed array with all the elements.
498 Handle<FixedArray> fast_elements =
499 factory()->NewFixedArray(elements.length(), pretenure_);
500 for (int i = 0, n = elements.length(); i < n; i++) {
501 fast_elements->set(i, *elements[i]);
502 }
503 Handle<Object> json_array = factory()->NewJSArrayWithElements(
504 fast_elements, FAST_ELEMENTS, pretenure_);
505 return scope.CloseAndEscape(json_array);
506 }
507
508
509 template <bool seq_one_byte>
ParseJsonNumber()510 Handle<Object> JsonParser<seq_one_byte>::ParseJsonNumber() {
511 bool negative = false;
512 int beg_pos = position_;
513 if (c0_ == '-') {
514 Advance();
515 negative = true;
516 }
517 if (c0_ == '0') {
518 Advance();
519 // Prefix zero is only allowed if it's the only digit before
520 // a decimal point or exponent.
521 if ('0' <= c0_ && c0_ <= '9') return ReportUnexpectedCharacter();
522 } else {
523 int i = 0;
524 int digits = 0;
525 if (c0_ < '1' || c0_ > '9') return ReportUnexpectedCharacter();
526 do {
527 i = i * 10 + c0_ - '0';
528 digits++;
529 Advance();
530 } while (c0_ >= '0' && c0_ <= '9');
531 if (c0_ != '.' && c0_ != 'e' && c0_ != 'E' && digits < 10) {
532 SkipWhitespace();
533 return Handle<Smi>(Smi::FromInt((negative ? -i : i)), isolate());
534 }
535 }
536 if (c0_ == '.') {
537 Advance();
538 if (c0_ < '0' || c0_ > '9') return ReportUnexpectedCharacter();
539 do {
540 Advance();
541 } while (c0_ >= '0' && c0_ <= '9');
542 }
543 if (AsciiAlphaToLower(c0_) == 'e') {
544 Advance();
545 if (c0_ == '-' || c0_ == '+') Advance();
546 if (c0_ < '0' || c0_ > '9') return ReportUnexpectedCharacter();
547 do {
548 Advance();
549 } while (c0_ >= '0' && c0_ <= '9');
550 }
551 int length = position_ - beg_pos;
552 double number;
553 if (seq_one_byte) {
554 Vector<const uint8_t> chars(seq_source_->GetChars() + beg_pos, length);
555 number = StringToDouble(isolate()->unicode_cache(),
556 chars,
557 NO_FLAGS, // Hex, octal or trailing junk.
558 base::OS::nan_value());
559 } else {
560 Vector<uint8_t> buffer = Vector<uint8_t>::New(length);
561 String::WriteToFlat(*source_, buffer.start(), beg_pos, position_);
562 Vector<const uint8_t> result =
563 Vector<const uint8_t>(buffer.start(), length);
564 number = StringToDouble(isolate()->unicode_cache(),
565 result,
566 NO_FLAGS, // Hex, octal or trailing junk.
567 0.0);
568 buffer.Dispose();
569 }
570 SkipWhitespace();
571 return factory()->NewNumber(number, pretenure_);
572 }
573
574
575 template <typename StringType>
576 inline void SeqStringSet(Handle<StringType> seq_str, int i, uc32 c);
577
578 template <>
SeqStringSet(Handle<SeqTwoByteString> seq_str,int i,uc32 c)579 inline void SeqStringSet(Handle<SeqTwoByteString> seq_str, int i, uc32 c) {
580 seq_str->SeqTwoByteStringSet(i, c);
581 }
582
583 template <>
SeqStringSet(Handle<SeqOneByteString> seq_str,int i,uc32 c)584 inline void SeqStringSet(Handle<SeqOneByteString> seq_str, int i, uc32 c) {
585 seq_str->SeqOneByteStringSet(i, c);
586 }
587
588 template <typename StringType>
589 inline Handle<StringType> NewRawString(Factory* factory,
590 int length,
591 PretenureFlag pretenure);
592
593 template <>
NewRawString(Factory * factory,int length,PretenureFlag pretenure)594 inline Handle<SeqTwoByteString> NewRawString(Factory* factory,
595 int length,
596 PretenureFlag pretenure) {
597 return factory->NewRawTwoByteString(length, pretenure).ToHandleChecked();
598 }
599
600 template <>
NewRawString(Factory * factory,int length,PretenureFlag pretenure)601 inline Handle<SeqOneByteString> NewRawString(Factory* factory,
602 int length,
603 PretenureFlag pretenure) {
604 return factory->NewRawOneByteString(length, pretenure).ToHandleChecked();
605 }
606
607
608 // Scans the rest of a JSON string starting from position_ and writes
609 // prefix[start..end] along with the scanned characters into a
610 // sequential string of type StringType.
611 template <bool seq_one_byte>
612 template <typename StringType, typename SinkChar>
SlowScanJsonString(Handle<String> prefix,int start,int end)613 Handle<String> JsonParser<seq_one_byte>::SlowScanJsonString(
614 Handle<String> prefix, int start, int end) {
615 int count = end - start;
616 int max_length = count + source_length_ - position_;
617 int length = Min(max_length, Max(kInitialSpecialStringLength, 2 * count));
618 Handle<StringType> seq_string =
619 NewRawString<StringType>(factory(), length, pretenure_);
620 // Copy prefix into seq_str.
621 SinkChar* dest = seq_string->GetChars();
622 String::WriteToFlat(*prefix, dest, start, end);
623
624 while (c0_ != '"') {
625 // Check for control character (0x00-0x1f) or unterminated string (<0).
626 if (c0_ < 0x20) return Handle<String>::null();
627 if (count >= length) {
628 // We need to create a longer sequential string for the result.
629 return SlowScanJsonString<StringType, SinkChar>(seq_string, 0, count);
630 }
631 if (c0_ != '\\') {
632 // If the sink can contain UC16 characters, or source_ contains only
633 // Latin1 characters, there's no need to test whether we can store the
634 // character. Otherwise check whether the UC16 source character can fit
635 // in the Latin1 sink.
636 if (sizeof(SinkChar) == kUC16Size || seq_one_byte ||
637 c0_ <= String::kMaxOneByteCharCode) {
638 SeqStringSet(seq_string, count++, c0_);
639 Advance();
640 } else {
641 // StringType is SeqOneByteString and we just read a non-Latin1 char.
642 return SlowScanJsonString<SeqTwoByteString, uc16>(seq_string, 0, count);
643 }
644 } else {
645 Advance(); // Advance past the \.
646 switch (c0_) {
647 case '"':
648 case '\\':
649 case '/':
650 SeqStringSet(seq_string, count++, c0_);
651 break;
652 case 'b':
653 SeqStringSet(seq_string, count++, '\x08');
654 break;
655 case 'f':
656 SeqStringSet(seq_string, count++, '\x0c');
657 break;
658 case 'n':
659 SeqStringSet(seq_string, count++, '\x0a');
660 break;
661 case 'r':
662 SeqStringSet(seq_string, count++, '\x0d');
663 break;
664 case 't':
665 SeqStringSet(seq_string, count++, '\x09');
666 break;
667 case 'u': {
668 uc32 value = 0;
669 for (int i = 0; i < 4; i++) {
670 Advance();
671 int digit = HexValue(c0_);
672 if (digit < 0) {
673 return Handle<String>::null();
674 }
675 value = value * 16 + digit;
676 }
677 if (sizeof(SinkChar) == kUC16Size ||
678 value <= String::kMaxOneByteCharCode) {
679 SeqStringSet(seq_string, count++, value);
680 break;
681 } else {
682 // StringType is SeqOneByteString and we just read a non-Latin1
683 // char.
684 position_ -= 6; // Rewind position_ to \ in \uxxxx.
685 Advance();
686 return SlowScanJsonString<SeqTwoByteString, uc16>(seq_string,
687 0,
688 count);
689 }
690 }
691 default:
692 return Handle<String>::null();
693 }
694 Advance();
695 }
696 }
697
698 DCHECK_EQ('"', c0_);
699 // Advance past the last '"'.
700 AdvanceSkipWhitespace();
701
702 // Shrink seq_string length to count and return.
703 return SeqString::Truncate(seq_string, count);
704 }
705
706
707 template <bool seq_one_byte>
708 template <bool is_internalized>
ScanJsonString()709 Handle<String> JsonParser<seq_one_byte>::ScanJsonString() {
710 DCHECK_EQ('"', c0_);
711 Advance();
712 if (c0_ == '"') {
713 AdvanceSkipWhitespace();
714 return factory()->empty_string();
715 }
716
717 if (seq_one_byte && is_internalized) {
718 // Fast path for existing internalized strings. If the the string being
719 // parsed is not a known internalized string, contains backslashes or
720 // unexpectedly reaches the end of string, return with an empty handle.
721 uint32_t running_hash = isolate()->heap()->HashSeed();
722 int position = position_;
723 uc32 c0 = c0_;
724 do {
725 if (c0 == '\\') {
726 c0_ = c0;
727 int beg_pos = position_;
728 position_ = position;
729 return SlowScanJsonString<SeqOneByteString, uint8_t>(source_,
730 beg_pos,
731 position_);
732 }
733 if (c0 < 0x20) return Handle<String>::null();
734 if (static_cast<uint32_t>(c0) >
735 unibrow::Utf16::kMaxNonSurrogateCharCode) {
736 running_hash =
737 StringHasher::AddCharacterCore(running_hash,
738 unibrow::Utf16::LeadSurrogate(c0));
739 running_hash =
740 StringHasher::AddCharacterCore(running_hash,
741 unibrow::Utf16::TrailSurrogate(c0));
742 } else {
743 running_hash = StringHasher::AddCharacterCore(running_hash, c0);
744 }
745 position++;
746 if (position >= source_length_) return Handle<String>::null();
747 c0 = seq_source_->SeqOneByteStringGet(position);
748 } while (c0 != '"');
749 int length = position - position_;
750 uint32_t hash = (length <= String::kMaxHashCalcLength)
751 ? StringHasher::GetHashCore(running_hash)
752 : static_cast<uint32_t>(length);
753 Vector<const uint8_t> string_vector(
754 seq_source_->GetChars() + position_, length);
755 StringTable* string_table = isolate()->heap()->string_table();
756 uint32_t capacity = string_table->Capacity();
757 uint32_t entry = StringTable::FirstProbe(hash, capacity);
758 uint32_t count = 1;
759 Handle<String> result;
760 while (true) {
761 Object* element = string_table->KeyAt(entry);
762 if (element == isolate()->heap()->undefined_value()) {
763 // Lookup failure.
764 result = factory()->InternalizeOneByteString(
765 seq_source_, position_, length);
766 break;
767 }
768 if (element != isolate()->heap()->the_hole_value() &&
769 String::cast(element)->IsOneByteEqualTo(string_vector)) {
770 result = Handle<String>(String::cast(element), isolate());
771 #ifdef DEBUG
772 uint32_t hash_field =
773 (hash << String::kHashShift) | String::kIsNotArrayIndexMask;
774 DCHECK_EQ(static_cast<int>(result->Hash()),
775 static_cast<int>(hash_field >> String::kHashShift));
776 #endif
777 break;
778 }
779 entry = StringTable::NextProbe(entry, count++, capacity);
780 }
781 position_ = position;
782 // Advance past the last '"'.
783 AdvanceSkipWhitespace();
784 return result;
785 }
786
787 int beg_pos = position_;
788 // Fast case for Latin1 only without escape characters.
789 do {
790 // Check for control character (0x00-0x1f) or unterminated string (<0).
791 if (c0_ < 0x20) return Handle<String>::null();
792 if (c0_ != '\\') {
793 if (seq_one_byte || c0_ <= String::kMaxOneByteCharCode) {
794 Advance();
795 } else {
796 return SlowScanJsonString<SeqTwoByteString, uc16>(source_,
797 beg_pos,
798 position_);
799 }
800 } else {
801 return SlowScanJsonString<SeqOneByteString, uint8_t>(source_,
802 beg_pos,
803 position_);
804 }
805 } while (c0_ != '"');
806 int length = position_ - beg_pos;
807 Handle<String> result =
808 factory()->NewRawOneByteString(length, pretenure_).ToHandleChecked();
809 uint8_t* dest = SeqOneByteString::cast(*result)->GetChars();
810 String::WriteToFlat(*source_, dest, beg_pos, position_);
811
812 DCHECK_EQ('"', c0_);
813 // Advance past the last '"'.
814 AdvanceSkipWhitespace();
815 return result;
816 }
817
818 } } // namespace v8::internal
819
820 #endif // V8_JSON_PARSER_H_
821