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/messages.h"
13 #include "src/spaces-inl.h"
14 #include "src/token.h"
15
16 namespace v8 {
17 namespace internal {
18
19 // A simple json parser.
20 template <bool seq_ascii>
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 ASCII characters.
43 if (seq_ascii) {
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_ascii) {
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.IsAscii()) {
107 ASSERT_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-ASCII 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 Handle<String> source_;
186 int source_length_;
187 Handle<SeqOneByteString> seq_source_;
188
189 PretenureFlag pretenure_;
190 Isolate* isolate_;
191 Factory* factory_;
192 Zone zone_;
193 Handle<JSFunction> object_constructor_;
194 uc32 c0_;
195 int position_;
196 };
197
198 template <bool seq_ascii>
ParseJson()199 MaybeHandle<Object> JsonParser<seq_ascii>::ParseJson() {
200 // Advance to the first character (possibly EOS)
201 AdvanceSkipWhitespace();
202 Handle<Object> result = ParseJsonValue();
203 if (result.is_null() || c0_ != kEndOfString) {
204 // Some exception (for example stack overflow) is already pending.
205 if (isolate_->has_pending_exception()) return Handle<Object>::null();
206
207 // Parse failed. Current character is the unexpected token.
208 const char* message;
209 Factory* factory = this->factory();
210 Handle<JSArray> array;
211
212 switch (c0_) {
213 case kEndOfString:
214 message = "unexpected_eos";
215 array = factory->NewJSArray(0);
216 break;
217 case '-':
218 case '0':
219 case '1':
220 case '2':
221 case '3':
222 case '4':
223 case '5':
224 case '6':
225 case '7':
226 case '8':
227 case '9':
228 message = "unexpected_token_number";
229 array = factory->NewJSArray(0);
230 break;
231 case '"':
232 message = "unexpected_token_string";
233 array = factory->NewJSArray(0);
234 break;
235 default:
236 message = "unexpected_token";
237 Handle<Object> name = factory->LookupSingleCharacterStringFromCode(c0_);
238 Handle<FixedArray> element = factory->NewFixedArray(1);
239 element->set(0, *name);
240 array = factory->NewJSArrayWithElements(element);
241 break;
242 }
243
244 MessageLocation location(factory->NewScript(source_),
245 position_,
246 position_ + 1);
247 Handle<Object> error = factory->NewSyntaxError(message, array);
248 return isolate()->template Throw<Object>(error, &location);
249 }
250 return result;
251 }
252
253
254 // Parse any JSON value.
255 template <bool seq_ascii>
ParseJsonValue()256 Handle<Object> JsonParser<seq_ascii>::ParseJsonValue() {
257 StackLimitCheck stack_check(isolate_);
258 if (stack_check.HasOverflowed()) {
259 isolate_->StackOverflow();
260 return Handle<Object>::null();
261 }
262
263 if (c0_ == '"') return ParseJsonString();
264 if ((c0_ >= '0' && c0_ <= '9') || c0_ == '-') return ParseJsonNumber();
265 if (c0_ == '{') return ParseJsonObject();
266 if (c0_ == '[') return ParseJsonArray();
267 if (c0_ == 'f') {
268 if (AdvanceGetChar() == 'a' && AdvanceGetChar() == 'l' &&
269 AdvanceGetChar() == 's' && AdvanceGetChar() == 'e') {
270 AdvanceSkipWhitespace();
271 return factory()->false_value();
272 }
273 return ReportUnexpectedCharacter();
274 }
275 if (c0_ == 't') {
276 if (AdvanceGetChar() == 'r' && AdvanceGetChar() == 'u' &&
277 AdvanceGetChar() == 'e') {
278 AdvanceSkipWhitespace();
279 return factory()->true_value();
280 }
281 return ReportUnexpectedCharacter();
282 }
283 if (c0_ == 'n') {
284 if (AdvanceGetChar() == 'u' && AdvanceGetChar() == 'l' &&
285 AdvanceGetChar() == 'l') {
286 AdvanceSkipWhitespace();
287 return factory()->null_value();
288 }
289 return ReportUnexpectedCharacter();
290 }
291 return ReportUnexpectedCharacter();
292 }
293
294
295 // Parse a JSON object. Position must be right at '{'.
296 template <bool seq_ascii>
ParseJsonObject()297 Handle<Object> JsonParser<seq_ascii>::ParseJsonObject() {
298 HandleScope scope(isolate());
299 Handle<JSObject> json_object =
300 factory()->NewJSObject(object_constructor(), pretenure_);
301 Handle<Map> map(json_object->map());
302 ZoneList<Handle<Object> > properties(8, zone());
303 ASSERT_EQ(c0_, '{');
304
305 bool transitioning = true;
306
307 AdvanceSkipWhitespace();
308 if (c0_ != '}') {
309 do {
310 if (c0_ != '"') return ReportUnexpectedCharacter();
311
312 int start_position = position_;
313 Advance();
314
315 uint32_t index = 0;
316 if (c0_ >= '0' && c0_ <= '9') {
317 // Maybe an array index, try to parse it.
318 if (c0_ == '0') {
319 // With a leading zero, the string has to be "0" only to be an index.
320 Advance();
321 } else {
322 do {
323 int d = c0_ - '0';
324 if (index > 429496729U - ((d > 5) ? 1 : 0)) break;
325 index = (index * 10) + d;
326 Advance();
327 } while (c0_ >= '0' && c0_ <= '9');
328 }
329
330 if (c0_ == '"') {
331 // Successfully parsed index, parse and store element.
332 AdvanceSkipWhitespace();
333
334 if (c0_ != ':') return ReportUnexpectedCharacter();
335 AdvanceSkipWhitespace();
336 Handle<Object> value = ParseJsonValue();
337 if (value.is_null()) return ReportUnexpectedCharacter();
338
339 JSObject::SetOwnElement(json_object, index, value, SLOPPY).Assert();
340 continue;
341 }
342 // Not an index, fallback to the slow path.
343 }
344
345 position_ = start_position;
346 #ifdef DEBUG
347 c0_ = '"';
348 #endif
349
350 Handle<String> key;
351 Handle<Object> value;
352
353 // Try to follow existing transitions as long as possible. Once we stop
354 // transitioning, no transition can be found anymore.
355 if (transitioning) {
356 // First check whether there is a single expected transition. If so, try
357 // to parse it first.
358 bool follow_expected = false;
359 Handle<Map> target;
360 if (seq_ascii) {
361 key = JSObject::ExpectedTransitionKey(map);
362 follow_expected = !key.is_null() && ParseJsonString(key);
363 }
364 // If the expected transition hits, follow it.
365 if (follow_expected) {
366 target = JSObject::ExpectedTransitionTarget(map);
367 } else {
368 // If the expected transition failed, parse an internalized string and
369 // try to find a matching transition.
370 key = ParseJsonInternalizedString();
371 if (key.is_null()) return ReportUnexpectedCharacter();
372
373 target = JSObject::FindTransitionToField(map, key);
374 // If a transition was found, follow it and continue.
375 transitioning = !target.is_null();
376 }
377 if (c0_ != ':') return ReportUnexpectedCharacter();
378
379 AdvanceSkipWhitespace();
380 value = ParseJsonValue();
381 if (value.is_null()) return ReportUnexpectedCharacter();
382
383 if (transitioning) {
384 int descriptor = map->NumberOfOwnDescriptors();
385 PropertyDetails details =
386 target->instance_descriptors()->GetDetails(descriptor);
387 Representation expected_representation = details.representation();
388
389 if (value->FitsRepresentation(expected_representation)) {
390 // If the target representation is double and the value is already
391 // double, use the existing box.
392 if (value->IsSmi() && expected_representation.IsDouble()) {
393 value = factory()->NewHeapNumber(
394 Handle<Smi>::cast(value)->value());
395 } else if (expected_representation.IsHeapObject() &&
396 !target->instance_descriptors()->GetFieldType(
397 descriptor)->NowContains(value)) {
398 Handle<HeapType> value_type(value->OptimalType(
399 isolate(), expected_representation));
400 Map::GeneralizeFieldType(target, descriptor, value_type);
401 }
402 ASSERT(target->instance_descriptors()->GetFieldType(
403 descriptor)->NowContains(value));
404 properties.Add(value, zone());
405 map = target;
406 continue;
407 } else {
408 transitioning = false;
409 }
410 }
411
412 // Commit the intermediate state to the object and stop transitioning.
413 JSObject::AllocateStorageForMap(json_object, map);
414 int length = properties.length();
415 for (int i = 0; i < length; i++) {
416 Handle<Object> value = properties[i];
417 FieldIndex index = FieldIndex::ForPropertyIndex(*map, i);
418 json_object->FastPropertyAtPut(index, *value);
419 }
420 } else {
421 key = ParseJsonInternalizedString();
422 if (key.is_null() || c0_ != ':') return ReportUnexpectedCharacter();
423
424 AdvanceSkipWhitespace();
425 value = ParseJsonValue();
426 if (value.is_null()) return ReportUnexpectedCharacter();
427 }
428
429 JSObject::SetOwnPropertyIgnoreAttributes(
430 json_object, key, value, NONE).Assert();
431 } while (MatchSkipWhiteSpace(','));
432 if (c0_ != '}') {
433 return ReportUnexpectedCharacter();
434 }
435
436 // If we transitioned until the very end, transition the map now.
437 if (transitioning) {
438 JSObject::AllocateStorageForMap(json_object, map);
439 int length = properties.length();
440 for (int i = 0; i < length; i++) {
441 Handle<Object> value = properties[i];
442 FieldIndex index = FieldIndex::ForPropertyIndex(*map, i);
443 json_object->FastPropertyAtPut(index, *value);
444 }
445 }
446 }
447 AdvanceSkipWhitespace();
448 return scope.CloseAndEscape(json_object);
449 }
450
451 // Parse a JSON array. Position must be right at '['.
452 template <bool seq_ascii>
ParseJsonArray()453 Handle<Object> JsonParser<seq_ascii>::ParseJsonArray() {
454 HandleScope scope(isolate());
455 ZoneList<Handle<Object> > elements(4, zone());
456 ASSERT_EQ(c0_, '[');
457
458 AdvanceSkipWhitespace();
459 if (c0_ != ']') {
460 do {
461 Handle<Object> element = ParseJsonValue();
462 if (element.is_null()) return ReportUnexpectedCharacter();
463 elements.Add(element, zone());
464 } while (MatchSkipWhiteSpace(','));
465 if (c0_ != ']') {
466 return ReportUnexpectedCharacter();
467 }
468 }
469 AdvanceSkipWhitespace();
470 // Allocate a fixed array with all the elements.
471 Handle<FixedArray> fast_elements =
472 factory()->NewFixedArray(elements.length(), pretenure_);
473 for (int i = 0, n = elements.length(); i < n; i++) {
474 fast_elements->set(i, *elements[i]);
475 }
476 Handle<Object> json_array = factory()->NewJSArrayWithElements(
477 fast_elements, FAST_ELEMENTS, pretenure_);
478 return scope.CloseAndEscape(json_array);
479 }
480
481
482 template <bool seq_ascii>
ParseJsonNumber()483 Handle<Object> JsonParser<seq_ascii>::ParseJsonNumber() {
484 bool negative = false;
485 int beg_pos = position_;
486 if (c0_ == '-') {
487 Advance();
488 negative = true;
489 }
490 if (c0_ == '0') {
491 Advance();
492 // Prefix zero is only allowed if it's the only digit before
493 // a decimal point or exponent.
494 if ('0' <= c0_ && c0_ <= '9') return ReportUnexpectedCharacter();
495 } else {
496 int i = 0;
497 int digits = 0;
498 if (c0_ < '1' || c0_ > '9') return ReportUnexpectedCharacter();
499 do {
500 i = i * 10 + c0_ - '0';
501 digits++;
502 Advance();
503 } while (c0_ >= '0' && c0_ <= '9');
504 if (c0_ != '.' && c0_ != 'e' && c0_ != 'E' && digits < 10) {
505 SkipWhitespace();
506 return Handle<Smi>(Smi::FromInt((negative ? -i : i)), isolate());
507 }
508 }
509 if (c0_ == '.') {
510 Advance();
511 if (c0_ < '0' || c0_ > '9') return ReportUnexpectedCharacter();
512 do {
513 Advance();
514 } while (c0_ >= '0' && c0_ <= '9');
515 }
516 if (AsciiAlphaToLower(c0_) == 'e') {
517 Advance();
518 if (c0_ == '-' || c0_ == '+') Advance();
519 if (c0_ < '0' || c0_ > '9') return ReportUnexpectedCharacter();
520 do {
521 Advance();
522 } while (c0_ >= '0' && c0_ <= '9');
523 }
524 int length = position_ - beg_pos;
525 double number;
526 if (seq_ascii) {
527 Vector<const uint8_t> chars(seq_source_->GetChars() + beg_pos, length);
528 number = StringToDouble(isolate()->unicode_cache(),
529 chars,
530 NO_FLAGS, // Hex, octal or trailing junk.
531 OS::nan_value());
532 } else {
533 Vector<uint8_t> buffer = Vector<uint8_t>::New(length);
534 String::WriteToFlat(*source_, buffer.start(), beg_pos, position_);
535 Vector<const uint8_t> result =
536 Vector<const uint8_t>(buffer.start(), length);
537 number = StringToDouble(isolate()->unicode_cache(),
538 result,
539 NO_FLAGS, // Hex, octal or trailing junk.
540 0.0);
541 buffer.Dispose();
542 }
543 SkipWhitespace();
544 return factory()->NewNumber(number, pretenure_);
545 }
546
547
548 template <typename StringType>
549 inline void SeqStringSet(Handle<StringType> seq_str, int i, uc32 c);
550
551 template <>
SeqStringSet(Handle<SeqTwoByteString> seq_str,int i,uc32 c)552 inline void SeqStringSet(Handle<SeqTwoByteString> seq_str, int i, uc32 c) {
553 seq_str->SeqTwoByteStringSet(i, c);
554 }
555
556 template <>
SeqStringSet(Handle<SeqOneByteString> seq_str,int i,uc32 c)557 inline void SeqStringSet(Handle<SeqOneByteString> seq_str, int i, uc32 c) {
558 seq_str->SeqOneByteStringSet(i, c);
559 }
560
561 template <typename StringType>
562 inline Handle<StringType> NewRawString(Factory* factory,
563 int length,
564 PretenureFlag pretenure);
565
566 template <>
NewRawString(Factory * factory,int length,PretenureFlag pretenure)567 inline Handle<SeqTwoByteString> NewRawString(Factory* factory,
568 int length,
569 PretenureFlag pretenure) {
570 return factory->NewRawTwoByteString(length, pretenure).ToHandleChecked();
571 }
572
573 template <>
NewRawString(Factory * factory,int length,PretenureFlag pretenure)574 inline Handle<SeqOneByteString> NewRawString(Factory* factory,
575 int length,
576 PretenureFlag pretenure) {
577 return factory->NewRawOneByteString(length, pretenure).ToHandleChecked();
578 }
579
580
581 // Scans the rest of a JSON string starting from position_ and writes
582 // prefix[start..end] along with the scanned characters into a
583 // sequential string of type StringType.
584 template <bool seq_ascii>
585 template <typename StringType, typename SinkChar>
SlowScanJsonString(Handle<String> prefix,int start,int end)586 Handle<String> JsonParser<seq_ascii>::SlowScanJsonString(
587 Handle<String> prefix, int start, int end) {
588 int count = end - start;
589 int max_length = count + source_length_ - position_;
590 int length = Min(max_length, Max(kInitialSpecialStringLength, 2 * count));
591 Handle<StringType> seq_string =
592 NewRawString<StringType>(factory(), length, pretenure_);
593 // Copy prefix into seq_str.
594 SinkChar* dest = seq_string->GetChars();
595 String::WriteToFlat(*prefix, dest, start, end);
596
597 while (c0_ != '"') {
598 // Check for control character (0x00-0x1f) or unterminated string (<0).
599 if (c0_ < 0x20) return Handle<String>::null();
600 if (count >= length) {
601 // We need to create a longer sequential string for the result.
602 return SlowScanJsonString<StringType, SinkChar>(seq_string, 0, count);
603 }
604 if (c0_ != '\\') {
605 // If the sink can contain UC16 characters, or source_ contains only
606 // ASCII characters, there's no need to test whether we can store the
607 // character. Otherwise check whether the UC16 source character can fit
608 // in the ASCII sink.
609 if (sizeof(SinkChar) == kUC16Size ||
610 seq_ascii ||
611 c0_ <= String::kMaxOneByteCharCode) {
612 SeqStringSet(seq_string, count++, c0_);
613 Advance();
614 } else {
615 // StringType is SeqOneByteString and we just read a non-ASCII char.
616 return SlowScanJsonString<SeqTwoByteString, uc16>(seq_string, 0, count);
617 }
618 } else {
619 Advance(); // Advance past the \.
620 switch (c0_) {
621 case '"':
622 case '\\':
623 case '/':
624 SeqStringSet(seq_string, count++, c0_);
625 break;
626 case 'b':
627 SeqStringSet(seq_string, count++, '\x08');
628 break;
629 case 'f':
630 SeqStringSet(seq_string, count++, '\x0c');
631 break;
632 case 'n':
633 SeqStringSet(seq_string, count++, '\x0a');
634 break;
635 case 'r':
636 SeqStringSet(seq_string, count++, '\x0d');
637 break;
638 case 't':
639 SeqStringSet(seq_string, count++, '\x09');
640 break;
641 case 'u': {
642 uc32 value = 0;
643 for (int i = 0; i < 4; i++) {
644 Advance();
645 int digit = HexValue(c0_);
646 if (digit < 0) {
647 return Handle<String>::null();
648 }
649 value = value * 16 + digit;
650 }
651 if (sizeof(SinkChar) == kUC16Size ||
652 value <= String::kMaxOneByteCharCode) {
653 SeqStringSet(seq_string, count++, value);
654 break;
655 } else {
656 // StringType is SeqOneByteString and we just read a non-ASCII char.
657 position_ -= 6; // Rewind position_ to \ in \uxxxx.
658 Advance();
659 return SlowScanJsonString<SeqTwoByteString, uc16>(seq_string,
660 0,
661 count);
662 }
663 }
664 default:
665 return Handle<String>::null();
666 }
667 Advance();
668 }
669 }
670
671 ASSERT_EQ('"', c0_);
672 // Advance past the last '"'.
673 AdvanceSkipWhitespace();
674
675 // Shrink seq_string length to count and return.
676 return SeqString::Truncate(seq_string, count);
677 }
678
679
680 template <bool seq_ascii>
681 template <bool is_internalized>
ScanJsonString()682 Handle<String> JsonParser<seq_ascii>::ScanJsonString() {
683 ASSERT_EQ('"', c0_);
684 Advance();
685 if (c0_ == '"') {
686 AdvanceSkipWhitespace();
687 return factory()->empty_string();
688 }
689
690 if (seq_ascii && is_internalized) {
691 // Fast path for existing internalized strings. If the the string being
692 // parsed is not a known internalized string, contains backslashes or
693 // unexpectedly reaches the end of string, return with an empty handle.
694 uint32_t running_hash = isolate()->heap()->HashSeed();
695 int position = position_;
696 uc32 c0 = c0_;
697 do {
698 if (c0 == '\\') {
699 c0_ = c0;
700 int beg_pos = position_;
701 position_ = position;
702 return SlowScanJsonString<SeqOneByteString, uint8_t>(source_,
703 beg_pos,
704 position_);
705 }
706 if (c0 < 0x20) return Handle<String>::null();
707 if (static_cast<uint32_t>(c0) >
708 unibrow::Utf16::kMaxNonSurrogateCharCode) {
709 running_hash =
710 StringHasher::AddCharacterCore(running_hash,
711 unibrow::Utf16::LeadSurrogate(c0));
712 running_hash =
713 StringHasher::AddCharacterCore(running_hash,
714 unibrow::Utf16::TrailSurrogate(c0));
715 } else {
716 running_hash = StringHasher::AddCharacterCore(running_hash, c0);
717 }
718 position++;
719 if (position >= source_length_) return Handle<String>::null();
720 c0 = seq_source_->SeqOneByteStringGet(position);
721 } while (c0 != '"');
722 int length = position - position_;
723 uint32_t hash = (length <= String::kMaxHashCalcLength)
724 ? StringHasher::GetHashCore(running_hash) : length;
725 Vector<const uint8_t> string_vector(
726 seq_source_->GetChars() + position_, length);
727 StringTable* string_table = isolate()->heap()->string_table();
728 uint32_t capacity = string_table->Capacity();
729 uint32_t entry = StringTable::FirstProbe(hash, capacity);
730 uint32_t count = 1;
731 Handle<String> result;
732 while (true) {
733 Object* element = string_table->KeyAt(entry);
734 if (element == isolate()->heap()->undefined_value()) {
735 // Lookup failure.
736 result = factory()->InternalizeOneByteString(
737 seq_source_, position_, length);
738 break;
739 }
740 if (element != isolate()->heap()->the_hole_value() &&
741 String::cast(element)->IsOneByteEqualTo(string_vector)) {
742 result = Handle<String>(String::cast(element), isolate());
743 #ifdef DEBUG
744 uint32_t hash_field =
745 (hash << String::kHashShift) | String::kIsNotArrayIndexMask;
746 ASSERT_EQ(static_cast<int>(result->Hash()),
747 static_cast<int>(hash_field >> String::kHashShift));
748 #endif
749 break;
750 }
751 entry = StringTable::NextProbe(entry, count++, capacity);
752 }
753 position_ = position;
754 // Advance past the last '"'.
755 AdvanceSkipWhitespace();
756 return result;
757 }
758
759 int beg_pos = position_;
760 // Fast case for ASCII only without escape characters.
761 do {
762 // Check for control character (0x00-0x1f) or unterminated string (<0).
763 if (c0_ < 0x20) return Handle<String>::null();
764 if (c0_ != '\\') {
765 if (seq_ascii || c0_ <= String::kMaxOneByteCharCode) {
766 Advance();
767 } else {
768 return SlowScanJsonString<SeqTwoByteString, uc16>(source_,
769 beg_pos,
770 position_);
771 }
772 } else {
773 return SlowScanJsonString<SeqOneByteString, uint8_t>(source_,
774 beg_pos,
775 position_);
776 }
777 } while (c0_ != '"');
778 int length = position_ - beg_pos;
779 Handle<String> result =
780 factory()->NewRawOneByteString(length, pretenure_).ToHandleChecked();
781 uint8_t* dest = SeqOneByteString::cast(*result)->GetChars();
782 String::WriteToFlat(*source_, dest, beg_pos, position_);
783
784 ASSERT_EQ('"', c0_);
785 // Advance past the last '"'.
786 AdvanceSkipWhitespace();
787 return result;
788 }
789
790 } } // namespace v8::internal
791
792 #endif // V8_JSON_PARSER_H_
793