Ben Murdoch | 4a90d5f | 2016-03-22 12:00:34 +0000 | [diff] [blame^] | 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_PARSING_JSON_PARSER_H_ |
| 6 | #define V8_PARSING_JSON_PARSER_H_ |
| 7 | |
| 8 | #include "src/char-predicates.h" |
| 9 | #include "src/conversions.h" |
| 10 | #include "src/debug/debug.h" |
| 11 | #include "src/factory.h" |
| 12 | #include "src/messages.h" |
| 13 | #include "src/parsing/scanner.h" |
| 14 | #include "src/parsing/token.h" |
| 15 | #include "src/transitions.h" |
| 16 | #include "src/types.h" |
| 17 | |
| 18 | namespace v8 { |
| 19 | namespace internal { |
| 20 | |
| 21 | enum ParseElementResult { kElementFound, kElementNotFound, kNullHandle }; |
| 22 | |
| 23 | |
| 24 | // A simple json parser. |
| 25 | template <bool seq_one_byte> |
| 26 | class JsonParser BASE_EMBEDDED { |
| 27 | public: |
| 28 | MUST_USE_RESULT static MaybeHandle<Object> Parse(Handle<String> source) { |
| 29 | return JsonParser(source).ParseJson(); |
| 30 | } |
| 31 | |
| 32 | static const int kEndOfString = -1; |
| 33 | |
| 34 | private: |
| 35 | explicit JsonParser(Handle<String> source) |
| 36 | : source_(source), |
| 37 | source_length_(source->length()), |
| 38 | isolate_(source->map()->GetHeap()->isolate()), |
| 39 | factory_(isolate_->factory()), |
| 40 | object_constructor_(isolate_->native_context()->object_function(), |
| 41 | isolate_), |
| 42 | position_(-1) { |
| 43 | source_ = String::Flatten(source_); |
| 44 | pretenure_ = (source_length_ >= kPretenureTreshold) ? TENURED : NOT_TENURED; |
| 45 | |
| 46 | // Optimized fast case where we only have Latin1 characters. |
| 47 | if (seq_one_byte) { |
| 48 | seq_source_ = Handle<SeqOneByteString>::cast(source_); |
| 49 | } |
| 50 | } |
| 51 | |
| 52 | // Parse a string containing a single JSON value. |
| 53 | MaybeHandle<Object> ParseJson(); |
| 54 | |
| 55 | inline void Advance() { |
| 56 | position_++; |
| 57 | if (position_ >= source_length_) { |
| 58 | c0_ = kEndOfString; |
| 59 | } else if (seq_one_byte) { |
| 60 | c0_ = seq_source_->SeqOneByteStringGet(position_); |
| 61 | } else { |
| 62 | c0_ = source_->Get(position_); |
| 63 | } |
| 64 | } |
| 65 | |
| 66 | // The JSON lexical grammar is specified in the ECMAScript 5 standard, |
| 67 | // section 15.12.1.1. The only allowed whitespace characters between tokens |
| 68 | // are tab, carriage-return, newline and space. |
| 69 | |
| 70 | inline void AdvanceSkipWhitespace() { |
| 71 | do { |
| 72 | Advance(); |
| 73 | } while (c0_ == ' ' || c0_ == '\t' || c0_ == '\n' || c0_ == '\r'); |
| 74 | } |
| 75 | |
| 76 | inline void SkipWhitespace() { |
| 77 | while (c0_ == ' ' || c0_ == '\t' || c0_ == '\n' || c0_ == '\r') { |
| 78 | Advance(); |
| 79 | } |
| 80 | } |
| 81 | |
| 82 | inline uc32 AdvanceGetChar() { |
| 83 | Advance(); |
| 84 | return c0_; |
| 85 | } |
| 86 | |
| 87 | // Checks that current charater is c. |
| 88 | // If so, then consume c and skip whitespace. |
| 89 | inline bool MatchSkipWhiteSpace(uc32 c) { |
| 90 | if (c0_ == c) { |
| 91 | AdvanceSkipWhitespace(); |
| 92 | return true; |
| 93 | } |
| 94 | return false; |
| 95 | } |
| 96 | |
| 97 | // A JSON string (production JSONString) is subset of valid JavaScript string |
| 98 | // literals. The string must only be double-quoted (not single-quoted), and |
| 99 | // the only allowed backslash-escapes are ", /, \, b, f, n, r, t and |
| 100 | // four-digit hex escapes (uXXXX). Any other use of backslashes is invalid. |
| 101 | Handle<String> ParseJsonString() { |
| 102 | return ScanJsonString<false>(); |
| 103 | } |
| 104 | |
| 105 | bool ParseJsonString(Handle<String> expected) { |
| 106 | int length = expected->length(); |
| 107 | if (source_->length() - position_ - 1 > length) { |
| 108 | DisallowHeapAllocation no_gc; |
| 109 | String::FlatContent content = expected->GetFlatContent(); |
| 110 | if (content.IsOneByte()) { |
| 111 | DCHECK_EQ('"', c0_); |
| 112 | const uint8_t* input_chars = seq_source_->GetChars() + position_ + 1; |
| 113 | const uint8_t* expected_chars = content.ToOneByteVector().start(); |
| 114 | for (int i = 0; i < length; i++) { |
| 115 | uint8_t c0 = input_chars[i]; |
| 116 | if (c0 != expected_chars[i] || c0 == '"' || c0 < 0x20 || c0 == '\\') { |
| 117 | return false; |
| 118 | } |
| 119 | } |
| 120 | if (input_chars[length] == '"') { |
| 121 | position_ = position_ + length + 1; |
| 122 | AdvanceSkipWhitespace(); |
| 123 | return true; |
| 124 | } |
| 125 | } |
| 126 | } |
| 127 | return false; |
| 128 | } |
| 129 | |
| 130 | Handle<String> ParseJsonInternalizedString() { |
| 131 | return ScanJsonString<true>(); |
| 132 | } |
| 133 | |
| 134 | template <bool is_internalized> |
| 135 | Handle<String> ScanJsonString(); |
| 136 | // Creates a new string and copies prefix[start..end] into the beginning |
| 137 | // of it. Then scans the rest of the string, adding characters after the |
| 138 | // prefix. Called by ScanJsonString when reaching a '\' or non-Latin1 char. |
| 139 | template <typename StringType, typename SinkChar> |
| 140 | Handle<String> SlowScanJsonString(Handle<String> prefix, int start, int end); |
| 141 | |
| 142 | // A JSON number (production JSONNumber) is a subset of the valid JavaScript |
| 143 | // decimal number literals. |
| 144 | // It includes an optional minus sign, must have at least one |
| 145 | // digit before and after a decimal point, may not have prefixed zeros (unless |
| 146 | // the integer part is zero), and may include an exponent part (e.g., "e-10"). |
| 147 | // Hexadecimal and octal numbers are not allowed. |
| 148 | Handle<Object> ParseJsonNumber(); |
| 149 | |
| 150 | // Parse a single JSON value from input (grammar production JSONValue). |
| 151 | // A JSON value is either a (double-quoted) string literal, a number literal, |
| 152 | // one of "true", "false", or "null", or an object or array literal. |
| 153 | Handle<Object> ParseJsonValue(); |
| 154 | |
| 155 | // Parse a JSON object literal (grammar production JSONObject). |
| 156 | // An object literal is a squiggly-braced and comma separated sequence |
| 157 | // (possibly empty) of key/value pairs, where the key is a JSON string |
| 158 | // literal, the value is a JSON value, and the two are separated by a colon. |
| 159 | // A JSON array doesn't allow numbers and identifiers as keys, like a |
| 160 | // JavaScript array. |
| 161 | Handle<Object> ParseJsonObject(); |
| 162 | |
| 163 | // Helper for ParseJsonObject. Parses the form "123": obj, which is recorded |
| 164 | // as an element, not a property. |
| 165 | ParseElementResult ParseElement(Handle<JSObject> json_object); |
| 166 | |
| 167 | // Parses a JSON array literal (grammar production JSONArray). An array |
| 168 | // literal is a square-bracketed and comma separated sequence (possibly empty) |
| 169 | // of JSON values. |
| 170 | // A JSON array doesn't allow leaving out values from the sequence, nor does |
| 171 | // it allow a terminal comma, like a JavaScript array does. |
| 172 | Handle<Object> ParseJsonArray(); |
| 173 | |
| 174 | |
| 175 | // Mark that a parsing error has happened at the current token, and |
| 176 | // return a null handle. Primarily for readability. |
| 177 | inline Handle<Object> ReportUnexpectedCharacter() { |
| 178 | return Handle<Object>::null(); |
| 179 | } |
| 180 | |
| 181 | inline Isolate* isolate() { return isolate_; } |
| 182 | inline Factory* factory() { return factory_; } |
| 183 | inline Handle<JSFunction> object_constructor() { return object_constructor_; } |
| 184 | |
| 185 | static const int kInitialSpecialStringLength = 32; |
| 186 | static const int kPretenureTreshold = 100 * 1024; |
| 187 | |
| 188 | |
| 189 | private: |
| 190 | Zone* zone() { return &zone_; } |
| 191 | |
| 192 | void CommitStateToJsonObject(Handle<JSObject> json_object, Handle<Map> map, |
| 193 | ZoneList<Handle<Object> >* properties); |
| 194 | |
| 195 | Handle<String> source_; |
| 196 | int source_length_; |
| 197 | Handle<SeqOneByteString> seq_source_; |
| 198 | |
| 199 | PretenureFlag pretenure_; |
| 200 | Isolate* isolate_; |
| 201 | Factory* factory_; |
| 202 | Zone zone_; |
| 203 | Handle<JSFunction> object_constructor_; |
| 204 | uc32 c0_; |
| 205 | int position_; |
| 206 | }; |
| 207 | |
| 208 | template <bool seq_one_byte> |
| 209 | MaybeHandle<Object> JsonParser<seq_one_byte>::ParseJson() { |
| 210 | // Advance to the first character (possibly EOS) |
| 211 | AdvanceSkipWhitespace(); |
| 212 | Handle<Object> result = ParseJsonValue(); |
| 213 | if (result.is_null() || c0_ != kEndOfString) { |
| 214 | // Some exception (for example stack overflow) is already pending. |
| 215 | if (isolate_->has_pending_exception()) return Handle<Object>::null(); |
| 216 | |
| 217 | // Parse failed. Current character is the unexpected token. |
| 218 | Factory* factory = this->factory(); |
| 219 | MessageTemplate::Template message; |
| 220 | Handle<String> argument; |
| 221 | |
| 222 | switch (c0_) { |
| 223 | case kEndOfString: |
| 224 | message = MessageTemplate::kUnexpectedEOS; |
| 225 | break; |
| 226 | case '-': |
| 227 | case '0': |
| 228 | case '1': |
| 229 | case '2': |
| 230 | case '3': |
| 231 | case '4': |
| 232 | case '5': |
| 233 | case '6': |
| 234 | case '7': |
| 235 | case '8': |
| 236 | case '9': |
| 237 | message = MessageTemplate::kUnexpectedTokenNumber; |
| 238 | break; |
| 239 | case '"': |
| 240 | message = MessageTemplate::kUnexpectedTokenString; |
| 241 | break; |
| 242 | default: |
| 243 | message = MessageTemplate::kUnexpectedToken; |
| 244 | argument = factory->LookupSingleCharacterStringFromCode(c0_); |
| 245 | break; |
| 246 | } |
| 247 | |
| 248 | Handle<Script> script(factory->NewScript(source_)); |
| 249 | // We should sent compile error event because we compile JSON object in |
| 250 | // separated source file. |
| 251 | isolate()->debug()->OnCompileError(script); |
| 252 | MessageLocation location(script, position_, position_ + 1); |
| 253 | Handle<Object> error = factory->NewSyntaxError(message, argument); |
| 254 | return isolate()->template Throw<Object>(error, &location); |
| 255 | } |
| 256 | return result; |
| 257 | } |
| 258 | |
| 259 | |
| 260 | // Parse any JSON value. |
| 261 | template <bool seq_one_byte> |
| 262 | Handle<Object> JsonParser<seq_one_byte>::ParseJsonValue() { |
| 263 | StackLimitCheck stack_check(isolate_); |
| 264 | if (stack_check.HasOverflowed()) { |
| 265 | isolate_->StackOverflow(); |
| 266 | return Handle<Object>::null(); |
| 267 | } |
| 268 | |
| 269 | if (stack_check.InterruptRequested()) { |
| 270 | ExecutionAccess access(isolate_); |
| 271 | // Avoid blocking GC in long running parser (v8:3974). |
| 272 | isolate_->stack_guard()->HandleGCInterrupt(); |
| 273 | } |
| 274 | |
| 275 | if (c0_ == '"') return ParseJsonString(); |
| 276 | if ((c0_ >= '0' && c0_ <= '9') || c0_ == '-') return ParseJsonNumber(); |
| 277 | if (c0_ == '{') return ParseJsonObject(); |
| 278 | if (c0_ == '[') return ParseJsonArray(); |
| 279 | if (c0_ == 'f') { |
| 280 | if (AdvanceGetChar() == 'a' && AdvanceGetChar() == 'l' && |
| 281 | AdvanceGetChar() == 's' && AdvanceGetChar() == 'e') { |
| 282 | AdvanceSkipWhitespace(); |
| 283 | return factory()->false_value(); |
| 284 | } |
| 285 | return ReportUnexpectedCharacter(); |
| 286 | } |
| 287 | if (c0_ == 't') { |
| 288 | if (AdvanceGetChar() == 'r' && AdvanceGetChar() == 'u' && |
| 289 | AdvanceGetChar() == 'e') { |
| 290 | AdvanceSkipWhitespace(); |
| 291 | return factory()->true_value(); |
| 292 | } |
| 293 | return ReportUnexpectedCharacter(); |
| 294 | } |
| 295 | if (c0_ == 'n') { |
| 296 | if (AdvanceGetChar() == 'u' && AdvanceGetChar() == 'l' && |
| 297 | AdvanceGetChar() == 'l') { |
| 298 | AdvanceSkipWhitespace(); |
| 299 | return factory()->null_value(); |
| 300 | } |
| 301 | return ReportUnexpectedCharacter(); |
| 302 | } |
| 303 | return ReportUnexpectedCharacter(); |
| 304 | } |
| 305 | |
| 306 | |
| 307 | template <bool seq_one_byte> |
| 308 | ParseElementResult JsonParser<seq_one_byte>::ParseElement( |
| 309 | Handle<JSObject> json_object) { |
| 310 | uint32_t index = 0; |
| 311 | // Maybe an array index, try to parse it. |
| 312 | if (c0_ == '0') { |
| 313 | // With a leading zero, the string has to be "0" only to be an index. |
| 314 | Advance(); |
| 315 | } else { |
| 316 | do { |
| 317 | int d = c0_ - '0'; |
| 318 | if (index > 429496729U - ((d + 3) >> 3)) break; |
| 319 | index = (index * 10) + d; |
| 320 | Advance(); |
| 321 | } while (IsDecimalDigit(c0_)); |
| 322 | } |
| 323 | |
| 324 | if (c0_ == '"') { |
| 325 | // Successfully parsed index, parse and store element. |
| 326 | AdvanceSkipWhitespace(); |
| 327 | |
| 328 | if (c0_ == ':') { |
| 329 | AdvanceSkipWhitespace(); |
| 330 | Handle<Object> value = ParseJsonValue(); |
| 331 | if (!value.is_null()) { |
| 332 | JSObject::SetOwnElementIgnoreAttributes(json_object, index, value, NONE) |
| 333 | .Assert(); |
| 334 | return kElementFound; |
| 335 | } else { |
| 336 | return kNullHandle; |
| 337 | } |
| 338 | } |
| 339 | } |
| 340 | return kElementNotFound; |
| 341 | } |
| 342 | |
| 343 | // Parse a JSON object. Position must be right at '{'. |
| 344 | template <bool seq_one_byte> |
| 345 | Handle<Object> JsonParser<seq_one_byte>::ParseJsonObject() { |
| 346 | HandleScope scope(isolate()); |
| 347 | Handle<JSObject> json_object = |
| 348 | factory()->NewJSObject(object_constructor(), pretenure_); |
| 349 | Handle<Map> map(json_object->map()); |
| 350 | int descriptor = 0; |
| 351 | ZoneList<Handle<Object> > properties(8, zone()); |
| 352 | DCHECK_EQ(c0_, '{'); |
| 353 | |
| 354 | bool transitioning = true; |
| 355 | |
| 356 | AdvanceSkipWhitespace(); |
| 357 | if (c0_ != '}') { |
| 358 | do { |
| 359 | if (c0_ != '"') return ReportUnexpectedCharacter(); |
| 360 | |
| 361 | int start_position = position_; |
| 362 | Advance(); |
| 363 | |
| 364 | if (IsDecimalDigit(c0_)) { |
| 365 | ParseElementResult element_result = ParseElement(json_object); |
| 366 | if (element_result == kNullHandle) return Handle<Object>::null(); |
| 367 | if (element_result == kElementFound) continue; |
| 368 | } |
| 369 | // Not an index, fallback to the slow path. |
| 370 | |
| 371 | position_ = start_position; |
| 372 | #ifdef DEBUG |
| 373 | c0_ = '"'; |
| 374 | #endif |
| 375 | |
| 376 | Handle<String> key; |
| 377 | Handle<Object> value; |
| 378 | |
| 379 | // Try to follow existing transitions as long as possible. Once we stop |
| 380 | // transitioning, no transition can be found anymore. |
| 381 | DCHECK(transitioning); |
| 382 | // First check whether there is a single expected transition. If so, try |
| 383 | // to parse it first. |
| 384 | bool follow_expected = false; |
| 385 | Handle<Map> target; |
| 386 | if (seq_one_byte) { |
| 387 | key = TransitionArray::ExpectedTransitionKey(map); |
| 388 | follow_expected = !key.is_null() && ParseJsonString(key); |
| 389 | } |
| 390 | // If the expected transition hits, follow it. |
| 391 | if (follow_expected) { |
| 392 | target = TransitionArray::ExpectedTransitionTarget(map); |
| 393 | } else { |
| 394 | // If the expected transition failed, parse an internalized string and |
| 395 | // try to find a matching transition. |
| 396 | key = ParseJsonInternalizedString(); |
| 397 | if (key.is_null()) return ReportUnexpectedCharacter(); |
| 398 | |
| 399 | target = TransitionArray::FindTransitionToField(map, key); |
| 400 | // If a transition was found, follow it and continue. |
| 401 | transitioning = !target.is_null(); |
| 402 | } |
| 403 | if (c0_ != ':') return ReportUnexpectedCharacter(); |
| 404 | |
| 405 | AdvanceSkipWhitespace(); |
| 406 | value = ParseJsonValue(); |
| 407 | if (value.is_null()) return ReportUnexpectedCharacter(); |
| 408 | |
| 409 | if (transitioning) { |
| 410 | PropertyDetails details = |
| 411 | target->instance_descriptors()->GetDetails(descriptor); |
| 412 | Representation expected_representation = details.representation(); |
| 413 | |
| 414 | if (value->FitsRepresentation(expected_representation)) { |
| 415 | if (expected_representation.IsHeapObject() && |
| 416 | !target->instance_descriptors() |
| 417 | ->GetFieldType(descriptor) |
| 418 | ->NowContains(value)) { |
| 419 | Handle<HeapType> value_type( |
| 420 | value->OptimalType(isolate(), expected_representation)); |
| 421 | Map::GeneralizeFieldType(target, descriptor, |
| 422 | expected_representation, value_type); |
| 423 | } |
| 424 | DCHECK(target->instance_descriptors() |
| 425 | ->GetFieldType(descriptor) |
| 426 | ->NowContains(value)); |
| 427 | properties.Add(value, zone()); |
| 428 | map = target; |
| 429 | descriptor++; |
| 430 | continue; |
| 431 | } else { |
| 432 | transitioning = false; |
| 433 | } |
| 434 | } |
| 435 | |
| 436 | DCHECK(!transitioning); |
| 437 | |
| 438 | // Commit the intermediate state to the object and stop transitioning. |
| 439 | CommitStateToJsonObject(json_object, map, &properties); |
| 440 | |
| 441 | JSObject::DefinePropertyOrElementIgnoreAttributes(json_object, key, value) |
| 442 | .Check(); |
| 443 | } while (transitioning && MatchSkipWhiteSpace(',')); |
| 444 | |
| 445 | // If we transitioned until the very end, transition the map now. |
| 446 | if (transitioning) { |
| 447 | CommitStateToJsonObject(json_object, map, &properties); |
| 448 | } else { |
| 449 | while (MatchSkipWhiteSpace(',')) { |
| 450 | HandleScope local_scope(isolate()); |
| 451 | if (c0_ != '"') return ReportUnexpectedCharacter(); |
| 452 | |
| 453 | int start_position = position_; |
| 454 | Advance(); |
| 455 | |
| 456 | if (IsDecimalDigit(c0_)) { |
| 457 | ParseElementResult element_result = ParseElement(json_object); |
| 458 | if (element_result == kNullHandle) return Handle<Object>::null(); |
| 459 | if (element_result == kElementFound) continue; |
| 460 | } |
| 461 | // Not an index, fallback to the slow path. |
| 462 | |
| 463 | position_ = start_position; |
| 464 | #ifdef DEBUG |
| 465 | c0_ = '"'; |
| 466 | #endif |
| 467 | |
| 468 | Handle<String> key; |
| 469 | Handle<Object> value; |
| 470 | |
| 471 | key = ParseJsonInternalizedString(); |
| 472 | if (key.is_null() || c0_ != ':') return ReportUnexpectedCharacter(); |
| 473 | |
| 474 | AdvanceSkipWhitespace(); |
| 475 | value = ParseJsonValue(); |
| 476 | if (value.is_null()) return ReportUnexpectedCharacter(); |
| 477 | |
| 478 | JSObject::DefinePropertyOrElementIgnoreAttributes(json_object, key, |
| 479 | value).Check(); |
| 480 | } |
| 481 | } |
| 482 | |
| 483 | if (c0_ != '}') { |
| 484 | return ReportUnexpectedCharacter(); |
| 485 | } |
| 486 | } |
| 487 | AdvanceSkipWhitespace(); |
| 488 | return scope.CloseAndEscape(json_object); |
| 489 | } |
| 490 | |
| 491 | |
| 492 | template <bool seq_one_byte> |
| 493 | void JsonParser<seq_one_byte>::CommitStateToJsonObject( |
| 494 | Handle<JSObject> json_object, Handle<Map> map, |
| 495 | ZoneList<Handle<Object> >* properties) { |
| 496 | JSObject::AllocateStorageForMap(json_object, map); |
| 497 | DCHECK(!json_object->map()->is_dictionary_map()); |
| 498 | |
| 499 | DisallowHeapAllocation no_gc; |
| 500 | |
| 501 | int length = properties->length(); |
| 502 | for (int i = 0; i < length; i++) { |
| 503 | Handle<Object> value = (*properties)[i]; |
| 504 | json_object->WriteToField(i, *value); |
| 505 | } |
| 506 | } |
| 507 | |
| 508 | |
| 509 | // Parse a JSON array. Position must be right at '['. |
| 510 | template <bool seq_one_byte> |
| 511 | Handle<Object> JsonParser<seq_one_byte>::ParseJsonArray() { |
| 512 | HandleScope scope(isolate()); |
| 513 | ZoneList<Handle<Object> > elements(4, zone()); |
| 514 | DCHECK_EQ(c0_, '['); |
| 515 | |
| 516 | AdvanceSkipWhitespace(); |
| 517 | if (c0_ != ']') { |
| 518 | do { |
| 519 | Handle<Object> element = ParseJsonValue(); |
| 520 | if (element.is_null()) return ReportUnexpectedCharacter(); |
| 521 | elements.Add(element, zone()); |
| 522 | } while (MatchSkipWhiteSpace(',')); |
| 523 | if (c0_ != ']') { |
| 524 | return ReportUnexpectedCharacter(); |
| 525 | } |
| 526 | } |
| 527 | AdvanceSkipWhitespace(); |
| 528 | // Allocate a fixed array with all the elements. |
| 529 | Handle<FixedArray> fast_elements = |
| 530 | factory()->NewFixedArray(elements.length(), pretenure_); |
| 531 | for (int i = 0, n = elements.length(); i < n; i++) { |
| 532 | fast_elements->set(i, *elements[i]); |
| 533 | } |
| 534 | Handle<Object> json_array = factory()->NewJSArrayWithElements( |
| 535 | fast_elements, FAST_ELEMENTS, Strength::WEAK, pretenure_); |
| 536 | return scope.CloseAndEscape(json_array); |
| 537 | } |
| 538 | |
| 539 | |
| 540 | template <bool seq_one_byte> |
| 541 | Handle<Object> JsonParser<seq_one_byte>::ParseJsonNumber() { |
| 542 | bool negative = false; |
| 543 | int beg_pos = position_; |
| 544 | if (c0_ == '-') { |
| 545 | Advance(); |
| 546 | negative = true; |
| 547 | } |
| 548 | if (c0_ == '0') { |
| 549 | Advance(); |
| 550 | // Prefix zero is only allowed if it's the only digit before |
| 551 | // a decimal point or exponent. |
| 552 | if (IsDecimalDigit(c0_)) return ReportUnexpectedCharacter(); |
| 553 | } else { |
| 554 | int i = 0; |
| 555 | int digits = 0; |
| 556 | if (c0_ < '1' || c0_ > '9') return ReportUnexpectedCharacter(); |
| 557 | do { |
| 558 | i = i * 10 + c0_ - '0'; |
| 559 | digits++; |
| 560 | Advance(); |
| 561 | } while (IsDecimalDigit(c0_)); |
| 562 | if (c0_ != '.' && c0_ != 'e' && c0_ != 'E' && digits < 10) { |
| 563 | SkipWhitespace(); |
| 564 | return Handle<Smi>(Smi::FromInt((negative ? -i : i)), isolate()); |
| 565 | } |
| 566 | } |
| 567 | if (c0_ == '.') { |
| 568 | Advance(); |
| 569 | if (!IsDecimalDigit(c0_)) return ReportUnexpectedCharacter(); |
| 570 | do { |
| 571 | Advance(); |
| 572 | } while (IsDecimalDigit(c0_)); |
| 573 | } |
| 574 | if (AsciiAlphaToLower(c0_) == 'e') { |
| 575 | Advance(); |
| 576 | if (c0_ == '-' || c0_ == '+') Advance(); |
| 577 | if (!IsDecimalDigit(c0_)) return ReportUnexpectedCharacter(); |
| 578 | do { |
| 579 | Advance(); |
| 580 | } while (IsDecimalDigit(c0_)); |
| 581 | } |
| 582 | int length = position_ - beg_pos; |
| 583 | double number; |
| 584 | if (seq_one_byte) { |
| 585 | Vector<const uint8_t> chars(seq_source_->GetChars() + beg_pos, length); |
| 586 | number = StringToDouble(isolate()->unicode_cache(), chars, |
| 587 | NO_FLAGS, // Hex, octal or trailing junk. |
| 588 | std::numeric_limits<double>::quiet_NaN()); |
| 589 | } else { |
| 590 | Vector<uint8_t> buffer = Vector<uint8_t>::New(length); |
| 591 | String::WriteToFlat(*source_, buffer.start(), beg_pos, position_); |
| 592 | Vector<const uint8_t> result = |
| 593 | Vector<const uint8_t>(buffer.start(), length); |
| 594 | number = StringToDouble(isolate()->unicode_cache(), |
| 595 | result, |
| 596 | NO_FLAGS, // Hex, octal or trailing junk. |
| 597 | 0.0); |
| 598 | buffer.Dispose(); |
| 599 | } |
| 600 | SkipWhitespace(); |
| 601 | return factory()->NewNumber(number, pretenure_); |
| 602 | } |
| 603 | |
| 604 | |
| 605 | template <typename StringType> |
| 606 | inline void SeqStringSet(Handle<StringType> seq_str, int i, uc32 c); |
| 607 | |
| 608 | template <> |
| 609 | inline void SeqStringSet(Handle<SeqTwoByteString> seq_str, int i, uc32 c) { |
| 610 | seq_str->SeqTwoByteStringSet(i, c); |
| 611 | } |
| 612 | |
| 613 | template <> |
| 614 | inline void SeqStringSet(Handle<SeqOneByteString> seq_str, int i, uc32 c) { |
| 615 | seq_str->SeqOneByteStringSet(i, c); |
| 616 | } |
| 617 | |
| 618 | template <typename StringType> |
| 619 | inline Handle<StringType> NewRawString(Factory* factory, |
| 620 | int length, |
| 621 | PretenureFlag pretenure); |
| 622 | |
| 623 | template <> |
| 624 | inline Handle<SeqTwoByteString> NewRawString(Factory* factory, |
| 625 | int length, |
| 626 | PretenureFlag pretenure) { |
| 627 | return factory->NewRawTwoByteString(length, pretenure).ToHandleChecked(); |
| 628 | } |
| 629 | |
| 630 | template <> |
| 631 | inline Handle<SeqOneByteString> NewRawString(Factory* factory, |
| 632 | int length, |
| 633 | PretenureFlag pretenure) { |
| 634 | return factory->NewRawOneByteString(length, pretenure).ToHandleChecked(); |
| 635 | } |
| 636 | |
| 637 | |
| 638 | // Scans the rest of a JSON string starting from position_ and writes |
| 639 | // prefix[start..end] along with the scanned characters into a |
| 640 | // sequential string of type StringType. |
| 641 | template <bool seq_one_byte> |
| 642 | template <typename StringType, typename SinkChar> |
| 643 | Handle<String> JsonParser<seq_one_byte>::SlowScanJsonString( |
| 644 | Handle<String> prefix, int start, int end) { |
| 645 | int count = end - start; |
| 646 | int max_length = count + source_length_ - position_; |
| 647 | int length = Min(max_length, Max(kInitialSpecialStringLength, 2 * count)); |
| 648 | Handle<StringType> seq_string = |
| 649 | NewRawString<StringType>(factory(), length, pretenure_); |
| 650 | // Copy prefix into seq_str. |
| 651 | SinkChar* dest = seq_string->GetChars(); |
| 652 | String::WriteToFlat(*prefix, dest, start, end); |
| 653 | |
| 654 | while (c0_ != '"') { |
| 655 | // Check for control character (0x00-0x1f) or unterminated string (<0). |
| 656 | if (c0_ < 0x20) return Handle<String>::null(); |
| 657 | if (count >= length) { |
| 658 | // We need to create a longer sequential string for the result. |
| 659 | return SlowScanJsonString<StringType, SinkChar>(seq_string, 0, count); |
| 660 | } |
| 661 | if (c0_ != '\\') { |
| 662 | // If the sink can contain UC16 characters, or source_ contains only |
| 663 | // Latin1 characters, there's no need to test whether we can store the |
| 664 | // character. Otherwise check whether the UC16 source character can fit |
| 665 | // in the Latin1 sink. |
| 666 | if (sizeof(SinkChar) == kUC16Size || seq_one_byte || |
| 667 | c0_ <= String::kMaxOneByteCharCode) { |
| 668 | SeqStringSet(seq_string, count++, c0_); |
| 669 | Advance(); |
| 670 | } else { |
| 671 | // StringType is SeqOneByteString and we just read a non-Latin1 char. |
| 672 | return SlowScanJsonString<SeqTwoByteString, uc16>(seq_string, 0, count); |
| 673 | } |
| 674 | } else { |
| 675 | Advance(); // Advance past the \. |
| 676 | switch (c0_) { |
| 677 | case '"': |
| 678 | case '\\': |
| 679 | case '/': |
| 680 | SeqStringSet(seq_string, count++, c0_); |
| 681 | break; |
| 682 | case 'b': |
| 683 | SeqStringSet(seq_string, count++, '\x08'); |
| 684 | break; |
| 685 | case 'f': |
| 686 | SeqStringSet(seq_string, count++, '\x0c'); |
| 687 | break; |
| 688 | case 'n': |
| 689 | SeqStringSet(seq_string, count++, '\x0a'); |
| 690 | break; |
| 691 | case 'r': |
| 692 | SeqStringSet(seq_string, count++, '\x0d'); |
| 693 | break; |
| 694 | case 't': |
| 695 | SeqStringSet(seq_string, count++, '\x09'); |
| 696 | break; |
| 697 | case 'u': { |
| 698 | uc32 value = 0; |
| 699 | for (int i = 0; i < 4; i++) { |
| 700 | Advance(); |
| 701 | int digit = HexValue(c0_); |
| 702 | if (digit < 0) { |
| 703 | return Handle<String>::null(); |
| 704 | } |
| 705 | value = value * 16 + digit; |
| 706 | } |
| 707 | if (sizeof(SinkChar) == kUC16Size || |
| 708 | value <= String::kMaxOneByteCharCode) { |
| 709 | SeqStringSet(seq_string, count++, value); |
| 710 | break; |
| 711 | } else { |
| 712 | // StringType is SeqOneByteString and we just read a non-Latin1 |
| 713 | // char. |
| 714 | position_ -= 6; // Rewind position_ to \ in \uxxxx. |
| 715 | Advance(); |
| 716 | return SlowScanJsonString<SeqTwoByteString, uc16>(seq_string, |
| 717 | 0, |
| 718 | count); |
| 719 | } |
| 720 | } |
| 721 | default: |
| 722 | return Handle<String>::null(); |
| 723 | } |
| 724 | Advance(); |
| 725 | } |
| 726 | } |
| 727 | |
| 728 | DCHECK_EQ('"', c0_); |
| 729 | // Advance past the last '"'. |
| 730 | AdvanceSkipWhitespace(); |
| 731 | |
| 732 | // Shrink seq_string length to count and return. |
| 733 | return SeqString::Truncate(seq_string, count); |
| 734 | } |
| 735 | |
| 736 | |
| 737 | template <bool seq_one_byte> |
| 738 | template <bool is_internalized> |
| 739 | Handle<String> JsonParser<seq_one_byte>::ScanJsonString() { |
| 740 | DCHECK_EQ('"', c0_); |
| 741 | Advance(); |
| 742 | if (c0_ == '"') { |
| 743 | AdvanceSkipWhitespace(); |
| 744 | return factory()->empty_string(); |
| 745 | } |
| 746 | |
| 747 | if (seq_one_byte && is_internalized) { |
| 748 | // Fast path for existing internalized strings. If the the string being |
| 749 | // parsed is not a known internalized string, contains backslashes or |
| 750 | // unexpectedly reaches the end of string, return with an empty handle. |
| 751 | uint32_t running_hash = isolate()->heap()->HashSeed(); |
| 752 | int position = position_; |
| 753 | uc32 c0 = c0_; |
| 754 | do { |
| 755 | if (c0 == '\\') { |
| 756 | c0_ = c0; |
| 757 | int beg_pos = position_; |
| 758 | position_ = position; |
| 759 | return SlowScanJsonString<SeqOneByteString, uint8_t>(source_, |
| 760 | beg_pos, |
| 761 | position_); |
| 762 | } |
| 763 | if (c0 < 0x20) return Handle<String>::null(); |
| 764 | running_hash = StringHasher::AddCharacterCore(running_hash, |
| 765 | static_cast<uint16_t>(c0)); |
| 766 | position++; |
| 767 | if (position >= source_length_) return Handle<String>::null(); |
| 768 | c0 = seq_source_->SeqOneByteStringGet(position); |
| 769 | } while (c0 != '"'); |
| 770 | int length = position - position_; |
| 771 | uint32_t hash = (length <= String::kMaxHashCalcLength) |
| 772 | ? StringHasher::GetHashCore(running_hash) |
| 773 | : static_cast<uint32_t>(length); |
| 774 | Vector<const uint8_t> string_vector( |
| 775 | seq_source_->GetChars() + position_, length); |
| 776 | StringTable* string_table = isolate()->heap()->string_table(); |
| 777 | uint32_t capacity = string_table->Capacity(); |
| 778 | uint32_t entry = StringTable::FirstProbe(hash, capacity); |
| 779 | uint32_t count = 1; |
| 780 | Handle<String> result; |
| 781 | while (true) { |
| 782 | Object* element = string_table->KeyAt(entry); |
| 783 | if (element == isolate()->heap()->undefined_value()) { |
| 784 | // Lookup failure. |
| 785 | result = factory()->InternalizeOneByteString( |
| 786 | seq_source_, position_, length); |
| 787 | break; |
| 788 | } |
| 789 | if (element != isolate()->heap()->the_hole_value() && |
| 790 | String::cast(element)->IsOneByteEqualTo(string_vector)) { |
| 791 | result = Handle<String>(String::cast(element), isolate()); |
| 792 | #ifdef DEBUG |
| 793 | uint32_t hash_field = |
| 794 | (hash << String::kHashShift) | String::kIsNotArrayIndexMask; |
| 795 | DCHECK_EQ(static_cast<int>(result->Hash()), |
| 796 | static_cast<int>(hash_field >> String::kHashShift)); |
| 797 | #endif |
| 798 | break; |
| 799 | } |
| 800 | entry = StringTable::NextProbe(entry, count++, capacity); |
| 801 | } |
| 802 | position_ = position; |
| 803 | // Advance past the last '"'. |
| 804 | AdvanceSkipWhitespace(); |
| 805 | return result; |
| 806 | } |
| 807 | |
| 808 | int beg_pos = position_; |
| 809 | // Fast case for Latin1 only without escape characters. |
| 810 | do { |
| 811 | // Check for control character (0x00-0x1f) or unterminated string (<0). |
| 812 | if (c0_ < 0x20) return Handle<String>::null(); |
| 813 | if (c0_ != '\\') { |
| 814 | if (seq_one_byte || c0_ <= String::kMaxOneByteCharCode) { |
| 815 | Advance(); |
| 816 | } else { |
| 817 | return SlowScanJsonString<SeqTwoByteString, uc16>(source_, |
| 818 | beg_pos, |
| 819 | position_); |
| 820 | } |
| 821 | } else { |
| 822 | return SlowScanJsonString<SeqOneByteString, uint8_t>(source_, |
| 823 | beg_pos, |
| 824 | position_); |
| 825 | } |
| 826 | } while (c0_ != '"'); |
| 827 | int length = position_ - beg_pos; |
| 828 | Handle<String> result = |
| 829 | factory()->NewRawOneByteString(length, pretenure_).ToHandleChecked(); |
| 830 | uint8_t* dest = SeqOneByteString::cast(*result)->GetChars(); |
| 831 | String::WriteToFlat(*source_, dest, beg_pos, position_); |
| 832 | |
| 833 | DCHECK_EQ('"', c0_); |
| 834 | // Advance past the last '"'. |
| 835 | AdvanceSkipWhitespace(); |
| 836 | return result; |
| 837 | } |
| 838 | |
| 839 | } // namespace internal |
| 840 | } // namespace v8 |
| 841 | |
| 842 | #endif // V8_PARSING_JSON_PARSER_H_ |