base/json/json_parser.cc

   1 // Copyright (c) 2012 The Chromium 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 #include "base/json/json_parser.h"
   6
   7 #include "base/float_util.h"
   8 #include "base/logging.h"
   9 #include "base/memory/scoped_ptr.h"
  10 #include "base/strings/string_number_conversions.h"
  11 #include "base/strings/string_piece.h"
  12 #include "base/strings/string_util.h"
  13 #include "base/strings/stringprintf.h"
  14 #include "base/strings/utf_string_conversion_utils.h"
  15 #include "base/strings/utf_string_conversions.h"
  16 #include "base/third_party/icu/icu_utf.h"
  17 #include "base/values.h"
  18
  19 namespace base {
  20 namespace internal {
  21
  22 namespace {
  23
  24 const int kStackMaxDepth = 100;
  25
  26 const int32 kExtendedASCIIStart = 0x80;
  27
  28 // This and the class below are used to own the JSON input string for when
  29 // string tokens are stored as StringPiece instead of std::string. This
  30 // optimization avoids about 2/3rds of string memory copies. The constructor
  31 // takes ownership of the input string. The real root value is Swap()ed into
  32 // the new instance.
  33 class DictionaryHiddenRootValue : public base::DictionaryValue {
  34  public:
  35   DictionaryHiddenRootValue(std::string* json, Value* root) : json_(json) {
  36     DCHECK(root->IsType(Value::TYPE_DICTIONARY));
  37     DictionaryValue::Swap(static_cast<DictionaryValue*>(root));
  38   }
  39
  40   void Swap(DictionaryValue* other) override {
  41     DVLOG(1) << "Swap()ing a DictionaryValue inefficiently.";
  42
  43     // First deep copy to convert JSONStringValue to std::string and swap that
  44     // copy with |other|, which contains the new contents of |this|.
  45     scoped_ptr<base::DictionaryValue> copy(DeepCopy());
  46     copy->Swap(other);
  47
  48     // Then erase the contents of the current dictionary and swap in the
  49     // new contents, originally from |other|.
  50     Clear();
  51     json_.reset();
  52     DictionaryValue::Swap(copy.get());
  53   }
  54
  55   // Not overriding DictionaryValue::Remove because it just calls through to
  56   // the method below.
  57
  58   bool RemoveWithoutPathExpansion(const std::string& key,
  59                                   scoped_ptr<Value>* out) override {
  60     // If the caller won't take ownership of the removed value, just call up.
  61     if (!out)
  62       return DictionaryValue::RemoveWithoutPathExpansion(key, out);
  63
  64     DVLOG(1) << "Remove()ing from a DictionaryValue inefficiently.";
  65
  66     // Otherwise, remove the value while its still "owned" by this and copy it
  67     // to convert any JSONStringValues to std::string.
  68     scoped_ptr<Value> out_owned;
  69     if (!DictionaryValue::RemoveWithoutPathExpansion(key, &out_owned))
  70       return false;
  71
  72     out->reset(out_owned->DeepCopy());
  73
  74     return true;
  75   }
  76
  77  private:
  78   scoped_ptr<std::string> json_;
  79
  80   DISALLOW_COPY_AND_ASSIGN(DictionaryHiddenRootValue);
  81 };
  82
  83 class ListHiddenRootValue : public base::ListValue {
  84  public:
  85   ListHiddenRootValue(std::string* json, Value* root) : json_(json) {
  86     DCHECK(root->IsType(Value::TYPE_LIST));
  87     ListValue::Swap(static_cast<ListValue*>(root));
  88   }
  89
  90   void Swap(ListValue* other) override {
  91     DVLOG(1) << "Swap()ing a ListValue inefficiently.";
  92
  93     // First deep copy to convert JSONStringValue to std::string and swap that
  94     // copy with |other|, which contains the new contents of |this|.
  95     scoped_ptr<base::ListValue> copy(DeepCopy());
  96     copy->Swap(other);
  97
  98     // Then erase the contents of the current list and swap in the new contents,
  99     // originally from |other|.
 100     Clear();
 101     json_.reset();
 102     ListValue::Swap(copy.get());
 103   }
 104
 105   bool Remove(size_t index, scoped_ptr<Value>* out) override {
 106     // If the caller won't take ownership of the removed value, just call up.
 107     if (!out)
 108       return ListValue::Remove(index, out);
 109
 110     DVLOG(1) << "Remove()ing from a ListValue inefficiently.";
 111
 112     // Otherwise, remove the value while its still "owned" by this and copy it
 113     // to convert any JSONStringValues to std::string.
 114     scoped_ptr<Value> out_owned;
 115     if (!ListValue::Remove(index, &out_owned))
 116       return false;
 117
 118     out->reset(out_owned->DeepCopy());
 119
 120     return true;
 121   }
 122
 123  private:
 124   scoped_ptr<std::string> json_;
 125
 126   DISALLOW_COPY_AND_ASSIGN(ListHiddenRootValue);
 127 };
 128
 129 // A variant on StringValue that uses StringPiece instead of copying the string
 130 // into the Value. This can only be stored in a child of hidden root (above),
 131 // otherwise the referenced string will not be guaranteed to outlive it.
 132 class JSONStringValue : public base::Value {
 133  public:
 134   explicit JSONStringValue(const base::StringPiece& piece)
 135       : Value(TYPE_STRING),
 136         string_piece_(piece) {
 137   }
 138
 139   // Overridden from base::Value:
 140   bool GetAsString(std::string* out_value) const override {
 141     string_piece_.CopyToString(out_value);
 142     return true;
 143   }
 144   bool GetAsString(string16* out_value) const override {
 145     *out_value = UTF8ToUTF16(string_piece_);
 146     return true;
 147   }
 148   Value* DeepCopy() const override {
 149     return new StringValue(string_piece_.as_string());
 150   }
 151   bool Equals(const Value* other) const override {
 152     std::string other_string;
 153     return other->IsType(TYPE_STRING) && other->GetAsString(&other_string) &&
 154         StringPiece(other_string) == string_piece_;
 155   }
 156
 157  private:
 158   // The location in the original input stream.
 159   base::StringPiece string_piece_;
 160
 161   DISALLOW_COPY_AND_ASSIGN(JSONStringValue);
 162 };
 163
 164 // Simple class that checks for maximum recursion/"stack overflow."
 165 class StackMarker {
 166  public:
 167   explicit StackMarker(int* depth) : depth_(depth) {
 168     ++(*depth_);
 169     DCHECK_LE(*depth_, kStackMaxDepth);
 170   }
 171   ~StackMarker() {
 172     --(*depth_);
 173   }
 174
 175   bool IsTooDeep() const {
 176     return *depth_ >= kStackMaxDepth;
 177   }
 178
 179  private:
 180   int* const depth_;
 181
 182   DISALLOW_COPY_AND_ASSIGN(StackMarker);
 183 };
 184
 185 }  // namespace
 186
 187 JSONParser::JSONParser(int options)
 188     : options_(options),
 189       start_pos_(NULL),
 190       pos_(NULL),
 191       end_pos_(NULL),
 192       index_(0),
 193       stack_depth_(0),
 194       line_number_(0),
 195       index_last_line_(0),
 196       error_code_(JSONReader::JSON_NO_ERROR),
 197       error_line_(0),
 198       error_column_(0) {
 199 }
 200
 201 JSONParser::~JSONParser() {
 202 }
 203
 204 Value* JSONParser::Parse(const StringPiece& input) {
 205   scoped_ptr<std::string> input_copy;
 206   // If the children of a JSON root can be detached, then hidden roots cannot
 207   // be used, so do not bother copying the input because StringPiece will not
 208   // be used anywhere.
 209   if (!(options_ & JSON_DETACHABLE_CHILDREN)) {
 210     input_copy.reset(new std::string(input.as_string()));
 211     start_pos_ = input_copy->data();
 212   } else {
 213     start_pos_ = input.data();
 214   }
 215   pos_ = start_pos_;
 216   end_pos_ = start_pos_ + input.length();
 217   index_ = 0;
 218   line_number_ = 1;
 219   index_last_line_ = 0;
 220
 221   error_code_ = JSONReader::JSON_NO_ERROR;
 222   error_line_ = 0;
 223   error_column_ = 0;
 224
 225   // When the input JSON string starts with a UTF-8 Byte-Order-Mark
 226   // <0xEF 0xBB 0xBF>, advance the start position to avoid the
 227   // ParseNextToken function mis-treating a Unicode BOM as an invalid
 228   // character and returning NULL.
 229   if (CanConsume(3) && static_cast<uint8>(*pos_) == 0xEF &&
 230       static_cast<uint8>(*(pos_ + 1)) == 0xBB &&
 231       static_cast<uint8>(*(pos_ + 2)) == 0xBF) {
 232     NextNChars(3);
 233   }
 234
 235   // Parse the first and any nested tokens.
 236   scoped_ptr<Value> root(ParseNextToken());
 237   if (!root.get())
 238     return NULL;
 239
 240   // Make sure the input stream is at an end.
 241   if (GetNextToken() != T_END_OF_INPUT) {
 242     if (!CanConsume(1) || (NextChar() && GetNextToken() != T_END_OF_INPUT)) {
 243       ReportError(JSONReader::JSON_UNEXPECTED_DATA_AFTER_ROOT, 1);
 244       return NULL;
 245     }
 246   }
 247
 248   // Dictionaries and lists can contain JSONStringValues, so wrap them in a
 249   // hidden root.
 250   if (!(options_ & JSON_DETACHABLE_CHILDREN)) {
 251     if (root->IsType(Value::TYPE_DICTIONARY)) {
 252       return new DictionaryHiddenRootValue(input_copy.release(), root.get());
 253     } else if (root->IsType(Value::TYPE_LIST)) {
 254       return new ListHiddenRootValue(input_copy.release(), root.get());
 255     } else if (root->IsType(Value::TYPE_STRING)) {
 256       // A string type could be a JSONStringValue, but because there's no
 257       // corresponding HiddenRootValue, the memory will be lost. Deep copy to
 258       // preserve it.
 259       return root->DeepCopy();
 260     }
 261   }
 262
 263   // All other values can be returned directly.
 264   return root.release();
 265 }
 266
 267 JSONReader::JsonParseError JSONParser::error_code() const {
 268   return error_code_;
 269 }
 270
 271 std::string JSONParser::GetErrorMessage() const {
 272   return FormatErrorMessage(error_line_, error_column_,
 273       JSONReader::ErrorCodeToString(error_code_));
 274 }
 275
 276 // StringBuilder ///////////////////////////////////////////////////////////////
 277
 278 JSONParser::StringBuilder::StringBuilder()
 279     : pos_(NULL),
 280       length_(0),
 281       string_(NULL) {
 282 }
 283
 284 JSONParser::StringBuilder::StringBuilder(const char* pos)
 285     : pos_(pos),
 286       length_(0),
 287       string_(NULL) {
 288 }
 289
 290 void JSONParser::StringBuilder::Swap(StringBuilder* other) {
 291   std::swap(other->string_, string_);
 292   std::swap(other->pos_, pos_);
 293   std::swap(other->length_, length_);
 294 }
 295
 296 JSONParser::StringBuilder::~StringBuilder() {
 297   delete string_;
 298 }
 299
 300 void JSONParser::StringBuilder::Append(const char& c) {
 301   DCHECK_GE(c, 0);
 302   DCHECK_LT(c, 128);
 303
 304   if (string_)
 305     string_->push_back(c);
 306   else
 307     ++length_;
 308 }
 309
 310 void JSONParser::StringBuilder::AppendString(const std::string& str) {
 311   DCHECK(string_);
 312   string_->append(str);
 313 }
 314
 315 void JSONParser::StringBuilder::Convert() {
 316   if (string_)
 317     return;
 318   string_  = new std::string(pos_, length_);
 319 }
 320
 321 bool JSONParser::StringBuilder::CanBeStringPiece() const {
 322   return !string_;
 323 }
 324
 325 StringPiece JSONParser::StringBuilder::AsStringPiece() {
 326   if (string_)
 327     return StringPiece();
 328   return StringPiece(pos_, length_);
 329 }
 330
 331 const std::string& JSONParser::StringBuilder::AsString() {
 332   if (!string_)
 333     Convert();
 334   return *string_;
 335 }
 336
 337 // JSONParser private //////////////////////////////////////////////////////////
 338
 339 inline bool JSONParser::CanConsume(int length) {
 340   return pos_ + length <= end_pos_;
 341 }
 342
 343 const char* JSONParser::NextChar() {
 344   DCHECK(CanConsume(1));
 345   ++index_;
 346   ++pos_;
 347   return pos_;
 348 }
 349
 350 void JSONParser::NextNChars(int n) {
 351   DCHECK(CanConsume(n));
 352   index_ += n;
 353   pos_ += n;
 354 }
 355
 356 JSONParser::Token JSONParser::GetNextToken() {
 357   EatWhitespaceAndComments();
 358   if (!CanConsume(1))
 359     return T_END_OF_INPUT;
 360
 361   switch (*pos_) {
 362     case '{':
 363       return T_OBJECT_BEGIN;
 364     case '}':
 365       return T_OBJECT_END;
 366     case '[':
 367       return T_ARRAY_BEGIN;
 368     case ']':
 369       return T_ARRAY_END;
 370     case '"':
 371       return T_STRING;
 372     case '0':
 373     case '1':
 374     case '2':
 375     case '3':
 376     case '4':
 377     case '5':
 378     case '6':
 379     case '7':
 380     case '8':
 381     case '9':
 382     case '-':
 383       return T_NUMBER;
 384     case 't':
 385       return T_BOOL_TRUE;
 386     case 'f':
 387       return T_BOOL_FALSE;
 388     case 'n':
 389       return T_NULL;
 390     case ',':
 391       return T_LIST_SEPARATOR;
 392     case ':':
 393       return T_OBJECT_PAIR_SEPARATOR;
 394     default:
 395       return T_INVALID_TOKEN;
 396   }
 397 }
 398
 399 void JSONParser::EatWhitespaceAndComments() {
 400   while (pos_ < end_pos_) {
 401     switch (*pos_) {
 402       case '\r':
 403       case '\n':
 404         index_last_line_ = index_;
 405         // Don't increment line_number_ twice for "\r\n".
 406         if (!(*pos_ == '\n' && pos_ > start_pos_ && *(pos_ - 1) == '\r'))
 407           ++line_number_;
 408         // Fall through.
 409       case ' ':
 410       case '\t':
 411         NextChar();
 412         break;
 413       case '/':
 414         if (!EatComment())
 415           return;
 416         break;
 417       default:
 418         return;
 419     }
 420   }
 421 }
 422
 423 bool JSONParser::EatComment() {
 424   if (*pos_ != '/' || !CanConsume(1))
 425     return false;
 426
 427   char next_char = *NextChar();
 428   if (next_char == '/') {
 429     // Single line comment, read to newline.
 430     while (CanConsume(1)) {
 431       next_char = *NextChar();
 432       if (next_char == '\n' || next_char == '\r')
 433         return true;
 434     }
 435   } else if (next_char == '*') {
 436     char previous_char = '\0';
 437     // Block comment, read until end marker.
 438     while (CanConsume(1)) {
 439       next_char = *NextChar();
 440       if (previous_char == '*' && next_char == '/') {
 441         // EatWhitespaceAndComments will inspect pos_, which will still be on
 442         // the last / of the comment, so advance once more (which may also be
 443         // end of input).
 444         NextChar();
 445         return true;
 446       }
 447       previous_char = next_char;
 448     }
 449
 450     // If the comment is unterminated, GetNextToken will report T_END_OF_INPUT.
 451   }
 452
 453   return false;
 454 }
 455
 456 Value* JSONParser::ParseNextToken() {
 457   return ParseToken(GetNextToken());
 458 }
 459
 460 Value* JSONParser::ParseToken(Token token) {
 461   switch (token) {
 462     case T_OBJECT_BEGIN:
 463       return ConsumeDictionary();
 464     case T_ARRAY_BEGIN:
 465       return ConsumeList();
 466     case T_STRING:
 467       return ConsumeString();
 468     case T_NUMBER:
 469       return ConsumeNumber();
 470     case T_BOOL_TRUE:
 471     case T_BOOL_FALSE:
 472     case T_NULL:
 473       return ConsumeLiteral();
 474     default:
 475       ReportError(JSONReader::JSON_UNEXPECTED_TOKEN, 1);
 476       return NULL;
 477   }
 478 }
 479
 480 Value* JSONParser::ConsumeDictionary() {
 481   if (*pos_ != '{') {
 482     ReportError(JSONReader::JSON_UNEXPECTED_TOKEN, 1);
 483     return NULL;
 484   }
 485
 486   StackMarker depth_check(&stack_depth_);
 487   if (depth_check.IsTooDeep()) {
 488     ReportError(JSONReader::JSON_TOO_MUCH_NESTING, 1);
 489     return NULL;
 490   }
 491
 492   scoped_ptr<DictionaryValue> dict(new DictionaryValue);
 493
 494   NextChar();
 495   Token token = GetNextToken();
 496   while (token != T_OBJECT_END) {
 497     if (token != T_STRING) {
 498       ReportError(JSONReader::JSON_UNQUOTED_DICTIONARY_KEY, 1);
 499       return NULL;
 500     }
 501
 502     // First consume the key.
 503     StringBuilder key;
 504     if (!ConsumeStringRaw(&key)) {
 505       return NULL;
 506     }
 507
 508     // Read the separator.
 509     NextChar();
 510     token = GetNextToken();
 511     if (token != T_OBJECT_PAIR_SEPARATOR) {
 512       ReportError(JSONReader::JSON_SYNTAX_ERROR, 1);
 513       return NULL;
 514     }
 515
 516     // The next token is the value. Ownership transfers to |dict|.
 517     NextChar();
 518     Value* value = ParseNextToken();
 519     if (!value) {
 520       // ReportError from deeper level.
 521       return NULL;
 522     }
 523
 524     dict->SetWithoutPathExpansion(key.AsString(), value);
 525
 526     NextChar();
 527     token = GetNextToken();
 528     if (token == T_LIST_SEPARATOR) {
 529       NextChar();
 530       token = GetNextToken();
 531       if (token == T_OBJECT_END && !(options_ & JSON_ALLOW_TRAILING_COMMAS)) {
 532         ReportError(JSONReader::JSON_TRAILING_COMMA, 1);
 533         return NULL;
 534       }
 535     } else if (token != T_OBJECT_END) {
 536       ReportError(JSONReader::JSON_SYNTAX_ERROR, 0);
 537       return NULL;
 538     }
 539   }
 540
 541   return dict.release();
 542 }
 543
 544 Value* JSONParser::ConsumeList() {
 545   if (*pos_ != '[') {
 546     ReportError(JSONReader::JSON_UNEXPECTED_TOKEN, 1);
 547     return NULL;
 548   }
 549
 550   StackMarker depth_check(&stack_depth_);
 551   if (depth_check.IsTooDeep()) {
 552     ReportError(JSONReader::JSON_TOO_MUCH_NESTING, 1);
 553     return NULL;
 554   }
 555
 556   scoped_ptr<ListValue> list(new ListValue);
 557
 558   NextChar();
 559   Token token = GetNextToken();
 560   while (token != T_ARRAY_END) {
 561     Value* item = ParseToken(token);
 562     if (!item) {
 563       // ReportError from deeper level.
 564       return NULL;
 565     }
 566
 567     list->Append(item);
 568
 569     NextChar();
 570     token = GetNextToken();
 571     if (token == T_LIST_SEPARATOR) {
 572       NextChar();
 573       token = GetNextToken();
 574       if (token == T_ARRAY_END && !(options_ & JSON_ALLOW_TRAILING_COMMAS)) {
 575         ReportError(JSONReader::JSON_TRAILING_COMMA, 1);
 576         return NULL;
 577       }
 578     } else if (token != T_ARRAY_END) {
 579       ReportError(JSONReader::JSON_SYNTAX_ERROR, 1);
 580       return NULL;
 581     }
 582   }
 583
 584   return list.release();
 585 }
 586
 587 Value* JSONParser::ConsumeString() {
 588   StringBuilder string;
 589   if (!ConsumeStringRaw(&string))
 590     return NULL;
 591
 592   // Create the Value representation, using a hidden root, if configured
 593   // to do so, and if the string can be represented by StringPiece.
 594   if (string.CanBeStringPiece() && !(options_ & JSON_DETACHABLE_CHILDREN)) {
 595     return new JSONStringValue(string.AsStringPiece());
 596   } else {
 597     if (string.CanBeStringPiece())
 598       string.Convert();
 599     return new StringValue(string.AsString());
 600   }
 601 }
 602
 603 bool JSONParser::ConsumeStringRaw(StringBuilder* out) {
 604   if (*pos_ != '"') {
 605     ReportError(JSONReader::JSON_UNEXPECTED_TOKEN, 1);
 606     return false;
 607   }
 608
 609   // StringBuilder will internally build a StringPiece unless a UTF-16
 610   // conversion occurs, at which point it will perform a copy into a
 611   // std::string.
 612   StringBuilder string(NextChar());
 613
 614   int length = end_pos_ - start_pos_;
 615   int32 next_char = 0;
 616
 617   while (CanConsume(1)) {
 618     pos_ = start_pos_ + index_;  // CBU8_NEXT is postcrement.
 619     CBU8_NEXT(start_pos_, index_, length, next_char);
 620     if (next_char < 0 || !IsValidCharacter(next_char)) {
 621       ReportError(JSONReader::JSON_UNSUPPORTED_ENCODING, 1);
 622       return false;
 623     }
 624
 625     // If this character is an escape sequence...
 626     if (next_char == '\\') {
 627       // The input string will be adjusted (either by combining the two
 628       // characters of an encoded escape sequence, or with a UTF conversion),
 629       // so using StringPiece isn't possible -- force a conversion.
 630       string.Convert();
 631
 632       if (!CanConsume(1)) {
 633         ReportError(JSONReader::JSON_INVALID_ESCAPE, 0);
 634         return false;
 635       }
 636
 637       switch (*NextChar()) {
 638         // Allowed esape sequences:
 639         case 'x': {  // UTF-8 sequence.
 640           // UTF-8 \x escape sequences are not allowed in the spec, but they
 641           // are supported here for backwards-compatiblity with the old parser.
 642           if (!CanConsume(2)) {
 643             ReportError(JSONReader::JSON_INVALID_ESCAPE, 1);
 644             return false;
 645           }
 646
 647           int hex_digit = 0;
 648           if (!HexStringToInt(StringPiece(NextChar(), 2), &hex_digit)) {
 649             ReportError(JSONReader::JSON_INVALID_ESCAPE, -1);
 650             return false;
 651           }
 652           NextChar();
 653
 654           if (hex_digit < kExtendedASCIIStart)
 655             string.Append(static_cast<char>(hex_digit));
 656           else
 657             DecodeUTF8(hex_digit, &string);
 658           break;
 659         }
 660         case 'u': {  // UTF-16 sequence.
 661           // UTF units are of the form \uXXXX.
 662           if (!CanConsume(5)) {  // 5 being 'u' and four HEX digits.
 663             ReportError(JSONReader::JSON_INVALID_ESCAPE, 0);
 664             return false;
 665           }
 666
 667           // Skip the 'u'.
 668           NextChar();
 669
 670           std::string utf8_units;
 671           if (!DecodeUTF16(&utf8_units)) {
 672             ReportError(JSONReader::JSON_INVALID_ESCAPE, -1);
 673             return false;
 674           }
 675
 676           string.AppendString(utf8_units);
 677           break;
 678         }
 679         case '"':
 680           string.Append('"');
 681           break;
 682         case '\\':
 683           string.Append('\\');
 684           break;
 685         case '/':
 686           string.Append('/');
 687           break;
 688         case 'b':
 689           string.Append('\b');
 690           break;
 691         case 'f':
 692           string.Append('\f');
 693           break;
 694         case 'n':
 695           string.Append('\n');
 696           break;
 697         case 'r':
 698           string.Append('\r');
 699           break;
 700         case 't':
 701           string.Append('\t');
 702           break;
 703         case 'v':  // Not listed as valid escape sequence in the RFC.
 704           string.Append('\v');
 705           break;
 706         // All other escape squences are illegal.
 707         default:
 708           ReportError(JSONReader::JSON_INVALID_ESCAPE, 0);
 709           return false;
 710       }
 711     } else if (next_char == '"') {
 712       --index_;  // Rewind by one because of CBU8_NEXT.
 713       out->Swap(&string);
 714       return true;
 715     } else {
 716       if (next_char < kExtendedASCIIStart)
 717         string.Append(static_cast<char>(next_char));
 718       else
 719         DecodeUTF8(next_char, &string);
 720     }
 721   }
 722
 723   ReportError(JSONReader::JSON_SYNTAX_ERROR, 0);
 724   return false;
 725 }
 726
 727 // Entry is at the first X in \uXXXX.
 728 bool JSONParser::DecodeUTF16(std::string* dest_string) {
 729   if (!CanConsume(4))
 730     return false;
 731
 732   // This is a 32-bit field because the shift operations in the
 733   // conversion process below cause MSVC to error about "data loss."
 734   // This only stores UTF-16 code units, though.
 735   // Consume the UTF-16 code unit, which may be a high surrogate.
 736   int code_unit16_high = 0;
 737   if (!HexStringToInt(StringPiece(pos_, 4), &code_unit16_high))
 738     return false;
 739
 740   // Only add 3, not 4, because at the end of this iteration, the parser has
 741   // finished working with the last digit of the UTF sequence, meaning that
 742   // the next iteration will advance to the next byte.
 743   NextNChars(3);
 744
 745   // Used to convert the UTF-16 code units to a code point and then to a UTF-8
 746   // code unit sequence.
 747   char code_unit8[8] = { 0 };
 748   size_t offset = 0;
 749
 750   // If this is a high surrogate, consume the next code unit to get the
 751   // low surrogate.
 752   if (CBU16_IS_SURROGATE(code_unit16_high)) {
 753     // Make sure this is the high surrogate. If not, it's an encoding
 754     // error.
 755     if (!CBU16_IS_SURROGATE_LEAD(code_unit16_high))
 756       return false;
 757
 758     // Make sure that the token has more characters to consume the
 759     // lower surrogate.
 760     if (!CanConsume(6))  // 6 being '\' 'u' and four HEX digits.
 761       return false;
 762     if (*NextChar() != '\\' || *NextChar() != 'u')
 763       return false;
 764
 765     NextChar();  // Read past 'u'.
 766     int code_unit16_low = 0;
 767     if (!HexStringToInt(StringPiece(pos_, 4), &code_unit16_low))
 768       return false;
 769
 770     NextNChars(3);
 771
 772     if (!CBU16_IS_TRAIL(code_unit16_low)) {
 773       return false;
 774     }
 775
 776     uint32 code_point = CBU16_GET_SUPPLEMENTARY(code_unit16_high,
 777                                                 code_unit16_low);
 778     offset = 0;
 779     CBU8_APPEND_UNSAFE(code_unit8, offset, code_point);
 780   } else {
 781     // Not a surrogate.
 782     DCHECK(CBU16_IS_SINGLE(code_unit16_high));
 783     CBU8_APPEND_UNSAFE(code_unit8, offset, code_unit16_high);
 784   }
 785
 786   dest_string->append(code_unit8);
 787   return true;
 788 }
 789
 790 void JSONParser::DecodeUTF8(const int32& point, StringBuilder* dest) {
 791   // Anything outside of the basic ASCII plane will need to be decoded from
 792   // int32 to a multi-byte sequence.
 793   if (point < kExtendedASCIIStart) {
 794     dest->Append(static_cast<char>(point));
 795   } else {
 796     char utf8_units[4] = { 0 };
 797     int offset = 0;
 798     CBU8_APPEND_UNSAFE(utf8_units, offset, point);
 799     dest->Convert();
 800     // CBU8_APPEND_UNSAFE can overwrite up to 4 bytes, so utf8_units may not be
 801     // zero terminated at this point.  |offset| contains the correct length.
 802     dest->AppendString(std::string(utf8_units, offset));
 803   }
 804 }
 805
 806 Value* JSONParser::ConsumeNumber() {
 807   const char* num_start = pos_;
 808   const int start_index = index_;
 809   int end_index = start_index;
 810
 811   if (*pos_ == '-')
 812     NextChar();
 813
 814   if (!ReadInt(false)) {
 815     ReportError(JSONReader::JSON_SYNTAX_ERROR, 1);
 816     return NULL;
 817   }
 818   end_index = index_;
 819
 820   // The optional fraction part.
 821   if (*pos_ == '.') {
 822     if (!CanConsume(1)) {
 823       ReportError(JSONReader::JSON_SYNTAX_ERROR, 1);
 824       return NULL;
 825     }
 826     NextChar();
 827     if (!ReadInt(true)) {
 828       ReportError(JSONReader::JSON_SYNTAX_ERROR, 1);
 829       return NULL;
 830     }
 831     end_index = index_;
 832   }
 833
 834   // Optional exponent part.
 835   if (*pos_ == 'e' || *pos_ == 'E') {
 836     NextChar();
 837     if (*pos_ == '-' || *pos_ == '+')
 838       NextChar();
 839     if (!ReadInt(true)) {
 840       ReportError(JSONReader::JSON_SYNTAX_ERROR, 1);
 841       return NULL;
 842     }
 843     end_index = index_;
 844   }
 845
 846   // ReadInt is greedy because numbers have no easily detectable sentinel,
 847   // so save off where the parser should be on exit (see Consume invariant at
 848   // the top of the header), then make sure the next token is one which is
 849   // valid.
 850   const char* exit_pos = pos_ - 1;
 851   int exit_index = index_ - 1;
 852
 853   switch (GetNextToken()) {
 854     case T_OBJECT_END:
 855     case T_ARRAY_END:
 856     case T_LIST_SEPARATOR:
 857     case T_END_OF_INPUT:
 858       break;
 859     default:
 860       ReportError(JSONReader::JSON_SYNTAX_ERROR, 1);
 861       return NULL;
 862   }
 863
 864   pos_ = exit_pos;
 865   index_ = exit_index;
 866
 867   StringPiece num_string(num_start, end_index - start_index);
 868
 869   int num_int;
 870   if (StringToInt(num_string, &num_int))
 871     return new FundamentalValue(num_int);
 872
 873   double num_double;
 874   if (base::StringToDouble(num_string.as_string(), &num_double) &&
 875       IsFinite(num_double)) {
 876     return new FundamentalValue(num_double);
 877   }
 878
 879   return NULL;
 880 }
 881
 882 bool JSONParser::ReadInt(bool allow_leading_zeros) {
 883   char first = *pos_;
 884   int len = 0;
 885
 886   char c = first;
 887   while (CanConsume(1) && IsAsciiDigit(c)) {
 888     c = *NextChar();
 889     ++len;
 890   }
 891
 892   if (len == 0)
 893     return false;
 894
 895   if (!allow_leading_zeros && len > 1 && first == '0')
 896     return false;
 897
 898   return true;
 899 }
 900
 901 Value* JSONParser::ConsumeLiteral() {
 902   switch (*pos_) {
 903     case 't': {
 904       const char kTrueLiteral[] = "true";
 905       const int kTrueLen = static_cast<int>(strlen(kTrueLiteral));
 906       if (!CanConsume(kTrueLen - 1) ||
 907           !StringsAreEqual(pos_, kTrueLiteral, kTrueLen)) {
 908         ReportError(JSONReader::JSON_SYNTAX_ERROR, 1);
 909         return NULL;
 910       }
 911       NextNChars(kTrueLen - 1);
 912       return new FundamentalValue(true);
 913     }
 914     case 'f': {
 915       const char kFalseLiteral[] = "false";
 916       const int kFalseLen = static_cast<int>(strlen(kFalseLiteral));
 917       if (!CanConsume(kFalseLen - 1) ||
 918           !StringsAreEqual(pos_, kFalseLiteral, kFalseLen)) {
 919         ReportError(JSONReader::JSON_SYNTAX_ERROR, 1);
 920         return NULL;
 921       }
 922       NextNChars(kFalseLen - 1);
 923       return new FundamentalValue(false);
 924     }
 925     case 'n': {
 926       const char kNullLiteral[] = "null";
 927       const int kNullLen = static_cast<int>(strlen(kNullLiteral));
 928       if (!CanConsume(kNullLen - 1) ||
 929           !StringsAreEqual(pos_, kNullLiteral, kNullLen)) {
 930         ReportError(JSONReader::JSON_SYNTAX_ERROR, 1);
 931         return NULL;
 932       }
 933       NextNChars(kNullLen - 1);
 934       return Value::CreateNullValue();
 935     }
 936     default:
 937       ReportError(JSONReader::JSON_UNEXPECTED_TOKEN, 1);
 938       return NULL;
 939   }
 940 }
 941
 942 // static
 943 bool JSONParser::StringsAreEqual(const char* one, const char* two, size_t len) {
 944   return strncmp(one, two, len) == 0;
 945 }
 946
 947 void JSONParser::ReportError(JSONReader::JsonParseError code,
 948                              int column_adjust) {
 949   error_code_ = code;
 950   error_line_ = line_number_;
 951   error_column_ = index_ - index_last_line_ + column_adjust;
 952 }
 953
 954 // static
 955 std::string JSONParser::FormatErrorMessage(int line, int column,
 956                                            const std::string& description) {
 957   if (line || column) {
 958     return StringPrintf("Line: %i, column: %i, %s",
 959         line, column, description.c_str());
 960   }
 961   return description;
 962 }
 963
 964 }  // namespace internal
 965 }  // namespace base