Merge up to 8597 to experimental/gc from the bleeding edge.

src/json-parser.h

Index: src/json-parser.h
===================================================================
--- src/json-parser.h	(revision 8618)
+++ src/json-parser.h	(working copy)
@@ -28,12 +28,19 @@
 #ifndef V8_JSON_PARSER_H_
 #define V8_JSON_PARSER_H_
 
+#include "v8.h"
+
+#include "char-predicates-inl.h"
+#include "v8conversions.h"
+#include "messages.h"
+#include "spaces-inl.h"
 #include "token.h"
 
 namespace v8 {
 namespace internal {
 
 // A simple json parser.
+template <bool seq_ascii>
 class JsonParser BASE_EMBEDDED {
  public:
   static Handle<Object> Parse(Handle<String> source) {
@@ -47,41 +54,64 @@
   Handle<Object> ParseJson(Handle<String> source);
 
   inline void Advance() {
+    position_++;
     if (position_ >= source_length_) {
-      position_++;
       c0_ = kEndOfString;
-    } else if (is_sequential_ascii_) {
-      position_++;
+    } else if (seq_ascii) {
       c0_ = seq_source_->SeqAsciiStringGet(position_);
     } else {
-      position_++;
       c0_ = source_->Get(position_);
     }
   }
 
-  inline Isolate* isolate() { return isolate_; }
+  // The JSON lexical grammar is specified in the ECMAScript 5 standard,
+  // section 15.12.1.1. The only allowed whitespace characters between tokens
+  // are tab, carriage-return, newline and space.
 
-  // Get the string for the current string token.
-  Handle<String> GetString(bool hint_symbol);
-  Handle<String> GetString();
-  Handle<String> GetSymbol();
+  inline void AdvanceSkipWhitespace() {
+    do {
+      Advance();
+    } while (c0_ == '\t' || c0_ == '\r' || c0_ == '\n' || c0_ == ' ');
+  }
 
-  // Scan a single JSON token. The JSON lexical grammar is specified in the
-  // ECMAScript 5 standard, section 15.12.1.1.
-  // Recognizes all of the single-character tokens directly, or calls a function
-  // to scan a number, string or identifier literal.
-  // The only allowed whitespace characters between tokens are tab,
-  // carriage-return, newline and space.
-  void ScanJson();
+  inline void SkipWhitespace() {
+    while (c0_ == '\t' || c0_ == '\r' || c0_ == '\n' || c0_ == ' ') {
+      Advance();
+    }
+  }
 
+  inline uc32 AdvanceGetChar() {
+    Advance();
+    return c0_;
+  }
+
+  // Checks that current charater is c.
+  // If so, then consume c and skip whitespace.
+  inline bool MatchSkipWhiteSpace(uc32 c) {
+    if (c0_ == c) {
+      AdvanceSkipWhitespace();
+      return true;
+    }
+    return false;
+  }
+
   // A JSON string (production JSONString) is subset of valid JavaScript string
   // literals. The string must only be double-quoted (not single-quoted), and
   // the only allowed backslash-escapes are ", /, \, b, f, n, r, t and
   // four-digit hex escapes (uXXXX). Any other use of backslashes is invalid.
-  Token::Value ScanJsonString();
-  // Slow version for unicode support, uses the first ascii_count characters,
-  // as first part of a ConsString
-  Token::Value SlowScanJsonString();
+  Handle<String> ParseJsonString() {
+    return ScanJsonString<false>();
+  }
+  Handle<String> ParseJsonSymbol() {
+    return ScanJsonString<true>();
+  }
+  template <bool is_symbol>
+  Handle<String> ScanJsonString();
+  // Creates a new string and copies prefix[start..end] into the beginning
+  // of it. Then scans the rest of the string, adding characters after the
+  // prefix. Called by ScanJsonString when reaching a '\' or non-ASCII char.
+  template <typename StringType, typename SinkChar>
+  Handle<String> SlowScanJsonString(Handle<String> prefix, int start, int end);
 
   // A JSON number (production JSONNumber) is a subset of the valid JavaScript
   // decimal number literals.
@@ -89,13 +119,8 @@
   // digit before and after a decimal point, may not have prefixed zeros (unless
   // the integer part is zero), and may include an exponent part (e.g., "e-10").
   // Hexadecimal and octal numbers are not allowed.
-  Token::Value ScanJsonNumber();
+  Handle<Object> ParseJsonNumber();
 
-  // Used to recognizes one of the literals "true", "false", or "null". These
-  // are the only valid JSON identifiers (productions JSONBooleanLiteral,
-  // JSONNullLiteral).
-  Token::Value ScanJsonIdentifier(const char* text, Token::Value token);
-
   // Parse a single JSON value from input (grammar production JSONValue).
   // A JSON value is either a (double-quoted) string literal, a number literal,
   // one of "true", "false", or "null", or an object or array literal.
@@ -119,43 +144,455 @@
 
   // Mark that a parsing error has happened at the current token, and
   // return a null handle. Primarily for readability.
-  Handle<Object> ReportUnexpectedToken() { return Handle<Object>::null(); }
+  inline Handle<Object> ReportUnexpectedCharacter() {
+    return Handle<Object>::null();
+  }
 
-  // Peek at the next token.
-  Token::Value Peek() { return next_.token; }
-  // Scan the next token and return the token scanned on the last call.
-  Token::Value Next();
+  inline Isolate* isolate() { return isolate_; }
 
-  struct TokenInfo {
-    TokenInfo() : token(Token::ILLEGAL),
-                  beg_pos(0),
-                  end_pos(0) { }
-    Token::Value token;
-    int beg_pos;
-    int end_pos;
-  };
+  static const int kInitialSpecialStringLength = 1024;
 
-  static const int kInitialSpecialStringSize = 100;
 
-
  private:
   Handle<String> source_;
   int source_length_;
   Handle<SeqAsciiString> seq_source_;
 
-  bool is_sequential_ascii_;
-  // Current and next token
-  TokenInfo current_;
-  TokenInfo next_;
   Isolate* isolate_;
   uc32 c0_;
   int position_;
+};
 
+template <bool seq_ascii>
+Handle<Object> JsonParser<seq_ascii>::ParseJson(Handle<String> source) {
+  isolate_ = source->map()->GetHeap()->isolate();
+  source_ = Handle<String>(source->TryFlattenGetString());
+  source_length_ = source_->length();
 
-  Handle<String> string_val_;
-  double number_;
-};
+  // Optimized fast case where we only have ASCII characters.
+  if (seq_ascii) {
+    seq_source_ = Handle<SeqAsciiString>::cast(source_);
+  }
 
+  // Set initial position right before the string.
+  position_ = -1;
+  // Advance to the first character (posibly EOS)
+  AdvanceSkipWhitespace();
+  Handle<Object> result = ParseJsonValue();
+  if (result.is_null() || c0_ != kEndOfString) {
+    // Parse failed. Current character is the unexpected token.
+
+    const char* message;
+    Factory* factory = isolate()->factory();
+    Handle<JSArray> array;
+
+    switch (c0_) {
+      case kEndOfString:
+        message = "unexpected_eos";
+        array = factory->NewJSArray(0);
+        break;
+      case '-':
+      case '0':
+      case '1':
+      case '2':
+      case '3':
+      case '4':
+      case '5':
+      case '6':
+      case '7':
+      case '8':
+      case '9':
+        message = "unexpected_token_number";
+        array = factory->NewJSArray(0);
+        break;
+      case '"':
+        message = "unexpected_token_string";
+        array = factory->NewJSArray(0);
+        break;
+      default:
+        message = "unexpected_token";
+        Handle<Object> name = LookupSingleCharacterStringFromCode(c0_);
+        Handle<FixedArray> element = factory->NewFixedArray(1);
+        element->set(0, *name);
+        array = factory->NewJSArrayWithElements(element);
+        break;
+    }
+
+    MessageLocation location(factory->NewScript(source),
+                             position_,
+                             position_ + 1);
+    Handle<Object> result = factory->NewSyntaxError(message, array);
+    isolate()->Throw(*result, &location);
+    return Handle<Object>::null();
+  }
+  return result;
+}
+
+
+// Parse any JSON value.
+template <bool seq_ascii>
+Handle<Object> JsonParser<seq_ascii>::ParseJsonValue() {
+  switch (c0_) {
+    case '"':
+      return ParseJsonString();
+    case '-':
+    case '0':
+    case '1':
+    case '2':
+    case '3':
+    case '4':
+    case '5':
+    case '6':
+    case '7':
+    case '8':
+    case '9':
+      return ParseJsonNumber();
+    case 'f':
+      if (AdvanceGetChar() == 'a' && AdvanceGetChar() == 'l' &&
+          AdvanceGetChar() == 's' && AdvanceGetChar() == 'e') {
+        AdvanceSkipWhitespace();
+        return isolate()->factory()->false_value();
+      } else {
+        return ReportUnexpectedCharacter();
+      }
+    case 't':
+      if (AdvanceGetChar() == 'r' && AdvanceGetChar() == 'u' &&
+          AdvanceGetChar() == 'e') {
+        AdvanceSkipWhitespace();
+        return isolate()->factory()->true_value();
+      } else {
+        return ReportUnexpectedCharacter();
+      }
+    case 'n':
+      if (AdvanceGetChar() == 'u' && AdvanceGetChar() == 'l' &&
+          AdvanceGetChar() == 'l') {
+        AdvanceSkipWhitespace();
+        return isolate()->factory()->null_value();
+      } else {
+        return ReportUnexpectedCharacter();
+      }
+    case '{':
+      return ParseJsonObject();
+    case '[':
+      return ParseJsonArray();
+    default:
+      return ReportUnexpectedCharacter();
+  }
+}
+
+
+// Parse a JSON object. Position must be right at '{'.
+template <bool seq_ascii>
+Handle<Object> JsonParser<seq_ascii>::ParseJsonObject() {
+  Handle<JSFunction> object_constructor(
+      isolate()->global_context()->object_function());
+  Handle<JSObject> json_object =
+      isolate()->factory()->NewJSObject(object_constructor);
+  ASSERT_EQ(c0_, '{');
+
+  AdvanceSkipWhitespace();
+  if (c0_ != '}') {
+    do {
+      if (c0_ != '"') return ReportUnexpectedCharacter();
+      Handle<String> key = ParseJsonSymbol();
+      if (key.is_null() || c0_ != ':') return ReportUnexpectedCharacter();
+      AdvanceSkipWhitespace();
+      Handle<Object> value = ParseJsonValue();
+      if (value.is_null()) return ReportUnexpectedCharacter();
+
+      uint32_t index;
+      if (key->AsArrayIndex(&index)) {
+        SetOwnElement(json_object, index, value, kNonStrictMode);
+      } else if (key->Equals(isolate()->heap()->Proto_symbol())) {
+        SetPrototype(json_object, value);
+      } else {
+        SetLocalPropertyIgnoreAttributes(json_object, key, value, NONE);
+      }
+    } while (MatchSkipWhiteSpace(','));
+    if (c0_ != '}') {
+      return ReportUnexpectedCharacter();
+    }
+  }
+  AdvanceSkipWhitespace();
+  return json_object;
+}
+
+// Parse a JSON array. Position must be right at '['.
+template <bool seq_ascii>
+Handle<Object> JsonParser<seq_ascii>::ParseJsonArray() {
+  ZoneScope zone_scope(isolate(), DELETE_ON_EXIT);
+  ZoneList<Handle<Object> > elements(4);
+  ASSERT_EQ(c0_, '[');
+
+  AdvanceSkipWhitespace();
+  if (c0_ != ']') {
+    do {
+      Handle<Object> element = ParseJsonValue();
+      if (element.is_null()) return ReportUnexpectedCharacter();
+      elements.Add(element);
+    } while (MatchSkipWhiteSpace(','));
+    if (c0_ != ']') {
+      return ReportUnexpectedCharacter();
+    }
+  }
+  AdvanceSkipWhitespace();
+  // Allocate a fixed array with all the elements.
+  Handle<FixedArray> fast_elements =
+      isolate()->factory()->NewFixedArray(elements.length());
+  for (int i = 0, n = elements.length(); i < n; i++) {
+    fast_elements->set(i, *elements[i]);
+  }
+  return isolate()->factory()->NewJSArrayWithElements(fast_elements);
+}
+
+
+template <bool seq_ascii>
+Handle<Object> JsonParser<seq_ascii>::ParseJsonNumber() {
+  bool negative = false;
+  int beg_pos = position_;
+  if (c0_ == '-') {
+    Advance();
+    negative = true;
+  }
+  if (c0_ == '0') {
+    Advance();
+    // Prefix zero is only allowed if it's the only digit before
+    // a decimal point or exponent.
+    if ('0' <= c0_ && c0_ <= '9') return ReportUnexpectedCharacter();
+  } else {
+    int i = 0;
+    int digits = 0;
+    if (c0_ < '1' || c0_ > '9') return ReportUnexpectedCharacter();
+    do {
+      i = i * 10 + c0_ - '0';
+      digits++;
+      Advance();
+    } while (c0_ >= '0' && c0_ <= '9');
+    if (c0_ != '.' && c0_ != 'e' && c0_ != 'E' && digits < 10) {
+      SkipWhitespace();
+      return Handle<Smi>(Smi::FromInt((negative ? -i : i)), isolate());
+    }
+  }
+  if (c0_ == '.') {
+    Advance();
+    if (c0_ < '0' || c0_ > '9') return ReportUnexpectedCharacter();
+    do {
+      Advance();
+    } while (c0_ >= '0' && c0_ <= '9');
+  }
+  if (AsciiAlphaToLower(c0_) == 'e') {
+    Advance();
+    if (c0_ == '-' || c0_ == '+') Advance();
+    if (c0_ < '0' || c0_ > '9') return ReportUnexpectedCharacter();
+    do {
+      Advance();
+    } while (c0_ >= '0' && c0_ <= '9');
+  }
+  int length = position_ - beg_pos;
+  double number;
+  if (seq_ascii) {
+    Vector<const char> chars(seq_source_->GetChars() +  beg_pos, length);
+    number = StringToDouble(isolate()->unicode_cache(),
+                             chars,
+                             NO_FLAGS,  // Hex, octal or trailing junk.
+                             OS::nan_value());
+  } else {
+    Vector<char> buffer = Vector<char>::New(length);
+    String::WriteToFlat(*source_, buffer.start(), beg_pos, position_);
+    Vector<const char> result =
+        Vector<const char>(reinterpret_cast<const char*>(buffer.start()),
+        length);
+    number = StringToDouble(isolate()->unicode_cache(),
+                             result,
+                             NO_FLAGS,  // Hex, octal or trailing junk.
+                             0.0);
+    buffer.Dispose();
+  }
+  SkipWhitespace();
+  return isolate()->factory()->NewNumber(number);
+}
+
+
+template <typename StringType>
+inline void SeqStringSet(Handle<StringType> seq_str, int i, uc32 c);
+
+template <>
+inline void SeqStringSet(Handle<SeqTwoByteString> seq_str, int i, uc32 c) {
+  seq_str->SeqTwoByteStringSet(i, c);
+}
+
+template <>
+inline void SeqStringSet(Handle<SeqAsciiString> seq_str, int i, uc32 c) {
+  seq_str->SeqAsciiStringSet(i, c);
+}
+
+template <typename StringType>
+inline Handle<StringType> NewRawString(Factory* factory, int length);
+
+template <>
+inline Handle<SeqTwoByteString> NewRawString(Factory* factory, int length) {
+  return factory->NewRawTwoByteString(length, NOT_TENURED);
+}
+
+template <>
+inline Handle<SeqAsciiString> NewRawString(Factory* factory, int length) {
+  return factory->NewRawAsciiString(length, NOT_TENURED);
+}
+
+
+// Scans the rest of a JSON string starting from position_ and writes
+// prefix[start..end] along with the scanned characters into a
+// sequential string of type StringType.
+template <bool seq_ascii>
+template <typename StringType, typename SinkChar>
+Handle<String> JsonParser<seq_ascii>::SlowScanJsonString(
+    Handle<String> prefix, int start, int end) {
+  int count = end - start;
+  int max_length = count + source_length_ - position_;
+  int length = Min(max_length, Max(kInitialSpecialStringLength, 2 * count));
+  Handle<StringType> seq_str = NewRawString<StringType>(isolate()->factory(),
+                                                        length);
+  // Copy prefix into seq_str.
+  SinkChar* dest = seq_str->GetChars();
+  String::WriteToFlat(*prefix, dest, start, end);
+
+  while (c0_ != '"') {
+    if (count >= length) {
+      // We need to create a longer sequential string for the result.
+      return SlowScanJsonString<StringType, SinkChar>(seq_str, 0, count);
+    }
+    // Check for control character (0x00-0x1f) or unterminated string (<0).
+    if (c0_ < 0x20) return Handle<String>::null();
+    if (c0_ != '\\') {
+      // If the sink can contain UC16 characters, or source_ contains only
+      // ASCII characters, there's no need to test whether we can store the
+      // character. Otherwise check whether the UC16 source character can fit
+      // in the ASCII sink.
+      if (sizeof(SinkChar) == kUC16Size ||
+          seq_ascii ||
+          c0_ <= kMaxAsciiCharCode) {
+        SeqStringSet(seq_str, count++, c0_);
+        Advance();
+      } else {
+        // StringType is SeqAsciiString and we just read a non-ASCII char.
+        return SlowScanJsonString<SeqTwoByteString, uc16>(seq_str, 0, count);
+      }
+    } else {
+      Advance();  // Advance past the \.
+      switch (c0_) {
+        case '"':
+        case '\\':
+        case '/':
+          SeqStringSet(seq_str, count++, c0_);
+          break;
+        case 'b':
+          SeqStringSet(seq_str, count++, '\x08');
+          break;
+        case 'f':
+          SeqStringSet(seq_str, count++, '\x0c');
+          break;
+        case 'n':
+          SeqStringSet(seq_str, count++, '\x0a');
+          break;
+        case 'r':
+          SeqStringSet(seq_str, count++, '\x0d');
+          break;
+        case 't':
+          SeqStringSet(seq_str, count++, '\x09');
+          break;
+        case 'u': {
+          uc32 value = 0;
+          for (int i = 0; i < 4; i++) {
+            Advance();
+            int digit = HexValue(c0_);
+            if (digit < 0) {
+              return Handle<String>::null();
+            }
+            value = value * 16 + digit;
+          }
+          if (sizeof(SinkChar) == kUC16Size || value <= kMaxAsciiCharCode) {
+            SeqStringSet(seq_str, count++, value);
+            break;
+          } else {
+            // StringType is SeqAsciiString and we just read a non-ASCII char.
+            position_ -= 6;  // Rewind position_ to \ in \uxxxx.
+            Advance();
+            return SlowScanJsonString<SeqTwoByteString, uc16>(seq_str,
+                                                              0,
+                                                              count);
+          }
+        }
+        default:
+          return Handle<String>::null();
+      }
+      Advance();
+    }
+  }
+  // Shrink seq_string length to count.
+  if (isolate()->heap()->InNewSpace(*seq_str)) {
+    isolate()->heap()->new_space()->
+        template ShrinkStringAtAllocationBoundary<StringType>(
+            *seq_str, count);
+  } else {
+    int string_size = StringType::SizeFor(count);
+    int allocated_string_size = StringType::SizeFor(length);
+    int delta = allocated_string_size - string_size;
+    Address start_filler_object = seq_str->address() + string_size;
+    seq_str->set_length(count);
+    isolate()->heap()->CreateFillerObjectAt(start_filler_object, delta);
+  }
+  ASSERT_EQ('"', c0_);
+  // Advance past the last '"'.
+  AdvanceSkipWhitespace();
+  return seq_str;
+}
+
+
+template <bool seq_ascii>
+template <bool is_symbol>
+Handle<String> JsonParser<seq_ascii>::ScanJsonString() {
+  ASSERT_EQ('"', c0_);
+  Advance();
+  if (c0_ == '"') {
+    AdvanceSkipWhitespace();
+    return Handle<String>(isolate()->heap()->empty_string());
+  }
+  int beg_pos = position_;
+  // Fast case for ASCII only without escape characters.
+  do {
+    // Check for control character (0x00-0x1f) or unterminated string (<0).
+    if (c0_ < 0x20) return Handle<String>::null();
+    if (c0_ != '\\') {
+      if (seq_ascii || c0_ <= kMaxAsciiCharCode) {
+        Advance();
+      } else {
+        return SlowScanJsonString<SeqTwoByteString, uc16>(source_,
+                                                          beg_pos,
+                                                          position_);
+      }
+    } else {
+      return SlowScanJsonString<SeqAsciiString, char>(source_,
+                                                      beg_pos,
+                                                      position_);
+    }
+  } while (c0_ != '"');
+  int length = position_ - beg_pos;
+  Handle<String> result;
+  if (seq_ascii && is_symbol) {
+    result = isolate()->factory()->LookupAsciiSymbol(seq_source_,
+                                                     beg_pos,
+                                                     length);
+  } else {
+    result = isolate()->factory()->NewRawAsciiString(length);
+    char* dest = SeqAsciiString::cast(*result)->GetChars();
+    String::WriteToFlat(*source_, dest, beg_pos, position_);
+  }
+  ASSERT_EQ('"', c0_);
+  // Advance past the last '"'.
+  AdvanceSkipWhitespace();
+  return result;
+}
+
 } }  // namespace v8::internal
 
 #endif  // V8_JSON_PARSER_H_