Remove scanner abstraction layer from JSON parsing. The change in api.cc is t...

src/json-parser.cc

Index: src/json-parser.cc
===================================================================
--- src/json-parser.cc	(revision 8119)
+++ src/json-parser.cc	(working copy)
@@ -53,50 +53,43 @@
   // Set initial position right before the string.
   position_ = -1;
   // Advance to the first character (posibly EOS)
-  Advance();
-  Next();
+  AdvanceWS();
   Handle<Object> result = ParseJsonValue();
-  if (result.is_null() || Next() != Token::EOS) {
-    // Parse failed. Scanner's current token is the unexpected token.
-    Token::Value token = current_.token;
+  if (result.is_null() || c0_ != kEndOfString) {
+    // Parse failed. Current character is the unexpected token.
 
     const char* message;
-    const char* name_opt = NULL;
 
-    switch (token) {
-      case Token::EOS:
+    switch (c0_) {
+      case kEndOfString:
         message = "unexpected_eos";
         break;
-      case Token::NUMBER:
+      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";
         break;
-      case Token::STRING:
+      case '"':
         message = "unexpected_token_string";

[Lasse Reichstein, 2011/06/01 11:03:58]

This might change the behavior for unterminated strings, but I guess that's
acceptable.

[sandholm, 2011/06/01 13:45:20]

Done.

         break;
-      case Token::IDENTIFIER:
-      case Token::FUTURE_RESERVED_WORD:
-        message = "unexpected_token_identifier";
-        break;
       default:

[Lasse Reichstein, 2011/06/01 11:03:58]

Agree, it's probably better to not have the unexpected_token_identifier message
for JSON.

[sandholm, 2011/06/01 13:45:20]

Done.

         message = "unexpected_token";

[Lasse Reichstein, 2011/06/01 11:03:58]

The unexpected_token message needs/expects second argument. See messages.js.

[sandholm, 2011/06/01 13:45:20]

Done.

-        name_opt = Token::String(token);
-        ASSERT(name_opt != NULL);
         break;
     }
 
     Factory* factory = isolate()->factory();
     MessageLocation location(factory->NewScript(source),
-                             current_.beg_pos,
-                             current_.end_pos);
-    Handle<JSArray> array;
-    if (name_opt == NULL) {
-      array = factory->NewJSArray(0);
-    } else {
-      Handle<String> name = factory->NewStringFromUtf8(CStrVector(name_opt));
-      Handle<FixedArray> element = factory->NewFixedArray(1);
-      element->set(0, *name);
-      array = factory->NewJSArrayWithElements(element);
-    }
+                             position_,
+                             position_ + 1);
+    Handle<JSArray> array = factory->NewJSArray(0);
     Handle<Object> result = factory->NewSyntaxError(message, array);
     isolate()->Throw(*result, &location);
     return Handle<Object>::null();
@@ -107,21 +100,48 @@
 
 // Parse any JSON value.
 Handle<Object> JsonParser::ParseJsonValue() {
-  Token::Value token = Next();
-  switch (token) {
-    case Token::STRING:
-      return GetString(false);
-    case Token::NUMBER:
-      return isolate()->factory()->NewNumber(number_);
-    case Token::FALSE_LITERAL:
-      return isolate()->factory()->false_value();
-    case Token::TRUE_LITERAL:
-      return isolate()->factory()->true_value();
-    case Token::NULL_LITERAL:
-      return isolate()->factory()->null_value();
-    case Token::LBRACE:
+  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') {
+        AdvanceWS();

[Lasse Reichstein, 2011/06/01 11:03:58]

Don't use AdvanceWS here, just check for non-identifier-part of c0_ after
matching 'e', and let the main scan loop skip whitespace.
Make sure that the cursor position is correct (pointing to the 'f') when
reporting the incorrect identifier.

[sandholm, 2011/06/01 13:45:20]

I think this is fine. The invariant is that we're at the next non-whitespace
char. Also, I think it's fine to use whatever position we're when failing. E.g.
with "falsse" we would report the error here: fals(s)e.

+        return isolate()->factory()->false_value();
+      } else {
+        return ReportUnexpectedToken();

[Lasse Reichstein, 2011/06/01 11:03:58]

Maybe change the name, now that we don't use Token values.

[sandholm, 2011/06/01 13:45:20]

Done.

+      }
+    case 't':
+      if (AdvanceGetChar() == 'r' && AdvanceGetChar() == 'u' &&
+          AdvanceGetChar() == 'e') {
+        AdvanceWS();

[Lasse Reichstein, 2011/06/01 11:03:58]

As above.

[sandholm, 2011/06/01 13:45:20]

ditto

+        return isolate()->factory()->true_value();
+      } else {
+        return ReportUnexpectedToken();
+      }
+    case 'n':
+      if (AdvanceGetChar() == 'u' && AdvanceGetChar() == 'l' &&
+          AdvanceGetChar() == 'l') {
+        AdvanceWS();

[Lasse Reichstein, 2011/06/01 11:03:58]

And again.

[sandholm, 2011/06/01 13:45:20]

ditto

+        return isolate()->factory()->null_value();
+      } else {
+        return ReportUnexpectedToken();
+      }
+    case '{':
       return ParseJsonObject();
-    case Token::LBRACK:
+    case '[':
       return ParseJsonArray();
     default:
       return ReportUnexpectedToken();
@@ -129,27 +149,21 @@
 }
 
 
-// Parse a JSON object. Scanner must be right after '{' token.
+// Parse a JSON object. Position must be right at '{'.
 Handle<Object> JsonParser::ParseJsonObject() {
   Handle<JSFunction> object_constructor(
       isolate()->global_context()->object_function());
   Handle<JSObject> json_object =
       isolate()->factory()->NewJSObject(object_constructor);
 
-  if (Peek() == Token::RBRACE) {
-    Next();
-  } else {
+  AdvanceWS();
+  if (c0_ != '}') {
     do {
-      if (Next() != Token::STRING) {
-        return ReportUnexpectedToken();
-      }
-      Handle<String> key = GetString(true);
-      if (Next() != Token::COLON) {
-        return ReportUnexpectedToken();
-      }
-
+      Handle<String> key = ParseJsonSymbol();
+      if (key.is_null() || c0_ != ':') return ReportUnexpectedToken();
+      AdvanceWS();
       Handle<Object> value = ParseJsonValue();
-      if (value.is_null()) return Handle<Object>::null();
+      if (value.is_null()) return ReportUnexpectedToken();
 
       uint32_t index;
       if (key->AsArrayIndex(&index)) {
@@ -159,150 +173,45 @@
       } else {
         SetLocalPropertyIgnoreAttributes(json_object, key, value, NONE);
       }
-    } while (Next() == Token::COMMA);
-    if (current_.token != Token::RBRACE) {
+    } while (c0_ == ',' && AdvanceWS());

[Lasse Reichstein, 2011/06/01 11:03:58]

Ah, so that's why AdvanceWS returns true.
Don't do that! In general, don't have side effects in condition expressions.

[sandholm, 2011/06/01 13:45:20]

Fixed with a AdvanceWhiteSpacesOnlyIfMatch(',') call.

+    if (c0_ != '}') {
       return ReportUnexpectedToken();
     }
   }
+  AdvanceWS();
   return json_object;
 }
 
-// Parse a JSON array. Scanner must be right after '[' token.
+// Parse a JSON array. Position must be right at '['.
 Handle<Object> JsonParser::ParseJsonArray() {

[Lasse Reichstein, 2011/06/01 11:03:58]

If position must be at '[', do 
 ASSERT_EQ(c0_, '[');

[sandholm, 2011/06/01 13:45:20]

Done.

   ZoneScope zone_scope(isolate(), DELETE_ON_EXIT);
   ZoneList<Handle<Object> > elements(4);
 
-  Token::Value token = Peek();
-  if (token == Token::RBRACK) {
-    Next();
-  } else {
+  AdvanceWS();
+  if (c0_ != ']') {
     do {
       Handle<Object> element = ParseJsonValue();
-      if (element.is_null()) return Handle<Object>::null();
+      if (element.is_null()) return ReportUnexpectedToken();
       elements.Add(element);
-      token = Next();
-    } while (token == Token::COMMA);
-    if (token != Token::RBRACK) {
+    } while (c0_ == ',' && AdvanceWS());
+    if (c0_ != ']') {
       return ReportUnexpectedToken();
     }
   }
-
+  AdvanceWS();
   // 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);
 }
 
 
-Token::Value JsonParser::Next() {
-  current_ = next_;
-  ScanJson();
-  return current_.token;
-}
-
-void JsonParser::ScanJson() {
-  if (source_->IsSeqAsciiString()) {
-    is_sequential_ascii_ = true;
-  } else {
-    is_sequential_ascii_ = false;
-  }
-
-  Token::Value token;
-  do {
-    // Remember the position of the next token
-    next_.beg_pos = position_;
-    switch (c0_) {
-      case '\t':
-      case '\r':
-      case '\n':
-      case ' ':
-        Advance();
-        token = Token::WHITESPACE;
-        break;
-      case '{':
-        Advance();
-        token = Token::LBRACE;
-        break;
-      case '}':
-        Advance();
-        token = Token::RBRACE;
-        break;
-      case '[':
-        Advance();
-        token = Token::LBRACK;
-        break;
-      case ']':
-        Advance();
-        token = Token::RBRACK;
-        break;
-      case ':':
-        Advance();
-        token = Token::COLON;
-        break;
-      case ',':
-        Advance();
-        token = Token::COMMA;
-        break;
-      case '"':
-        token = ScanJsonString();
-        break;
-      case '-':
-      case '0':
-      case '1':
-      case '2':
-      case '3':
-      case '4':
-      case '5':
-      case '6':
-      case '7':
-      case '8':
-      case '9':
-        token = ScanJsonNumber();
-        break;
-      case 't':
-        token = ScanJsonIdentifier("true", Token::TRUE_LITERAL);
-        break;
-      case 'f':
-        token = ScanJsonIdentifier("false", Token::FALSE_LITERAL);
-        break;
-      case 'n':
-        token = ScanJsonIdentifier("null", Token::NULL_LITERAL);
-        break;
-      default:
-        if (c0_ < 0) {
-          Advance();
-          token = Token::EOS;
-        } else {
-          Advance();
-          token = Token::ILLEGAL;
-        }
-    }
-  } while (token == Token::WHITESPACE);
-
-  next_.end_pos = position_;
-  next_.token = token;
-}
-
-
-Token::Value JsonParser::ScanJsonIdentifier(const char* text,
-                                            Token::Value token) {
-  while (*text != '\0') {
-    if (c0_ != *text) return Token::ILLEGAL;
-    Advance();
-    text++;
-  }
-  return token;
-}
-
-
-Token::Value JsonParser::ScanJsonNumber() {
+Handle<Object> JsonParser::ParseJsonNumber() {
   bool negative = false;
-
+  beg_pos_ = position_;
   if (c0_ == '-') {
     Advance();
     negative = true;
@@ -311,11 +220,11 @@
     Advance();
     // Prefix zero is only allowed if it's the only digit before
     // a decimal point or exponent.
-    if ('0' <= c0_ && c0_ <= '9') return Token::ILLEGAL;
+    if ('0' <= c0_ && c0_ <= '9') return ReportUnexpectedToken();
   } else {
     int i = 0;
     int digits = 0;
-    if (c0_ < '1' || c0_ > '9') return Token::ILLEGAL;
+    if (c0_ < '1' || c0_ > '9') return ReportUnexpectedToken();
     do {
       i = i * 10 + c0_ - '0';
       digits++;
@@ -323,12 +232,13 @@
     } while (c0_ >= '0' && c0_ <= '9');
     if (c0_ != '.' && c0_ != 'e' && c0_ != 'E' && digits < 10) {
       number_ = (negative ? -i : i);
-      return Token::NUMBER;
+      SkipWS();
+      return isolate()->factory()->NewNumber(number_);
     }
   }
   if (c0_ == '.') {
     Advance();
-    if (c0_ < '0' || c0_ > '9') return Token::ILLEGAL;
+    if (c0_ < '0' || c0_ > '9') return ReportUnexpectedToken();
     do {
       Advance();
     } while (c0_ >= '0' && c0_ <= '9');
@@ -336,37 +246,38 @@
   if (AsciiAlphaToLower(c0_) == 'e') {
     Advance();
     if (c0_ == '-' || c0_ == '+') Advance();
-    if (c0_ < '0' || c0_ > '9') return Token::ILLEGAL;
+    if (c0_ < '0' || c0_ > '9') return ReportUnexpectedToken();
     do {
       Advance();
     } while (c0_ >= '0' && c0_ <= '9');
   }
+  int length = position_ - beg_pos_;
   if (is_sequential_ascii_) {
-    Vector<const char> chars(seq_source_->GetChars() +  next_.beg_pos,
-                             position_ - next_.beg_pos);
+    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(position_ - next_.beg_pos);
-    String::WriteToFlat(*source_, buffer.start(), next_.beg_pos, position_);
+    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()),
-        position_ - next_.beg_pos);
+        length);
     number_ = StringToDouble(isolate()->unicode_cache(),
                              result,
                              NO_FLAGS,  // Hex, octal or trailing junk.
                              0.0);
     buffer.Dispose();
   }
-  return Token::NUMBER;
+  SkipWS();
+  return isolate()->factory()->NewNumber(number_);
 }
 
-Token::Value JsonParser::SlowScanJsonString() {
+Handle<Object> JsonParser::SlowScanJsonString() {
   // The currently scanned ascii characters.
   Handle<String> ascii(isolate()->factory()->NewSubString(source_,
-                                                          next_.beg_pos + 1,
+                                                          beg_pos_,
                                                           position_));
   Handle<String> two_byte =
       isolate()->factory()->NewRawTwoByteString(kInitialSpecialStringSize,
@@ -392,7 +303,7 @@
     }
 
     // Check for control character (0x00-0x1f) or unterminated string (<0).
-    if (c0_ < 0x20) return Token::ILLEGAL;
+    if (c0_ < 0x20) return ReportUnexpectedToken();
     if (c0_ != '\\') {
       seq_two_byte->SeqTwoByteStringSet(count++, c0_);
       Advance();
@@ -425,7 +336,7 @@
             Advance();
             int digit = HexValue(c0_);
             if (digit < 0) {
-              return Token::ILLEGAL;
+              return ReportUnexpectedToken();
             }
             value = value * 16 + digit;
           }
@@ -433,14 +344,14 @@
           break;
         }
         default:
-          return Token::ILLEGAL;
+          return ReportUnexpectedToken();
       }
       Advance();
     }
   }
   // Advance past the last '"'.
   ASSERT_EQ('"', c0_);
-  Advance();
+  AdvanceWS();
 
   // Shrink the the string to our length.
   if (isolate()->heap()->InNewSpace(*seq_two_byte)) {
@@ -456,21 +367,19 @@
     seq_two_byte->set_length(count);
     isolate()->heap()->CreateFillerObjectAt(start_filler_object, delta);
   }
-  string_val_ = isolate()->factory()->NewConsString(ascii, seq_two_byte);
-  return Token::STRING;
+  return isolate()->factory()->NewConsString(ascii, seq_two_byte);
 }
 
 
-Token::Value JsonParser::ScanJsonString() {
+template <bool is_symbol>
+Handle<Object> JsonParser::ScanJsonString() {
   ASSERT_EQ('"', c0_);
-  // Set string_val to null. If string_val is not set we assume an
-  // ascii string begining at next_.beg_pos + 1 to next_.end_pos - 1.
-  string_val_ = Handle<String>::null();
   Advance();
+  beg_pos_ = position_;
   // Fast case for ascii only without escape characters.
   while (c0_ != '"') {
     // Check for control character (0x00-0x1f) or unterminated string (<0).
-    if (c0_ < 0x20) return Token::ILLEGAL;
+    if (c0_ < 0x20) return ReportUnexpectedToken();
     if (c0_ != '\\' && c0_ < kMaxAsciiCharCode) {
       Advance();
     } else {
@@ -478,36 +387,16 @@
     }
   }
   ASSERT_EQ('"', c0_);
+  end_pos_ = position_;
   // Advance past the last '"'.
-  Advance();
-  return Token::STRING;
-}
-
-Handle<String> JsonParser::GetString() {
-  return GetString(false);
-}
-
-Handle<String> JsonParser::GetSymbol() {
-  Handle<String> result = GetString(true);
-  if (result->IsSymbol()) return result;
-  return isolate()->factory()->LookupSymbol(result);
-}
-
-Handle<String> JsonParser::GetString(bool hint_symbol) {
-  // We have a non ascii string, return that.
-  if (!string_val_.is_null()) return string_val_;
-
-  if (is_sequential_ascii_ && hint_symbol) {
-    Handle<SeqAsciiString> seq = Handle<SeqAsciiString>::cast(source_);
-    // The current token includes the '"' in both ends.
-    int length = current_.end_pos - current_.beg_pos - 2;
+  AdvanceWS();
+  if (is_sequential_ascii_ && is_symbol) {
     return isolate()->factory()->LookupAsciiSymbol(seq_source_,
-                                                   current_.beg_pos + 1,
-                                                   length);
+                                                   beg_pos_,
+                                                   end_pos_ - beg_pos_);
+  } else {
+    return isolate()->factory()->NewSubString(source_, beg_pos_, end_pos_);
   }
-  // The current token includes the '"' in both ends.
-  return  isolate()->factory()->NewSubString(
-      source_, current_.beg_pos + 1, current_.end_pos - 1);
 }
 
 } }  // namespace v8::internal