Changed keyword token recognition to be done inline in the identifier scanner.

src/scanner.cc

Index: src/scanner.cc
diff --git a/src/scanner.cc b/src/scanner.cc
index 3dae414f9d6c60f59cc3bcfa7f3cc5c3318600fc..ec2b2c3fcf4b5df8bfbb2d4464a727009aad29a8 100644
--- a/src/scanner.cc
+++ b/src/scanner.cc
@@ -194,6 +194,139 @@ void TwoByteStringUTF16Buffer::SeekForward(int pos) {
 
 
 // ----------------------------------------------------------------------------
+// Keyword Matcher
+KeywordMatcher::FirstState KeywordMatcher::first_states_[] = {
+  { "break",  KEYWORD_PREFIX, Token::BREAK },
+  { NULL,     C,              Token::ILLEGAL },
+  { NULL,     D,              Token::ILLEGAL },
+  { "else",   KEYWORD_PREFIX, Token::ELSE },
+  { NULL,     F,              Token::ILLEGAL },
+  { NULL,     UNMATCHABLE,    Token::ILLEGAL },
+  { NULL,     UNMATCHABLE,    Token::ILLEGAL },
+  { NULL,     I,              Token::ILLEGAL },
+  { NULL,     UNMATCHABLE,    Token::ILLEGAL },
+  { NULL,     UNMATCHABLE,    Token::ILLEGAL },
+  { NULL,     UNMATCHABLE,    Token::ILLEGAL },
+  { NULL,     UNMATCHABLE,    Token::ILLEGAL },
+  { NULL,     N,              Token::ILLEGAL },
+  { NULL,     UNMATCHABLE,    Token::ILLEGAL },
+  { NULL,     UNMATCHABLE,    Token::ILLEGAL },
+  { NULL,     UNMATCHABLE,    Token::ILLEGAL },
+  { "return", KEYWORD_PREFIX, Token::RETURN },
+  { "switch", KEYWORD_PREFIX, Token::SWITCH },
+  { NULL,     T,              Token::ILLEGAL },
+  { NULL,     UNMATCHABLE,    Token::ILLEGAL },
+  { NULL,     V,              Token::ILLEGAL },
+  { NULL,     W,              Token::ILLEGAL }
+};
+
+
+void KeywordMatcher::Step(uc32 input) {
+  switch (state_) {
+    case INITIAL: {
+      // matching the first character is the only state with significant fanout.
+      // Match only lower-case letters in range 'b'..'w'.
+      unsigned int offset = input - kFirstCharRangeMin;
+      if (offset < kFirstCharRangeLength) {
+        state_ = first_states_[offset].state;
+        if (state_ == KEYWORD_PREFIX) {
+          keyword_ = first_states_[offset].keyword;
+          counter_ = 1;
+          keyword_token_ = first_states_[offset].token;
+        }
+        return;
+      }
+      break;
+    }
+    case KEYWORD_PREFIX:
+      if (keyword_[counter_] == input) {
+        ASSERT_NE(input, '\0');
+        counter_++;
+        if (keyword_[counter_] == '\0') {
+          state_ = KEYWORD_MATCHED;
+          token_ = keyword_token_;
+        }
+        return;
+      }
+      break;
+    case KEYWORD_MATCHED:
+      token_ = Token::IDENTIFIER;
+      break;
+    case C:
+      if (MatchState(input, 'a', CA)) return;
+      if (MatchState(input, 'o', CO)) return;
+      break;
+    case CA:
+      if (MatchKeywordStart(input, "case", 2, Token::CASE)) return;
+      if (MatchKeywordStart(input, "catch", 2, Token::CATCH)) return;
+      break;
+    case CO:
+      if (MatchState(input, 'n', CON)) return;
+      break;
+    case CON:
+      if (MatchKeywordStart(input, "const", 3, Token::CONST)) return;
+      if (MatchKeywordStart(input, "continue", 3, Token::CONTINUE)) return;
+      break;
+    case D:
+      if (MatchState(input, 'e', DE)) return;
+      if (MatchKeyword(input, 'o', KEYWORD_MATCHED, Token::DO)) return;
+      break;
+    case DE:
+      if (MatchKeywordStart(input, "debugger", 2, Token::DEBUGGER)) return;
+      if (MatchKeywordStart(input, "default", 2, Token::DEFAULT)) return;
+      if (MatchKeywordStart(input, "delete", 2, Token::DELETE)) return;
+      break;
+    case F:
+      if (MatchKeywordStart(input, "false", 1, Token::FALSE_LITERAL)) return;
+      if (MatchKeywordStart(input, "finally", 1, Token::FINALLY)) return;
+      if (MatchKeywordStart(input, "for", 1, Token::FOR)) return;
+      if (MatchKeywordStart(input, "function", 1, Token::FUNCTION)) return;
+      break;
+    case I:
+      if (MatchKeyword(input, 'f', KEYWORD_MATCHED, Token::IF)) return;
+      if (MatchKeyword(input, 'n', IN, Token::IN)) return;
+      break;
+    case IN:
+      token_ = Token::IDENTIFIER;
+      if (MatchKeywordStart(input, "instanceof", 2, Token::INSTANCEOF)) {
+        return;
+      }
+      break;
+    case N:
+      if (MatchKeywordStart(input, "native", 1, Token::NATIVE)) return;
+      if (MatchKeywordStart(input, "new", 1, Token::NEW)) return;
+      if (MatchKeywordStart(input, "null", 1, Token::NULL_LITERAL)) return;
+      break;
+    case T:
+      if (MatchState(input, 'h', TH)) return;
+      if (MatchState(input, 'r', TR)) return;
+      if (MatchKeywordStart(input, "typeof", 1, Token::TYPEOF)) return;
+      break;
+    case TH:
+      if (MatchKeywordStart(input, "this", 2, Token::THIS)) return;
+      if (MatchKeywordStart(input, "throw", 2, Token::THROW)) return;
+      break;
+    case TR:
+      if (MatchKeywordStart(input, "true", 2, Token::TRUE_LITERAL)) return;
+      if (MatchKeyword(input, 'y', KEYWORD_MATCHED, Token::TRY)) return;
+      break;
+    case V:
+      if (MatchKeywordStart(input, "var", 1, Token::VAR)) return;
+      if (MatchKeywordStart(input, "void", 1, Token::VOID)) return;
+      break;
+    case W:
+      if (MatchKeywordStart(input, "while", 1, Token::WHILE)) return;
+      if (MatchKeywordStart(input, "with", 1, Token::WITH)) return;
+      break;
+    default:
+      UNREACHABLE();
+  }
+  // On fallthrough, it's a failure.
+  state_ = UNMATCHABLE;
+}
+
+
+// ----------------------------------------------------------------------------
 // Scanner
 
 Scanner::Scanner(bool pre) : stack_overflow_(false), is_pre_parsing_(pre) {
@@ -855,48 +988,40 @@ uc32 Scanner::ScanIdentifierUnicodeEscape() {
 
 Token::Value Scanner::ScanIdentifier() {
   ASSERT(kIsIdentifierStart.get(c0_));
-  bool has_escapes = false;
 
   StartLiteral();
+  KeywordMatcher keyword_match;
+
   // Scan identifier start character.
   if (c0_ == '\\') {
-    has_escapes = true;
     uc32 c = ScanIdentifierUnicodeEscape();
     // Only allow legal identifier start characters.
     if (!kIsIdentifierStart.get(c)) return Token::ILLEGAL;
     AddChar(c);
+    keyword_match.Fail();
   } else {
     AddChar(c0_);
+    keyword_match.AddChar(c0_);
     Advance();
   }
 
   // Scan the rest of the identifier characters.
   while (kIsIdentifierPart.get(c0_)) {
     if (c0_ == '\\') {
-      has_escapes = true;
       uc32 c = ScanIdentifierUnicodeEscape();
       // Only allow legal identifier part characters.
       if (!kIsIdentifierPart.get(c)) return Token::ILLEGAL;
       AddChar(c);
+      keyword_match.Fail();
     } else {
       AddChar(c0_);
+      keyword_match.AddChar(c0_);
       Advance();
     }
   }
   TerminateLiteral();
 
-  // We don't have any 1-letter keywords (this is probably a common case).
-  if ((next_.literal_end - next_.literal_pos) == 1) {
-    return Token::IDENTIFIER;
-  }
-
-  // If the identifier contains unicode escapes, it must not be
-  // resolved to a keyword.
-  if (has_escapes) {
-    return Token::IDENTIFIER;
-  }
-
-  return Token::Lookup(&literals_.data()[next_.literal_pos]);
+  return keyword_match.token();
 }