Implement a 'trial parse' step, that will abort pre-parsing excessively

src/scanner.h

Index: src/scanner.h
diff --git a/src/scanner.h b/src/scanner.h
index 5162fbec12d0bfec37a7c8847e7a6400b84b58c0..61176bcfd95536cfd602ac01c9078bca725748c8 100644
--- a/src/scanner.h
+++ b/src/scanner.h
@@ -89,6 +89,9 @@ class Utf16CharacterStream {
   // Must not be used right after calling SeekForward.
   virtual void PushBack(int32_t code_unit) = 0;
 
+  virtual bool SetBookmark();
+  virtual void ResetToBookmark();
+
  protected:
   static const uc32 kEndOfInput = -1;
 
@@ -268,6 +271,17 @@ class LiteralBuffer {
 
   Handle<String> Internalize(Isolate* isolate) const;
 
+  void CopyFrom(const LiteralBuffer* other) {
+    if (other == nullptr) {
+      Reset();
+    } else {
+      is_one_byte_ = other->is_one_byte_;
+      position_ = other->position_;
+      backing_store_.Dispose();
+      backing_store_ = other->backing_store_.Clone();
+    }
+  }
+
  private:
   static const int kInitialCapacity = 16;
   static const int kGrowthFactory = 4;
@@ -342,6 +356,23 @@ class Scanner {
     bool complete_;
   };
 
+  // Scoped helper for a re-settable bookmark.
+  class BookmarkScope {
+   public:
+    explicit BookmarkScope(Scanner* scanner) : scanner_(scanner) {
+      DCHECK_NOT_NULL(scanner_);
+    }
+    ~BookmarkScope() { scanner_->DropBookmark(); }
+
+    bool Set() { return scanner_->SetBookmark(); }
+    void Reset() { scanner_->ResetToBookmark(); }
+    bool HasBeenSet() { return scanner_->BookmarkHasBeenSet(); }
+    bool HasBeenReset() { return scanner_->BookmarkHasBeenReset(); }
+
+   private:
+    Scanner* scanner_;
+  };

[jochen (gone - plz use gerrit), 2015/04/30 13:20:18]

disallow copy/assign

[vogelheim, 2015/04/30 15:35:05]

Done.

+
   // Representation of an interval of source positions.
   struct Location {
     Location(int b, int e) : beg_pos(b), end_pos(e) { }
@@ -509,6 +540,14 @@ class Scanner {
     current_.raw_literal_chars = NULL;
   }
 
+  // Support BookmarkScope functionality.
+  bool SetBookmark();
+  void ResetToBookmark();
+  bool BookmarkHasBeenSet();
+  bool BookmarkHasBeenReset();
+  void DropBookmark();
+  static void CopyTokenDesc(TokenDesc* to, TokenDesc* from);
+
   // Literal buffer support
   inline void StartLiteral() {
     LiteralBuffer* free_buffer = (current_.literal_chars == &literal_buffer1_) ?
@@ -711,6 +750,35 @@ class Scanner {
   TokenDesc current_;  // desc for current token (as returned by Next())
   TokenDesc next_;     // desc for next token (one token look-ahead)
 
+  // Variables for Scanner::BookmarkScope and the *Bookmark implementation.
+  // These variables contain the scanner state when a bookmark is set.
+  //
+  // We will use bookmark_c0_ as a 'control' variable, where:
+  // - bookmark_c0_ >= 0: A bookmark has been set and this contains c0_.
+  // - bookmark_c0_ == -1: No bookmark has been set.

[jochen (gone - plz use gerrit), 2015/04/30 13:20:18]

how about defining an enum for -1 and -2?

[vogelheim, 2015/04/30 15:35:06]

Done.

+  // - bookmark_c0_ == -2: The bookmark has been applied (ResetToBookmark).
+  //
+  // Which state is being bookmarked? The parser state is distributed over
+  // several variables, roughly like this:
+  //   ...    1234        +       5678 ..... [character stream]
+  //       [current_] [next_] c0_ |      [scanner state]
+  // So when the scanner is logically at the beginning of an expression
+  // like "1234 + 4567", then:
+  // - current_ contains "1234"
+  // - next_ contains "+"
+  // - c0_ contains ' ' (the space between "+" and "5678",
+  // - the source_ character stream points to the beginning of "5678".
+  // To be able to restore this state, we will keep copies of current_, next_,
+  // and c0_; we'll ask the stream to bookmark itself, and we'll copy the
+  // contents of current_'s and next_'s literal buffers to bookmark_*_literal_.
+  TokenDesc bookmark_current_;
+  TokenDesc bookmark_next_;
+  uc32 bookmark_c0_;
+  LiteralBuffer bookmark_current_literal_;
+  LiteralBuffer bookmark_current_raw_literal_;
+  LiteralBuffer bookmark_next_literal_;
+  LiteralBuffer bookmark_next_raw_literal_;
+
   // Input stream. Must be initialized to an Utf16CharacterStream.
   Utf16CharacterStream* source_;