Change VM's string-buffer patch to use a Uin16Array as backing buffer.

runtime/lib/string_buffer_patch.dart

Index: runtime/lib/string_buffer_patch.dart
diff --git a/runtime/lib/string_buffer_patch.dart b/runtime/lib/string_buffer_patch.dart
index 8fadafaa1259f867a224224f38a8a58969df5692..8c5b525d2a9940f4fb70adfd61fdf094a673d247 100644
--- a/runtime/lib/string_buffer_patch.dart
+++ b/runtime/lib/string_buffer_patch.dart
@@ -3,13 +3,22 @@
 // BSD-style license that can be found in the LICENSE file.
 
 patch class StringBuffer {
-  List<String> _buffer;
-  int _length;
+  /** Backing store for collected UTF-16 code units. */
+  Uint16List _buffer;
+  /** Number of code units collected. */
+  int _length = 0;
+  /**
+   * Collects the approximate maximal magnitude of the added code units.
+   *
+   * The value of each added code unit is or'ed with this variable, so the
+   * most significant bit set in any code unit is also set in this value.
+   * If below 256, the string is a Latin-1 string.
+   */
+  int _code_magnitude = 0;

[srdjan, 2013/03/07 17:58:34]

s/_code_magnitude/_codeMagnitude/

[Lasse Reichstein Nielsen, 2013/03/08 07:18:44]

_codeUnitMagnitude even.

 
   /// Creates the string buffer with an initial content.
-  /* patch */ StringBuffer([Object content = ""]) {
-    _buffer = new List<String>();
-    _length = 0;
+  /* patch */ StringBuffer([Object content = ""])
+      : _buffer = new Uint16List(16) {
     write(content);
   }
 
@@ -17,37 +26,75 @@ patch class StringBuffer {
 
   /// Adds [obj] to the buffer.
   /* patch */ void write(Object obj) {
-    // TODO(srdjan): The following four lines could be replaced by
-    // '$obj', but apparently this is too slow on the Dart VM.
     String str;
     if (obj is String) {
       str = obj;
     } else {
+      // TODO(srdjan): The following four lines could be replaced by
+      // '$obj', but apparently this is too slow on the Dart VM.
       str = obj.toString();
       if (str is! String) {
         throw new ArgumentError('toString() did not return a string');
       }
     }
     if (str.isEmpty) return;
-    _buffer.add(str);
+    _ensureCapacity(str.length);
+    for (int i = 0; i < str.length; i++) {
+      int unit = str.codeUnitAt(i);
+      _buffer[_length + i] = unit;
+      _code_magnitude |= unit;
+    }
     _length += str.length;
   }
 
-  /// Clears the string buffer.
+  /* patch */ writeCharCode(int charCode) {
+    if (charCode <= 0xFFFF) {
+      if (charCode < 0) {
+        throw new RangeError.range(charCode, 0, 0x10FFFF);
+      }
+      _ensureCapacity(1);
+      _buffer[_length++] = charCode;
+      _code_magnitude |= charCode;
+    } else {
+      if (charCode > 0x10FFFF) {
+        throw new RangeError.range(charCode, 0, 0x10FFFF);
+      }
+      _ensureCapacity(2);
+      int bits = charCode - 0x10000;
+      _buffer[_length++] = 0xD800 | (bits >> 10);
+      _buffer[_length++] = 0xDC00 | (bits & 0x3FF);
+      _code_magnitude |= 0xFFFF;
+    }
+  }
+
+  /** Makes the buffer empty. */
   /* patch */ void clear() {
-    _buffer = new List<String>();
     _length = 0;
+    _code_magnitude = 0;
   }
 
-  /// Returns the contents of buffer as a concatenated string.
+  /** Returns the contents of buffer as a string. */
   /* patch */ String toString() {
-    if (_buffer.length == 0) return "";
-    if (_buffer.length == 1) return _buffer[0];
-    String result = _StringBase.concatAll(_buffer);
-    _buffer.clear();
-    _buffer.add(result);
-    // Since we track the length at each add operation, there is no
-    // need to update it in this function.
-    return result;
+    if (_length == 0) return "";
+    return _create(_buffer, _length, _code_magnitude <= 0xFF);

[Vyacheslav Egorov (Google), 2013/03/07 13:50:52]

consider:

final isLatin1 =  _code_magnitude <= 0xFF;
_create(..., isLatin1)

[Lasse Reichstein Nielsen, 2013/03/08 07:18:44]

Done.

   }
+
+  /** Ensures that the buffer has enough capacity to contain n code units. */

[srdjan, 2013/03/07 17:58:34]

s/contain/add/

[Lasse Reichstein Nielsen, 2013/03/08 07:18:44]

Done.

+  void _ensureCapacity(int n) {
+    int requiredCapacity = _length + n;
+    if (requiredCapacity <= _buffer.length) return;

[Vyacheslav Egorov (Google), 2013/03/07 13:50:52]

consider

  {
return;
}

and more vertical white space for readability.

[srdjan, 2013/03/07 17:58:34]

I would split the rare case out:
if (requiredCapacity > buffer.length) {
  _grow(requiredCapacity);
}

That way there is a larger chance that ensureCapacity gets inlined.

[Lasse Reichstein Nielsen, 2013/03/08 07:18:44]

Done.

+    int newCapacity = _buffer.length;
+    do {
+      newCapacity *= 2;

[srdjan, 2013/03/07 17:58:34]

Optional: This growth may be too aggressive. Opinions?

[Lasse Reichstein Nielsen, 2013/03/08 07:18:44]

I was wondering if it was too small, and considering *4 or more for the first
steps (16, 64, 256, 1024, maybe even skipping 64). It obviously depends on
usage, and I have no real data. For string literals in the V8 parser, growing
fast in the beginning was an advantage, but string literals are artificially
limited by line length, so they tended to be either short, ~72 chars or very
long.

+    } while (newCapacity < requiredCapacity);
+    List<int> newBuffer = new Uint16List(newCapacity);
+    newBuffer.setRange(0, _length, _buffer);
+    _buffer = newBuffer;
+  }
+
+  /**
+   * Create a [String] from the UFT-16 code units in buffer.
+   */
+  static String _create(List<int> buffer, int length, bool isLatin1)

[srdjan, 2013/03/07 17:58:34]

s/List<int>/Uint16List/

[Lasse Reichstein Nielsen, 2013/03/08 07:18:44]

Done.

+      native "StringBuffer_createStringFromUint16Array";
 }