Replace old utf8 decoder with new one.

sdk/lib/convert/utf.dart

 // Copyright (c) 2013, the Dart project authors.  Please see the AUTHORS file
 // for details. All rights reserved. Use of this source code is governed by a
 // BSD-style license that can be found in the LICENSE file.
 
 part of dart.convert;
 
 /**
  * A [Utf8Encoder] converts strings to their UTF-8 code units (a list of
  * unsigned 8-bit integers).
  */
 class Utf8Encoder extends Converter<String, List<int>> {
   /**
    * Converts [string] to its UTF-8 code units (a list of
    * unsigned 8-bit integers).
    */
   List<int> convert(String string) => OLD_UTF_LIB.encodeUtf8(string);
 }
 
 /**
  * A [Utf8Decoder] converts UTF-8 code units (lists of unsigned 8-bit integers)
  * to a string.
  */
 class Utf8Decoder extends Converter<List<int>, String> {
+  final bool _allowMalformed;
+
+  /**
+   * Instantiates a new [Utf8Decoder].
+   *
+   * The optional [allowMalformed] argument defines how [convert] deals
+   * with invalid or unterminated character sequences.
+   *
+   * If it is `true` [convert] replaces invalid (or unterminated) character
+   * sequences with the Unicode Replacement character `0xFFFD` (�). Otherwise

[Lasse Reichstein Nielsen, 2013/07/16 12:23:03]

U+FFFD

[floitsch, 2013/07/16 14:25:24]

Done.

+   * it throws a [FormatException].
+   */
+  Utf8Decoder({ bool allowMalformed: false })
+      : this._allowMalformed = allowMalformed;
+
   /**
    * Converts the UTF-8 [codeUnits] (a list of unsigned 8-bit integers) to the
    * corresponding string.
    */
-  // TODO(floitsch): allow to configure the decoder (for example the replacement
-  // character).
-  String convert(List<int> codeUnits) => OLD_UTF_LIB.decodeUtf8(codeUnits);
+  String convert(List<int> codeUnits) {
+    StringBuffer buffer = new StringBuffer();
+    _Utf8Decoder decoder = new _Utf8Decoder(_allowMalformed);
+    decoder.convert(codeUnits, 0, codeUnits.length, buffer);
+    decoder.close(buffer);
+    return buffer.toString();
+  }
 }
+
+// UTF-8 constants.
+const int _ONE_BYTE_LIMIT = 0x7f;   // 7 bytes
+const int _TWO_BYTE_LIMIT = 0x7ff;  // 11 bytes
+const int _THREE_BYTE_LIMIT = 0xffff;  // 16 bytes
+const int _FOUR_BYTE_LIMIT = 0x10ffff;  // 21 bytes, truncated to Unicode max.
+
+// UTF-16 constants.
+const int _SURROGATE_MASK = 0xF800;
+const int _SURROGATE_TAG_MASK = 0xFC00;
+const int _SURROGATE_VALUE_MASK = 0x3FF;
+const int _LEAD_SURROGATE_MIN = 0xD800;
+const int _TAIL_SURROGATE_MIN = 0xDC00;
+
+const int _REPLACEMENT_CHARACTER = 0xFFFD;
+
+bool _isSurrogate(int codeUnit) =>
+    (codeUnit & _SURROGATE_MASK) == _LEAD_SURROGATE_MIN;
+bool _isLeadSurrogate(int codeUnit) =>
+    (codeUnit & _SURROGATE_TAG_MASK) == _LEAD_SURROGATE_MIN;
+bool _isTailSurrogate(int codeUnit) =>
+    (codeUnit & _SURROGATE_TAG_MASK) == _TAIL_SURROGATE_MIN;
+int _combineSurrogatePair(int lead, int tail) =>
+    0x10000 | ((lead & _SURROGATE_VALUE_MASK) << 10)
+            | (tail & _SURROGATE_VALUE_MASK);
+
+
+/**
+ * Decodes utf-8.

[Lasse Reichstein Nielsen, 2013/07/16 12:23:03]

UTF-8.

[floitsch, 2013/07/16 14:25:24]

Done.

+ *
+ * The decoder handles chunked input.
+ */
+// TODO(floitsch): do we want to make this class public?

[Lasse Reichstein Nielsen, 2013/07/16 12:23:03]

Sure, why not?

[floitsch, 2013/07/16 14:25:24]

Later. But rephrased TODO.

+class _Utf8Decoder {
+  final bool _allowMalformed;
+  int _value = 0;
+  int _expectedUnits = 0;
+  int _extraUnits = 0;
+
+  _Utf8Decoder(this._allowMalformed);
+
+  bool get hasPartialInput => _expectedUnits > 0;
+
+  // Limits of one through four byte encodings.
+  static const List<int> _LIMITS = const <int>[
+      _ONE_BYTE_LIMIT,
+      _TWO_BYTE_LIMIT,
+      _THREE_BYTE_LIMIT,
+      _FOUR_BYTE_LIMIT ];
+
+  void close(StringSink sink) {
+    if (hasPartialInput) {
+      _throwIfNecessary("Unfinished UTF-8 encoding");
+      sink.writeCharCode(_REPLACEMENT_CHARACTER);
+    }
+  }
+
+  void convert(List<int> codeUnits, int startIndex, int endIndex,
+               StringSink sink) {
+    int value = _value;
+    int expectedUnits = _expectedUnits;
+    int extraUnits = _extraUnits;
+    _value = 0;
+    _expectedUnits = 0;
+    _extraUnits = 0;
+
+    int i = startIndex;
+    loop: while (true) {
+      multibyte: if (expectedUnits > 0) {
+        do {
+          if (i == endIndex) {
+            break loop;
+          }
+          int unit = codeUnits[i];
+          if ((unit & 0xC0) != 0x80) {
+            expectedUnits = 0;
+            _throwIfNecessary(
+                "Bad UTF-8 encoding 0x${unit.toRadixString(16)}");
+            sink.writeCharCode(_REPLACEMENT_CHARACTER);
+            break multibyte;
+          } else {
+            value = (value << 6) | (unit & 0x3f);
+            expectedUnits--;
+            i++;
+          }
+        } while (expectedUnits > 0);
+        if (value <= _LIMITS[extraUnits - 1]) {
+          // Overly long encoding. The value could be encoded with a shorter
+          // encoding.
+          _throwIfNecessary(
+              "Overlong encoding of 0x${value.toRadixString(16)}");
+          value = _REPLACEMENT_CHARACTER;
+        }
+        sink.writeCharCode(value);
+      }
+
+      while (i < endIndex) {
+        int unit = codeUnits[i++];
+        if (unit <= _ONE_BYTE_LIMIT) {
+          sink.writeCharCode(unit);
+        } else {
+          if ((unit & 0xE0) == 0xC0) {
+            value = unit & 0x1F;
+            expectedUnits = extraUnits = 1;
+            continue loop;
+          }
+          if ((unit & 0xF0) == 0xE0) {
+            value = unit & 0x0F;
+            expectedUnits = extraUnits = 2;
+            continue loop;
+          }
+          if ((unit & 0xF8) == 0xF0) {
+            value = unit & 0x07;
+            expectedUnits = extraUnits = 3;
+            continue loop;
+          }
+          _throwIfNecessary("Bad UTF-8 encoding 0x${unit.toRadixString(16)}");

[Lasse Reichstein Nielsen, 2013/07/16 12:23:03]

Seems inefficient to create the string and not use it.
I'd rather inline the _throwIfNecessary, or move the string interpolation into
_throwIfNecessary (e.g., only pass the constant string before "0x", and the bad
value, as arguments, and build the rest if necessary)

[floitsch, 2013/07/16 14:25:24]

inlined.

+          value = _REPLACEMENT_CHARACTER;
+          expectedUnits = extraUnits = 0;
+          sink.writeCharCode(value);
+        }
+      }
+      break loop;
+    }
+    if (expectedUnits > 0) {
+      _value = value;
+      _expectedUnits = expectedUnits;
+      _extraUnits = extraUnits;
+    }
+  }
+
+  void _throwIfNecessary(String message) {
+    if (!_allowMalformed) {
+      throw new FormatException(message);
+    }
+  }
+}