Fix Unicode issues in dart2js and dart2dart.

sdk/lib/_internal/compiler/implementation/util/util.dart

Index: sdk/lib/_internal/compiler/implementation/util/util.dart
diff --git a/sdk/lib/_internal/compiler/implementation/util/util.dart b/sdk/lib/_internal/compiler/implementation/util/util.dart
index 64194ea4d4c79f0db3550227497397912ccf5ab8..ed095398a9ad94707341eea0d884a80a026b94a1 100644
--- a/sdk/lib/_internal/compiler/implementation/util/util.dart
+++ b/sdk/lib/_internal/compiler/implementation/util/util.dart
@@ -24,52 +24,76 @@ class SpannableAssertionFailure {
   String toString() => 'Compiler crashed: $message.';
 }
 
-/// Writes the characters of [iterator] on [buffer].  The characters
+/// Writes the characters of [string] on [buffer].  The characters
 /// are escaped as suitable for JavaScript and JSON.  [buffer] is
 /// anything which supports [:add:] and [:addCharCode:], for example,
-/// [StringBuffer].
-void writeJsonEscapedCharsOn(Iterator<int> iterator, buffer, onError(code)) {
-  while (iterator.hasNext) {
-    int code = iterator.next();
-    if (identical(code, $SQ)) {
-      buffer.add(r"\'");
-    } else if (identical(code, $LF)) {
-      buffer.add(r'\n');
-    } else if (identical(code, $CR)) {
-      buffer.add(r'\r');
-    } else if (identical(code, $LS)) {
-      // This Unicode line terminator and $PS are invalid in JS string
-      // literals.
-      buffer.add(r'\u2028');
-    } else if (identical(code, $PS)) {
-      buffer.add(r'\u2029');
-    } else if (identical(code, $BACKSLASH)) {
-      buffer.add(r'\\');
-    } else {
-      if (code > 0xffff) {
-        if (onError != null) onError(code);
-        throw 'Unhandled non-BMP character: ${code.toRadixString(16)}';
+/// [StringBuffer].  Note that JS supports \xnn and \unnnn whereas JSON only
+/// supports the \unnnn notation.  Therefore we use the \unnnn notation.
+

[floitsch, 2012/11/21 16:16:50]

remove new line.

+void writeJsonEscapedCharsOn(String string, buffer) {

[floitsch, 2012/11/21 16:16:50]

var buffer

[erikcorry, 2012/11/22 12:42:15]

Done.

+  void addCodeUnitEscaped(buffer, int code) {
+    assert(code < 0x10000);
+    buffer.add(r'\u');
+    if (code < 0x1000) {
+      buffer.add('0');
+      if (code < 0x100) {
+        buffer.add('0');
+        if (code < 0x10) {
+          buffer.add('0');
+        }
       }
-      // TODO(lrn): Consider whether all codes above 0x7f really need to
-      // be escaped. We build a Dart string here, so it should be a literal
-      // stage that converts it to, e.g., UTF-8 for a JS interpreter.
-      if (code < 0x20) {
-        buffer.add(r'\x');
-        if (code < 0x10) buffer.add('0');
-        buffer.add(code.toRadixString(16));
-      } else if (code >= 0x80) {
-        if (code < 0x100) {
-          buffer.add(r'\x');
+    }
+    buffer.add(code.toRadixString(16));
+  }
+
+  void writeEscaped(String string, buffer) {

[floitsch, 2012/11/21 16:16:50]

writeEscapedOn (or change name above)

[erikcorry, 2012/11/22 12:42:15]

Done.

+    for (int i = 0; i < string.length; i++) {
+      int code = string.charCodeAt(i);
+      if (identical(code, $DQ)) {

[floitsch, 2012/11/21 16:16:50]

nyc: remove identicals.
numbers should never be compared with identical.

[erikcorry, 2012/11/22 12:42:15]

Done.

+        buffer.add(r'\"');
+      } else if (identical(code, $TAB)) {
+        buffer.add(r'\t');
+      } else if (identical(code, $LF)) {
+        buffer.add(r'\n');
+      } else if (identical(code, $CR)) {
+        buffer.add(r'\r');
+      } else if (identical(code, $DEL)) {
+        addCodeUnitEscaped(buffer, $DEL);
+      } else if (identical(code, $LS)) {
+        // This Unicode line terminator and $PS are invalid in JS string
+        // literals.
+        addCodeUnitEscaped(buffer, $LS);  // 0x2028.
+      } else if (identical(code, $PS)) {
+        addCodeUnitEscaped(buffer, $PS);  // 0x2029.
+      } else if (identical(code, $BACKSLASH)) {
+        buffer.add(r'\\');
+      } else {
+        if (code < 0x20) {
+          addCodeUnitEscaped(buffer, code);
+          // We emit DEL (ASCII 0x7f) as an escape because it would be confusing
+          // to have it unescaped in a string literal.  We also escape
+          // everything above 0x7f because that means we don't have to worry
+          // about whether the web server serves it up as Latin1 or UTF-8.
+        } else if (code < 0x7f) {
+          buffer.addCharCode(code);
         } else {
-          buffer.add(r'\u');
-          if (code < 0x1000) {
-            buffer.add('0');
-          }
+          // This will output surrogate pairs in the form \udxxx\udyyy, rather
+          // than the more logical \u{zzzzzz}.  This should work in JavaScript
+          // (especially old UCS-2 based implementations) and is the only
+          // format that is allowed in JSON.
+          addCodeUnitEscaped(buffer, code);
         }
-        buffer.add(code.toRadixString(16));
-      } else {
-        buffer.addCharCode(code);
       }
     }
   }
+
+  for (int i = 0; i < string.length; i++) {
+    int code = string.charCodeAt(i);
+    if (code < 0x20 || code == $DEL || code == $DQ || code == $LS ||
+        code == $PS || code == $BACKSLASH || code >= 0x80) {
+      writeEscaped(string, buffer);
+      return;
+    }
+  }
+  buffer.add(string);
 }