String:WriteUtf8: Add REPLACE_INVALID_UTF8 option

src/unicode-inl.h

Index: src/unicode-inl.h
diff --git a/src/unicode-inl.h b/src/unicode-inl.h
index f861f9f2d47449945d62a6fbc8044abbcd0b2a2b..0210b93cb9251229262cc380013e7e7245377a7a 100644
--- a/src/unicode-inl.h
+++ b/src/unicode-inl.h
@@ -107,8 +107,13 @@ unsigned Utf8::EncodeOneByte(char* str, uint8_t c) {
   return 2;
 }
 
-
-unsigned Utf8::Encode(char* str, uchar c, int previous) {
+// Encode encodes the UTF-16 code units c and previous into the given str
+// buffer. Unless allow_invalid is set to true, surrogate code points will be
+// replaced with kReplacementCharacter.
+unsigned Utf8::Encode(char* str,
+                      uchar c,
+                      int previous,
+                      bool allow_invalid = true) {

[dcarney, 2014/01/13 09:19:56]

default should be in declaration

[haimuiba, 2014/01/15 10:52:34]

Done.

   static const int kMask = ~(1 << 6);
   if (c <= kMaxOneByteChar) {
     str[0] = c;
@@ -118,12 +123,16 @@ unsigned Utf8::Encode(char* str, uchar c, int previous) {
     str[1] = 0x80 | (c & kMask);
     return 2;
   } else if (c <= kMaxThreeByteChar) {
-    if (Utf16::IsTrailSurrogate(c) &&
-        Utf16::IsLeadSurrogate(previous)) {
+    if (Utf16::IsSurrogatePair(previous, c)) {
       const int kUnmatchedSize = kSizeOfUnmatchedSurrogate;
       return Encode(str - kUnmatchedSize,
                     Utf16::CombineSurrogatePair(previous, c),
-                    Utf16::kNoPreviousCharacter) - kUnmatchedSize;
+                    Utf16::kNoPreviousCharacter,
+                    allow_invalid) - kUnmatchedSize;
+    } else if (!allow_invalid &&
+               (Utf16::IsLeadSurrogate(c) ||
+               Utf16::IsTrailSurrogate(c))) {
+      c = kBadChar;
     }
     str[0] = 0xE0 | (c >> 12);
     str[1] = 0x80 | ((c >> 6) & kMask);