const char *
DecodedBitStreamParser::guessEncoding(unsigned char *bytes, int length) {
+ const bool ASSUME_SHIFT_JIS = false;
+ char const* const PLATFORM_DEFAULT_ENCODING="UTF-8";
+
// Does it start with the UTF-8 byte order mark? then guess it's UTF-8
if (length > 3 && bytes[0] == (unsigned char)0xEF && bytes[1] == (unsigned char)0xBB && bytes[2]
== (unsigned char)0xBF) {
// If we see something else in that second byte, we'll make the risky guess
// that it's UTF-8.
bool canBeISO88591 = true;
+ bool canBeShiftJIS = true;
+ bool canBeUTF8 = true;
+ int utf8BytesLeft = 0;
+ int maybeDoubleByteCount = 0;
+ int maybeSingleByteKatakanaCount = 0;
+ bool sawLatin1Supplement = false;
+ bool sawUTF8Start = false;
bool lastWasPossibleDoubleByteStart = false;
- for (int i = 0; i < length; i++) {
+ for (int i = 0;
+ i < length && (canBeISO88591 || canBeShiftJIS || canBeUTF8);
+ i++) {
int value = bytes[i] & 0xFF;
- if (value >= 0x80 && value <= 0x9F && i < length - 1) {
- canBeISO88591 = false;
- // ISO-8859-1 shouldn't use this, but before we decide it is Shift_JIS,
- // just double check that it is followed by a byte that's valid in
- // the Shift_JIS encoding
+
+ // UTF-8 stuff
+ if (value >= 0x80 && value <= 0xBF) {
+ if (utf8BytesLeft > 0) {
+ utf8BytesLeft--;
+ }
+ } else {
+ if (utf8BytesLeft > 0) {
+ canBeUTF8 = false;
+ }
+ if (value >= 0xC0 && value <= 0xFD) {
+ sawUTF8Start = true;
+ int valueCopy = value;
+ while ((valueCopy & 0x40) != 0) {
+ utf8BytesLeft++;
+ valueCopy <<= 1;
+ }
+ }
+ }
+
+ // Shift_JIS stuff
+
+ if (value >= 0xA1 && value <= 0xDF) {
+ // count the number of characters that might be a Shift_JIS single-byte Katakana character
+ if (!lastWasPossibleDoubleByteStart) {
+ maybeSingleByteKatakanaCount++;
+ }
+ }
+ if (!lastWasPossibleDoubleByteStart &&
+ ((value >= 0xF0 && value <= 0xFF) || value == 0x80 || value == 0xA0)) {
+ canBeShiftJIS = false;
+ }
+ if (((value >= 0x81 && value <= 0x9F) || (value >= 0xE0 && value <= 0xEF))) {
+ // These start double-byte characters in Shift_JIS. Let's see if it's followed by a valid
+ // second byte.
if (lastWasPossibleDoubleByteStart) {
// If we just checked this and the last byte for being a valid double-byte
- // char, don't check starting on this byte. If the this and the last byte
+ // char, don't check starting on this byte. If this and the last byte
// formed a valid pair, then this shouldn't be checked to see if it starts
// a double byte pair of course.
lastWasPossibleDoubleByteStart = false;
// ... otherwise do check to see if this plus the next byte form a valid
// double byte pair encoding a character.
lastWasPossibleDoubleByteStart = true;
- int nextValue = bytes[i + 1] & 0xFF;
- if ((value & 0x1) == 0) {
- // if even, next value should be in [0x9F,0xFC]
- // if not, we'll guess UTF-8
- if (nextValue < 0x9F || nextValue > 0xFC) {
- return UTF8;
- }
+ if (i >= length - 1) {
+ canBeShiftJIS = false;
} else {
- // if odd, next value should be in [0x40,0x9E]
- // if not, we'll guess UTF-8
- if (nextValue < 0x40 || nextValue > 0x9E) {
- return UTF8;
+ int nextValue = bytes[i + 1] & 0xFF;
+ if (nextValue < 0x40 || nextValue > 0xFC) {
+ canBeShiftJIS = false;
+ } else {
+ maybeDoubleByteCount++;
}
+ // There is some conflicting information out there about which bytes can follow which in
+ // double-byte Shift_JIS characters. The rule above seems to be the one that matches practice.
}
}
+ } else {
+ lastWasPossibleDoubleByteStart = false;
}
}
- return canBeISO88591 ? ISO88591 : SHIFT_JIS;
+ if (utf8BytesLeft > 0) {
+ canBeUTF8 = false;
+ }
+
+ // Easy -- if assuming Shift_JIS and no evidence it can't be, done
+ if (canBeShiftJIS && ASSUME_SHIFT_JIS) {
+ return SHIFT_JIS;
+ }
+ if (canBeUTF8 && sawUTF8Start) {
+ return UTF8;
+ }
+ // Distinguishing Shift_JIS and ISO-8859-1 can be a little tough. The crude heuristic is:
+ // - If we saw
+ // - at least 3 bytes that starts a double-byte value (bytes that are rare in ISO-8859-1), or
+ // - over 5% of bytes could be single-byte Katakana (also rare in ISO-8859-1),
+ // - and, saw no sequences that are invalid in Shift_JIS, then we conclude Shift_JIS
+ if (canBeShiftJIS && (maybeDoubleByteCount >= 3 || 20 * maybeSingleByteKatakanaCount > length)) {
+ return SHIFT_JIS;
+ }
+ // Otherwise, we default to ISO-8859-1 unless we know it can't be
+ if (!sawLatin1Supplement && canBeISO88591) {
+ return ISO88591;
+ }
+ // Otherwise, we take a wild guess with platform encoding
+ return PLATFORM_DEFAULT_ENCODING;
}
string DecodedBitStreamParser::decode(ArrayRef<unsigned char> bytes, Version *version) {