QR Codes can encode text as bits in one of several modes, and can use multiple modes @@ -27,7 +30,7 @@ import java.io.UnsupportedEncodingException; * *
See ISO 18004:2006, 6.4.3 - 6.4.7
* - * @author srowen@google.com (Sean Owen) + * @author Sean Owen */ final class DecodedBitStreamParser { @@ -54,23 +57,28 @@ final class DecodedBitStreamParser { private DecodedBitStreamParser() { } - static String decode(byte[] bytes, Version version) throws ReaderException { + static DecoderResult decode(byte[] bytes, Version version) throws ReaderException { BitSource bits = new BitSource(bytes); StringBuffer result = new StringBuffer(); CharacterSetECI currentCharacterSetECI = null; + boolean fc1InEffect = false; + Vector byteSegments = new Vector(1); Mode mode; do { // While still another segment to read... - if (bits.available() == 0) { + if (bits.available() < 4) { // OK, assume we're done. Really, a TERMINATOR mode should have been recorded here mode = Mode.TERMINATOR; } else { mode = Mode.forBits(bits.readBits(4)); // mode is encoded by 4 bits } if (!mode.equals(Mode.TERMINATOR)) { - if (mode.equals(Mode.ECI)) { + if (mode.equals(Mode.FNC1_FIRST_POSITION) || mode.equals(Mode.FNC1_SECOND_POSITION)) { + // We do little with FNC1 except alter the parsed result a bit according to the spec + fc1InEffect = true; + } else if (mode.equals(Mode.ECI)) { // Count doesn't apply to ECI - int value = ECI.parseECI(bits); + int value = parseECIValue(bits); try { currentCharacterSetECI = CharacterSetECI.getCharacterSetECIByValue(value); } catch (IllegalArgumentException iae) { @@ -78,13 +86,13 @@ final class DecodedBitStreamParser { } } else { // How many characters will follow, encoded in this mode? - int count = bits.readBits(mode.getCharacterCountBits(version)); + int count = bits.readBits(mode.getCharacterCountBits(version)); if (mode.equals(Mode.NUMERIC)) { decodeNumericSegment(bits, result, count); } else if (mode.equals(Mode.ALPHANUMERIC)) { - decodeAlphanumericSegment(bits, result, count); + decodeAlphanumericSegment(bits, result, count, fc1InEffect); } else if (mode.equals(Mode.BYTE)) { - decodeByteSegment(bits, result, count, currentCharacterSetECI); + decodeByteSegment(bits, result, count, currentCharacterSetECI, byteSegments); } else if (mode.equals(Mode.KANJI)) { decodeKanjiSegment(bits, result, count); } else { @@ -94,16 +102,7 @@ final class DecodedBitStreamParser { } } while (!mode.equals(Mode.TERMINATOR)); - // I thought it wasn't allowed to leave extra bytes after the terminator but it happens - /* - int bitsLeft = bits.available(); - if (bitsLeft > 0) { - if (bitsLeft > 6 || bits.readBits(bitsLeft) != 0) { - throw new ReaderException("Excess bits or non-zero bits after terminator mode indicator"); - } - } - */ - return result.toString(); + return new DecoderResult(bytes, result.toString(), byteSegments.isEmpty() ? null : byteSegments); } private static void decodeKanjiSegment(BitSource bits, @@ -140,7 +139,8 @@ final class DecodedBitStreamParser { private static void decodeByteSegment(BitSource bits, StringBuffer result, int count, - CharacterSetECI currentCharacterSetECI) throws ReaderException { + CharacterSetECI currentCharacterSetECI, + Vector byteSegments) throws ReaderException { byte[] readBytes = new byte[count]; if (count << 3 > bits.available()) { throw new ReaderException("Count too large: " + count); @@ -164,12 +164,15 @@ final class DecodedBitStreamParser { } catch (UnsupportedEncodingException uce) { throw new ReaderException(uce.toString()); } + byteSegments.addElement(readBytes); } private static void decodeAlphanumericSegment(BitSource bits, StringBuffer result, - int count) { + int count, + boolean fc1InEffect) { // Read two characters at a time + int start = result.length(); while (count > 1) { int nextTwoCharsBits = bits.readBits(11); result.append(ALPHANUMERIC_CHARS[nextTwoCharsBits / 45]); @@ -180,6 +183,21 @@ final class DecodedBitStreamParser { // special case: one character left result.append(ALPHANUMERIC_CHARS[bits.readBits(6)]); } + // See section 6.4.8.1, 6.4.8.2 + if (fc1InEffect) { + // We need to massage the result a bit if in an FNC1 mode: + for (int i = start; i < result.length(); i++) { + if (result.charAt(i) == '%') { + if (i < result.length() - 1 && result.charAt(i + 1) == '%') { + // %% is rendered as % + result.deleteCharAt(i + 1); + } else { + // In alpha mode, % should be converted to FNC1 separator 0x1D + result.setCharAt(i, (char) 0x1D); + } + } + } + } } private static void decodeNumericSegment(BitSource bits, @@ -232,17 +250,31 @@ final class DecodedBitStreamParser { // that it's UTF-8. int length = bytes.length; boolean canBeISO88591 = true; + boolean canBeShiftJIS = true; + boolean sawDoubleByteStart = false; + int maybeSingleByteKatakanaCount = 0; boolean lastWasPossibleDoubleByteStart = false; - for (int i = 0; i < length; i++) { + for (int i = 0; i < length && (canBeISO88591 || canBeShiftJIS); i++) { int value = bytes[i] & 0xFF; - if (value >= 0x80 && value <= 0x9F && i < length - 1) { + if (value >= 0x7F && value <= 0x9F) { 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 + } + 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)) && i < length - 1) { + // These start double-byte characters in Shift_JIS. Let's see if it's followed by a valid + // second byte. + sawDoubleByteStart = true; 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; @@ -251,23 +283,47 @@ final class DecodedBitStreamParser { // 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; - } - } else { - // if odd, next value should be in [0x40,0x9E] - // if not, we'll guess UTF-8 - if (nextValue < 0x40 || nextValue > 0x9E) { - return UTF8; - } + if (nextValue < 0x40 || nextValue > 0xFC) { + canBeShiftJIS = false; } + // 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; + // Distinguishing Shift_JIS and ISO-8859-1 can be a little tough. The crude heuristic is: + // - If we saw + // - at least one byte that starts a double-byte value (bytes that are rare in ISO-8859-1), or + // - over 5% of bytes that 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 ((sawDoubleByteStart || 20 * maybeSingleByteKatakanaCount > length) && canBeShiftJIS) { + return SHIFT_JIS; + } + // Otherwise, we default to ISO-8859-1 unless we know it can't be + if (canBeISO88591) { + return ISO88591; + } + // Otherwise, we take a wild guess with UTF-8 + return UTF8; + } + + private static int parseECIValue(BitSource bits) { + int firstByte = bits.readBits(8); + if ((firstByte & 0x80) == 0) { + // just one byte + return firstByte & 0x7F; + } else if ((firstByte & 0xC0) == 0x80) { + // two bytes + int secondByte = bits.readBits(8); + return ((firstByte & 0x3F) << 8) | secondByte; + } else if ((firstByte & 0xE0) == 0xC0) { + // three bytes + int secondThirdBytes = bits.readBits(16); + return ((firstByte & 0x1F) << 16) | secondThirdBytes; + } + throw new IllegalArgumentException("Bad ECI bits starting with byte " + firstByte); } }