QR Codes can encode text as bits in one of several modes, and can use multiple modes @@ -28,7 +32,7 @@ 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 { @@ -55,11 +59,13 @@ final class DecodedBitStreamParser { private DecodedBitStreamParser() { } - static String decode(byte[] bytes, Version version) throws ReaderException { + static DecoderResult decode(byte[] bytes, Version version, ErrorCorrectionLevel ecLevel, Hashtable hints) + throws ReaderException { BitSource bits = new BitSource(bytes); - StringBuffer result = new StringBuffer(); + StringBuffer result = new StringBuffer(50); CharacterSetECI currentCharacterSetECI = null; boolean fc1InEffect = false; + Vector byteSegments = new Vector(1); Mode mode; do { // While still another segment to read... @@ -67,19 +73,26 @@ final class DecodedBitStreamParser { // 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 + try { + mode = Mode.forBits(bits.readBits(4)); // mode is encoded by 4 bits + } catch (IllegalArgumentException iae) { + throw ReaderException.getInstance(); + } } if (!mode.equals(Mode.TERMINATOR)) { 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.STRUCTURED_APPEND)) { + // not really supported; all we do is ignore it + // Read next 8 bits (symbol sequence #) and 8 bits (parity data), then continue + bits.readBits(16); } else if (mode.equals(Mode.ECI)) { // Count doesn't apply to ECI int value = parseECIValue(bits); - try { - currentCharacterSetECI = CharacterSetECI.getCharacterSetECIByValue(value); - } catch (IllegalArgumentException iae) { - // unsupported... just continue? + currentCharacterSetECI = CharacterSetECI.getCharacterSetECIByValue(value); + if (currentCharacterSetECI == null) { + throw ReaderException.getInstance(); } } else { // How many characters will follow, encoded in this mode? @@ -89,26 +102,17 @@ final class DecodedBitStreamParser { } else if (mode.equals(Mode.ALPHANUMERIC)) { decodeAlphanumericSegment(bits, result, count, fc1InEffect); } else if (mode.equals(Mode.BYTE)) { - decodeByteSegment(bits, result, count, currentCharacterSetECI); + decodeByteSegment(bits, result, count, currentCharacterSetECI, byteSegments, hints); } else if (mode.equals(Mode.KANJI)) { decodeKanjiSegment(bits, result, count); } else { - throw new ReaderException("Unsupported mode indicator"); + throw ReaderException.getInstance(); } } } } 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, ecLevel); } private static void decodeKanjiSegment(BitSource bits, @@ -138,17 +142,19 @@ final class DecodedBitStreamParser { try { result.append(new String(buffer, SHIFT_JIS)); } catch (UnsupportedEncodingException uee) { - throw new ReaderException(SHIFT_JIS + " encoding is not supported on this device"); + throw ReaderException.getInstance(); } } private static void decodeByteSegment(BitSource bits, StringBuffer result, int count, - CharacterSetECI currentCharacterSetECI) throws ReaderException { + CharacterSetECI currentCharacterSetECI, + Vector byteSegments, + Hashtable hints) throws ReaderException { byte[] readBytes = new byte[count]; if (count << 3 > bits.available()) { - throw new ReaderException("Count too large: " + count); + throw ReaderException.getInstance(); } for (int i = 0; i < count; i++) { readBytes[i] = (byte) bits.readBits(8); @@ -160,15 +166,16 @@ final class DecodedBitStreamParser { // upon decoding. I have seen ISO-8859-1 used as well as // Shift_JIS -- without anything like an ECI designator to // give a hint. - encoding = guessEncoding(readBytes); + encoding = guessEncoding(readBytes, hints); } else { encoding = currentCharacterSetECI.getEncodingName(); } try { result.append(new String(readBytes, encoding)); } catch (UnsupportedEncodingException uce) { - throw new ReaderException(uce.toString()); + throw ReaderException.getInstance(); } + byteSegments.addElement(readBytes); } private static void decodeAlphanumericSegment(BitSource bits, @@ -212,7 +219,7 @@ final class DecodedBitStreamParser { // Each 10 bits encodes three digits int threeDigitsBits = bits.readBits(10); if (threeDigitsBits >= 1000) { - throw new ReaderException("Illegal value for 3-digit unit: " + threeDigitsBits); + throw ReaderException.getInstance(); } result.append(ALPHANUMERIC_CHARS[threeDigitsBits / 100]); result.append(ALPHANUMERIC_CHARS[(threeDigitsBits / 10) % 10]); @@ -223,7 +230,7 @@ final class DecodedBitStreamParser { // Two digits left over to read, encoded in 7 bits int twoDigitsBits = bits.readBits(7); if (twoDigitsBits >= 100) { - throw new ReaderException("Illegal value for 2-digit unit: " + twoDigitsBits); + throw ReaderException.getInstance(); } result.append(ALPHANUMERIC_CHARS[twoDigitsBits / 10]); result.append(ALPHANUMERIC_CHARS[twoDigitsBits % 10]); @@ -231,13 +238,19 @@ final class DecodedBitStreamParser { // One digit left over to read int digitBits = bits.readBits(4); if (digitBits >= 10) { - throw new ReaderException("Illegal value for digit unit: " + digitBits); + throw ReaderException.getInstance(); } result.append(ALPHANUMERIC_CHARS[digitBits]); } } - private static String guessEncoding(byte[] bytes) { + private static String guessEncoding(byte[] bytes, Hashtable hints) { + if (hints != null) { + String characterSet = (String) hints.get(DecodeHintType.CHARACTER_SET); + if (characterSet != null) { + return characterSet; + } + } if (ASSUME_SHIFT_JIS) { return SHIFT_JIS; } @@ -255,22 +268,36 @@ final class DecodedBitStreamParser { int length = bytes.length; boolean canBeISO88591 = true; boolean canBeShiftJIS = true; - boolean sawDoubleByteStart = false; + int maybeDoubleByteCount = 0; int maybeSingleByteKatakanaCount = 0; + boolean sawLatin1Supplement = false; boolean lastWasPossibleDoubleByteStart = false; for (int i = 0; i < length && (canBeISO88591 || canBeShiftJIS); i++) { int value = bytes[i] & 0xFF; + if ((value == 0xC2 || value == 0xC3) && i < length - 1) { + // This is really a poor hack. The slightly more exotic characters people might want to put in + // a QR Code, by which I mean the Latin-1 supplement characters (e.g. u-umlaut) have encodings + // that start with 0xC2 followed by [0xA0,0xBF], or start with 0xC3 followed by [0x80,0xBF]. + int nextValue = bytes[i + 1] & 0xFF; + if (nextValue <= 0xBF && ((value == 0xC2 && nextValue >= 0xA0) || (value == 0xC3 && nextValue >= 0x80))) { + sawLatin1Supplement = true; + } + } if (value >= 0x7F && value <= 0x9F) { canBeISO88591 = false; } if (value >= 0xA1 && value <= 0xDF) { // count the number of characters that might be a Shift_JIS single-byte Katakana character - maybeSingleByteKatakanaCount++; + 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) { + 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. - 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 this and the last byte @@ -281,25 +308,33 @@ final class DecodedBitStreamParser { // ... 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 (nextValue < 0x40 || nextValue > 0xFC) { + if (i >= bytes.length - 1) { canBeShiftJIS = false; + } else { + 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. } - // 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; } } // 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 + // - at least three 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) { + 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 (canBeISO88591) { + if (!sawLatin1Supplement && canBeISO88591) { return ISO88591; } // Otherwise, we take a wild guess with UTF-8