2 * Copyright 2007 ZXing authors
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at
8 * http://www.apache.org/licenses/LICENSE-2.0
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
17 package com.google.zxing.qrcode.decoder;
19 import com.google.zxing.DecodeHintType;
20 import com.google.zxing.FormatException;
21 import com.google.zxing.common.BitSource;
22 import com.google.zxing.common.CharacterSetECI;
23 import com.google.zxing.common.DecoderResult;
25 import java.io.UnsupportedEncodingException;
26 import java.util.Hashtable;
27 import java.util.Vector;
30 * <p>QR Codes can encode text as bits in one of several modes, and can use multiple modes
31 * in one QR Code. This class decodes the bits back into text.</p>
33 * <p>See ISO 18004:2006, 6.4.3 - 6.4.7</p>
37 final class DecodedBitStreamParser {
40 * See ISO 18004:2006, 6.4.4 Table 5
42 private static final char[] ALPHANUMERIC_CHARS = {
43 '0', '1', '2', '3', '4', '5', '6', '7', '8', '9', 'A', 'B',
44 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M', 'N',
45 'O', 'P', 'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X', 'Y', 'Z',
46 ' ', '$', '%', '*', '+', '-', '.', '/', ':'
48 private static final String SHIFT_JIS = "SJIS";
49 private static final String EUC_JP = "EUC_JP";
50 private static final boolean ASSUME_SHIFT_JIS;
51 private static final String UTF8 = "UTF8";
52 private static final String ISO88591 = "ISO8859_1";
55 String platformDefault = System.getProperty("file.encoding");
56 ASSUME_SHIFT_JIS = SHIFT_JIS.equalsIgnoreCase(platformDefault) || EUC_JP.equalsIgnoreCase(platformDefault);
59 private DecodedBitStreamParser() {
62 static DecoderResult decode(byte[] bytes, Version version, ErrorCorrectionLevel ecLevel, Hashtable hints)
63 throws FormatException {
64 BitSource bits = new BitSource(bytes);
65 StringBuffer result = new StringBuffer(50);
66 CharacterSetECI currentCharacterSetECI = null;
67 boolean fc1InEffect = false;
68 Vector byteSegments = new Vector(1);
71 // While still another segment to read...
72 if (bits.available() < 4) {
73 // OK, assume we're done. Really, a TERMINATOR mode should have been recorded here
74 mode = Mode.TERMINATOR;
77 mode = Mode.forBits(bits.readBits(4)); // mode is encoded by 4 bits
78 } catch (IllegalArgumentException iae) {
79 throw FormatException.getFormatInstance();
82 if (!mode.equals(Mode.TERMINATOR)) {
83 if (mode.equals(Mode.FNC1_FIRST_POSITION) || mode.equals(Mode.FNC1_SECOND_POSITION)) {
84 // We do little with FNC1 except alter the parsed result a bit according to the spec
86 } else if (mode.equals(Mode.STRUCTURED_APPEND)) {
87 // not really supported; all we do is ignore it
88 // Read next 8 bits (symbol sequence #) and 8 bits (parity data), then continue
90 } else if (mode.equals(Mode.ECI)) {
91 // Count doesn't apply to ECI
92 int value = parseECIValue(bits);
93 currentCharacterSetECI = CharacterSetECI.getCharacterSetECIByValue(value);
94 if (currentCharacterSetECI == null) {
95 throw FormatException.getFormatInstance();
98 // How many characters will follow, encoded in this mode?
99 int count = bits.readBits(mode.getCharacterCountBits(version));
100 if (mode.equals(Mode.NUMERIC)) {
101 decodeNumericSegment(bits, result, count);
102 } else if (mode.equals(Mode.ALPHANUMERIC)) {
103 decodeAlphanumericSegment(bits, result, count, fc1InEffect);
104 } else if (mode.equals(Mode.BYTE)) {
105 decodeByteSegment(bits, result, count, currentCharacterSetECI, byteSegments, hints);
106 } else if (mode.equals(Mode.KANJI)) {
107 decodeKanjiSegment(bits, result, count);
109 throw FormatException.getFormatInstance();
113 } while (!mode.equals(Mode.TERMINATOR));
115 return new DecoderResult(bytes, result.toString(), byteSegments.isEmpty() ? null : byteSegments, ecLevel);
118 private static void decodeKanjiSegment(BitSource bits,
120 int count) throws FormatException {
121 // Each character will require 2 bytes. Read the characters as 2-byte pairs
122 // and decode as Shift_JIS afterwards
123 byte[] buffer = new byte[2 * count];
126 // Each 13 bits encodes a 2-byte character
127 int twoBytes = bits.readBits(13);
128 int assembledTwoBytes = ((twoBytes / 0x0C0) << 8) | (twoBytes % 0x0C0);
129 if (assembledTwoBytes < 0x01F00) {
130 // In the 0x8140 to 0x9FFC range
131 assembledTwoBytes += 0x08140;
133 // In the 0xE040 to 0xEBBF range
134 assembledTwoBytes += 0x0C140;
136 buffer[offset] = (byte) (assembledTwoBytes >> 8);
137 buffer[offset + 1] = (byte) assembledTwoBytes;
141 // Shift_JIS may not be supported in some environments:
143 result.append(new String(buffer, SHIFT_JIS));
144 } catch (UnsupportedEncodingException uee) {
145 throw FormatException.getFormatInstance();
149 private static void decodeByteSegment(BitSource bits,
152 CharacterSetECI currentCharacterSetECI,
154 Hashtable hints) throws FormatException {
155 byte[] readBytes = new byte[count];
156 if (count << 3 > bits.available()) {
157 throw FormatException.getFormatInstance();
159 for (int i = 0; i < count; i++) {
160 readBytes[i] = (byte) bits.readBits(8);
163 if (currentCharacterSetECI == null) {
164 // The spec isn't clear on this mode; see
165 // section 6.4.5: t does not say which encoding to assuming
166 // upon decoding. I have seen ISO-8859-1 used as well as
167 // Shift_JIS -- without anything like an ECI designator to
169 encoding = guessEncoding(readBytes, hints);
171 encoding = currentCharacterSetECI.getEncodingName();
174 result.append(new String(readBytes, encoding));
175 } catch (UnsupportedEncodingException uce) {
176 throw FormatException.getFormatInstance();
178 byteSegments.addElement(readBytes);
181 private static void decodeAlphanumericSegment(BitSource bits,
184 boolean fc1InEffect) {
185 // Read two characters at a time
186 int start = result.length();
188 int nextTwoCharsBits = bits.readBits(11);
189 result.append(ALPHANUMERIC_CHARS[nextTwoCharsBits / 45]);
190 result.append(ALPHANUMERIC_CHARS[nextTwoCharsBits % 45]);
194 // special case: one character left
195 result.append(ALPHANUMERIC_CHARS[bits.readBits(6)]);
197 // See section 6.4.8.1, 6.4.8.2
199 // We need to massage the result a bit if in an FNC1 mode:
200 for (int i = start; i < result.length(); i++) {
201 if (result.charAt(i) == '%') {
202 if (i < result.length() - 1 && result.charAt(i + 1) == '%') {
203 // %% is rendered as %
204 result.deleteCharAt(i + 1);
206 // In alpha mode, % should be converted to FNC1 separator 0x1D
207 result.setCharAt(i, (char) 0x1D);
214 private static void decodeNumericSegment(BitSource bits,
216 int count) throws FormatException {
217 // Read three digits at a time
219 // Each 10 bits encodes three digits
220 int threeDigitsBits = bits.readBits(10);
221 if (threeDigitsBits >= 1000) {
222 throw FormatException.getFormatInstance();
224 result.append(ALPHANUMERIC_CHARS[threeDigitsBits / 100]);
225 result.append(ALPHANUMERIC_CHARS[(threeDigitsBits / 10) % 10]);
226 result.append(ALPHANUMERIC_CHARS[threeDigitsBits % 10]);
230 // Two digits left over to read, encoded in 7 bits
231 int twoDigitsBits = bits.readBits(7);
232 if (twoDigitsBits >= 100) {
233 throw FormatException.getFormatInstance();
235 result.append(ALPHANUMERIC_CHARS[twoDigitsBits / 10]);
236 result.append(ALPHANUMERIC_CHARS[twoDigitsBits % 10]);
237 } else if (count == 1) {
238 // One digit left over to read
239 int digitBits = bits.readBits(4);
240 if (digitBits >= 10) {
241 throw FormatException.getFormatInstance();
243 result.append(ALPHANUMERIC_CHARS[digitBits]);
247 private static String guessEncoding(byte[] bytes, Hashtable hints) {
249 String characterSet = (String) hints.get(DecodeHintType.CHARACTER_SET);
250 if (characterSet != null) {
254 if (ASSUME_SHIFT_JIS) {
257 // Does it start with the UTF-8 byte order mark? then guess it's UTF-8
258 if (bytes.length > 3 && bytes[0] == (byte) 0xEF && bytes[1] == (byte) 0xBB && bytes[2] == (byte) 0xBF) {
261 // For now, merely tries to distinguish ISO-8859-1, UTF-8 and Shift_JIS,
262 // which should be by far the most common encodings. ISO-8859-1
263 // should not have bytes in the 0x80 - 0x9F range, while Shift_JIS
264 // uses this as a first byte of a two-byte character. If we see this
265 // followed by a valid second byte in Shift_JIS, assume it is Shift_JIS.
266 // If we see something else in that second byte, we'll make the risky guess
268 int length = bytes.length;
269 boolean canBeISO88591 = true;
270 boolean canBeShiftJIS = true;
271 int maybeDoubleByteCount = 0;
272 int maybeSingleByteKatakanaCount = 0;
273 boolean sawLatin1Supplement = false;
274 boolean lastWasPossibleDoubleByteStart = false;
275 for (int i = 0; i < length && (canBeISO88591 || canBeShiftJIS); i++) {
276 int value = bytes[i] & 0xFF;
277 if ((value == 0xC2 || value == 0xC3) && i < length - 1) {
278 // This is really a poor hack. The slightly more exotic characters people might want to put in
279 // a QR Code, by which I mean the Latin-1 supplement characters (e.g. u-umlaut) have encodings
280 // that start with 0xC2 followed by [0xA0,0xBF], or start with 0xC3 followed by [0x80,0xBF].
281 int nextValue = bytes[i + 1] & 0xFF;
282 if (nextValue <= 0xBF && ((value == 0xC2 && nextValue >= 0xA0) || (value == 0xC3 && nextValue >= 0x80))) {
283 sawLatin1Supplement = true;
286 if (value >= 0x7F && value <= 0x9F) {
287 canBeISO88591 = false;
289 if (value >= 0xA1 && value <= 0xDF) {
290 // count the number of characters that might be a Shift_JIS single-byte Katakana character
291 if (!lastWasPossibleDoubleByteStart) {
292 maybeSingleByteKatakanaCount++;
295 if (!lastWasPossibleDoubleByteStart && ((value >= 0xF0 && value <= 0xFF) || value == 0x80 || value == 0xA0)) {
296 canBeShiftJIS = false;
298 if (((value >= 0x81 && value <= 0x9F) || (value >= 0xE0 && value <= 0xEF))) {
299 // These start double-byte characters in Shift_JIS. Let's see if it's followed by a valid
301 if (lastWasPossibleDoubleByteStart) {
302 // If we just checked this and the last byte for being a valid double-byte
303 // char, don't check starting on this byte. If this and the last byte
304 // formed a valid pair, then this shouldn't be checked to see if it starts
305 // a double byte pair of course.
306 lastWasPossibleDoubleByteStart = false;
308 // ... otherwise do check to see if this plus the next byte form a valid
309 // double byte pair encoding a character.
310 lastWasPossibleDoubleByteStart = true;
311 if (i >= bytes.length - 1) {
312 canBeShiftJIS = false;
314 int nextValue = bytes[i + 1] & 0xFF;
315 if (nextValue < 0x40 || nextValue > 0xFC) {
316 canBeShiftJIS = false;
318 maybeDoubleByteCount++;
320 // There is some conflicting information out there about which bytes can follow which in
321 // double-byte Shift_JIS characters. The rule above seems to be the one that matches practice.
325 lastWasPossibleDoubleByteStart = false;
328 // Distinguishing Shift_JIS and ISO-8859-1 can be a little tough. The crude heuristic is:
330 // - at least three byte that starts a double-byte value (bytes that are rare in ISO-8859-1), or
331 // - over 5% of bytes that could be single-byte Katakana (also rare in ISO-8859-1),
332 // - and, saw no sequences that are invalid in Shift_JIS, then we conclude Shift_JIS
333 if (canBeShiftJIS && (maybeDoubleByteCount >= 3 || 20 * maybeSingleByteKatakanaCount > length)) {
336 // Otherwise, we default to ISO-8859-1 unless we know it can't be
337 if (!sawLatin1Supplement && canBeISO88591) {
340 // Otherwise, we take a wild guess with UTF-8
344 private static int parseECIValue(BitSource bits) {
345 int firstByte = bits.readBits(8);
346 if ((firstByte & 0x80) == 0) {
348 return firstByte & 0x7F;
349 } else if ((firstByte & 0xC0) == 0x80) {
351 int secondByte = bits.readBits(8);
352 return ((firstByte & 0x3F) << 8) | secondByte;
353 } else if ((firstByte & 0xE0) == 0xC0) {
355 int secondThirdBytes = bits.readBits(16);
356 return ((firstByte & 0x1F) << 16) | secondThirdBytes;
358 throw new IllegalArgumentException("Bad ECI bits starting with byte " + firstByte);