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.ReaderException;
20 import com.google.zxing.common.BitSource;
21 import com.google.zxing.common.CharacterSetECI;
22 import com.google.zxing.common.DecoderResult;
24 import java.io.UnsupportedEncodingException;
25 import java.util.Vector;
28 * <p>QR Codes can encode text as bits in one of several modes, and can use multiple modes
29 * in one QR Code. This class decodes the bits back into text.</p>
31 * <p>See ISO 18004:2006, 6.4.3 - 6.4.7</p>
35 final class DecodedBitStreamParser {
38 * See ISO 18004:2006, 6.4.4 Table 5
40 private static final char[] ALPHANUMERIC_CHARS = {
41 '0', '1', '2', '3', '4', '5', '6', '7', '8', '9', 'A', 'B',
42 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M', 'N',
43 'O', 'P', 'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X', 'Y', 'Z',
44 ' ', '$', '%', '*', '+', '-', '.', '/', ':'
46 private static final String SHIFT_JIS = "SJIS";
47 private static final String EUC_JP = "EUC_JP";
48 private static final boolean ASSUME_SHIFT_JIS;
49 private static final String UTF8 = "UTF8";
50 private static final String ISO88591 = "ISO8859_1";
53 String platformDefault = System.getProperty("file.encoding");
54 ASSUME_SHIFT_JIS = SHIFT_JIS.equalsIgnoreCase(platformDefault) || EUC_JP.equalsIgnoreCase(platformDefault);
57 private DecodedBitStreamParser() {
60 static DecoderResult decode(byte[] bytes, Version version) throws ReaderException {
61 BitSource bits = new BitSource(bytes);
62 StringBuffer result = new StringBuffer();
63 CharacterSetECI currentCharacterSetECI = null;
64 boolean fc1InEffect = false;
65 Vector byteSegments = new Vector(1);
68 // While still another segment to read...
69 if (bits.available() < 4) {
70 // OK, assume we're done. Really, a TERMINATOR mode should have been recorded here
71 mode = Mode.TERMINATOR;
74 mode = Mode.forBits(bits.readBits(4)); // mode is encoded by 4 bits
75 } catch (IllegalArgumentException iae) {
76 throw ReaderException.getInstance();
79 if (!mode.equals(Mode.TERMINATOR)) {
80 if (mode.equals(Mode.FNC1_FIRST_POSITION) || mode.equals(Mode.FNC1_SECOND_POSITION)) {
81 // We do little with FNC1 except alter the parsed result a bit according to the spec
83 } else if (mode.equals(Mode.STRUCTURED_APPEND)) {
84 // not really supported; all we do is ignore it
85 // Read next 8 bits (symbol sequence #) and 8 bits (parity data), then continue
87 } else if (mode.equals(Mode.ECI)) {
88 // Count doesn't apply to ECI
89 int value = parseECIValue(bits);
90 currentCharacterSetECI = CharacterSetECI.getCharacterSetECIByValue(value);
91 if (currentCharacterSetECI == null) {
92 throw ReaderException.getInstance();
95 // How many characters will follow, encoded in this mode?
96 int count = bits.readBits(mode.getCharacterCountBits(version));
97 if (mode.equals(Mode.NUMERIC)) {
98 decodeNumericSegment(bits, result, count);
99 } else if (mode.equals(Mode.ALPHANUMERIC)) {
100 decodeAlphanumericSegment(bits, result, count, fc1InEffect);
101 } else if (mode.equals(Mode.BYTE)) {
102 decodeByteSegment(bits, result, count, currentCharacterSetECI, byteSegments);
103 } else if (mode.equals(Mode.KANJI)) {
104 decodeKanjiSegment(bits, result, count);
106 throw ReaderException.getInstance();
110 } while (!mode.equals(Mode.TERMINATOR));
112 return new DecoderResult(bytes, result.toString(), byteSegments.isEmpty() ? null : byteSegments);
115 private static void decodeKanjiSegment(BitSource bits,
117 int count) throws ReaderException {
118 // Each character will require 2 bytes. Read the characters as 2-byte pairs
119 // and decode as Shift_JIS afterwards
120 byte[] buffer = new byte[2 * count];
123 // Each 13 bits encodes a 2-byte character
124 int twoBytes = bits.readBits(13);
125 int assembledTwoBytes = ((twoBytes / 0x0C0) << 8) | (twoBytes % 0x0C0);
126 if (assembledTwoBytes < 0x01F00) {
127 // In the 0x8140 to 0x9FFC range
128 assembledTwoBytes += 0x08140;
130 // In the 0xE040 to 0xEBBF range
131 assembledTwoBytes += 0x0C140;
133 buffer[offset] = (byte) (assembledTwoBytes >> 8);
134 buffer[offset + 1] = (byte) assembledTwoBytes;
138 // Shift_JIS may not be supported in some environments:
140 result.append(new String(buffer, SHIFT_JIS));
141 } catch (UnsupportedEncodingException uee) {
142 throw ReaderException.getInstance();
146 private static void decodeByteSegment(BitSource bits,
149 CharacterSetECI currentCharacterSetECI,
150 Vector byteSegments) throws ReaderException {
151 byte[] readBytes = new byte[count];
152 if (count << 3 > bits.available()) {
153 throw ReaderException.getInstance();
155 for (int i = 0; i < count; i++) {
156 readBytes[i] = (byte) bits.readBits(8);
159 if (currentCharacterSetECI == null) {
160 // The spec isn't clear on this mode; see
161 // section 6.4.5: t does not say which encoding to assuming
162 // upon decoding. I have seen ISO-8859-1 used as well as
163 // Shift_JIS -- without anything like an ECI designator to
165 encoding = guessEncoding(readBytes);
167 encoding = currentCharacterSetECI.getEncodingName();
170 result.append(new String(readBytes, encoding));
171 } catch (UnsupportedEncodingException uce) {
172 throw ReaderException.getInstance();
174 byteSegments.addElement(readBytes);
177 private static void decodeAlphanumericSegment(BitSource bits,
180 boolean fc1InEffect) {
181 // Read two characters at a time
182 int start = result.length();
184 int nextTwoCharsBits = bits.readBits(11);
185 result.append(ALPHANUMERIC_CHARS[nextTwoCharsBits / 45]);
186 result.append(ALPHANUMERIC_CHARS[nextTwoCharsBits % 45]);
190 // special case: one character left
191 result.append(ALPHANUMERIC_CHARS[bits.readBits(6)]);
193 // See section 6.4.8.1, 6.4.8.2
195 // We need to massage the result a bit if in an FNC1 mode:
196 for (int i = start; i < result.length(); i++) {
197 if (result.charAt(i) == '%') {
198 if (i < result.length() - 1 && result.charAt(i + 1) == '%') {
199 // %% is rendered as %
200 result.deleteCharAt(i + 1);
202 // In alpha mode, % should be converted to FNC1 separator 0x1D
203 result.setCharAt(i, (char) 0x1D);
210 private static void decodeNumericSegment(BitSource bits,
212 int count) throws ReaderException {
213 // Read three digits at a time
215 // Each 10 bits encodes three digits
216 int threeDigitsBits = bits.readBits(10);
217 if (threeDigitsBits >= 1000) {
218 throw ReaderException.getInstance();
220 result.append(ALPHANUMERIC_CHARS[threeDigitsBits / 100]);
221 result.append(ALPHANUMERIC_CHARS[(threeDigitsBits / 10) % 10]);
222 result.append(ALPHANUMERIC_CHARS[threeDigitsBits % 10]);
226 // Two digits left over to read, encoded in 7 bits
227 int twoDigitsBits = bits.readBits(7);
228 if (twoDigitsBits >= 100) {
229 throw ReaderException.getInstance();
231 result.append(ALPHANUMERIC_CHARS[twoDigitsBits / 10]);
232 result.append(ALPHANUMERIC_CHARS[twoDigitsBits % 10]);
233 } else if (count == 1) {
234 // One digit left over to read
235 int digitBits = bits.readBits(4);
236 if (digitBits >= 10) {
237 throw ReaderException.getInstance();
239 result.append(ALPHANUMERIC_CHARS[digitBits]);
243 private static String guessEncoding(byte[] bytes) {
244 if (ASSUME_SHIFT_JIS) {
247 // Does it start with the UTF-8 byte order mark? then guess it's UTF-8
248 if (bytes.length > 3 && bytes[0] == (byte) 0xEF && bytes[1] == (byte) 0xBB && bytes[2] == (byte) 0xBF) {
251 // For now, merely tries to distinguish ISO-8859-1, UTF-8 and Shift_JIS,
252 // which should be by far the most common encodings. ISO-8859-1
253 // should not have bytes in the 0x80 - 0x9F range, while Shift_JIS
254 // uses this as a first byte of a two-byte character. If we see this
255 // followed by a valid second byte in Shift_JIS, assume it is Shift_JIS.
256 // If we see something else in that second byte, we'll make the risky guess
258 int length = bytes.length;
259 boolean canBeISO88591 = true;
260 boolean canBeShiftJIS = true;
261 boolean sawDoubleByteStart = false;
262 int maybeSingleByteKatakanaCount = 0;
263 boolean sawLatin1Supplement = false;
264 boolean lastWasPossibleDoubleByteStart = false;
265 for (int i = 0; i < length && (canBeISO88591 || canBeShiftJIS); i++) {
266 int value = bytes[i] & 0xFF;
267 if (value == 0xC2 || value == 0xC3 && i < length - 1) {
268 // This is really a poor hack. The slightly more exotic characters people might want to put in
269 // a QR Code, by which I mean the Latin-1 supplement characters (e.g. u-umlaut) have encodings
270 // that start with 0xC2 followed by [0xA0,0xBF], or start with 0xC3 followed by [0x80,0xBF].
271 int nextValue = bytes[i + 1] & 0xFF;
272 if (nextValue <= 0xBF && ((value == 0xC2 && nextValue >= 0xA0) || (value == 0xC3 && nextValue >= 0x80))) {
273 sawLatin1Supplement = true;
276 if (value >= 0x7F && value <= 0x9F) {
277 canBeISO88591 = false;
279 if (value >= 0xA1 && value <= 0xDF) {
280 // count the number of characters that might be a Shift_JIS single-byte Katakana character
281 if (!lastWasPossibleDoubleByteStart) {
282 maybeSingleByteKatakanaCount++;
285 if (!lastWasPossibleDoubleByteStart && ((value >= 0xF0 && value <= 0xFF) || value == 0x80 || value == 0xA0)) {
286 canBeShiftJIS = false;
288 if (((value >= 0x81 && value <= 0x9F) || (value >= 0xE0 && value <= 0xEF)) && i < length - 1) {
289 // These start double-byte characters in Shift_JIS. Let's see if it's followed by a valid
291 sawDoubleByteStart = true;
292 if (lastWasPossibleDoubleByteStart) {
293 // If we just checked this and the last byte for being a valid double-byte
294 // char, don't check starting on this byte. If this and the last byte
295 // formed a valid pair, then this shouldn't be checked to see if it starts
296 // a double byte pair of course.
297 lastWasPossibleDoubleByteStart = false;
299 // ... otherwise do check to see if this plus the next byte form a valid
300 // double byte pair encoding a character.
301 lastWasPossibleDoubleByteStart = true;
302 int nextValue = bytes[i + 1] & 0xFF;
303 if (nextValue < 0x40 || nextValue > 0xFC) {
304 canBeShiftJIS = false;
306 // There is some conflicting information out there about which bytes can follow which in
307 // double-byte Shift_JIS characters. The rule above seems to be the one that matches practice.
310 lastWasPossibleDoubleByteStart = false;
313 // Distinguishing Shift_JIS and ISO-8859-1 can be a little tough. The crude heuristic is:
315 // - at least one byte that starts a double-byte value (bytes that are rare in ISO-8859-1), or
316 // - over 5% of bytes that could be single-byte Katakana (also rare in ISO-8859-1),
317 // - and, saw no sequences that are invalid in Shift_JIS, then we conclude Shift_JIS
318 if (canBeShiftJIS && (sawDoubleByteStart || 20 * maybeSingleByteKatakanaCount > length)) {
321 // Otherwise, we default to ISO-8859-1 unless we know it can't be
322 if (!sawLatin1Supplement && canBeISO88591) {
325 // Otherwise, we take a wild guess with UTF-8
329 private static int parseECIValue(BitSource bits) {
330 int firstByte = bits.readBits(8);
331 if ((firstByte & 0x80) == 0) {
333 return firstByte & 0x7F;
334 } else if ((firstByte & 0xC0) == 0x80) {
336 int secondByte = bits.readBits(8);
337 return ((firstByte & 0x3F) << 8) | secondByte;
338 } else if ((firstByte & 0xE0) == 0xC0) {
340 int secondThirdBytes = bits.readBits(16);
341 return ((firstByte & 0x1F) << 16) | secondThirdBytes;
343 throw new IllegalArgumentException("Bad ECI bits starting with byte " + firstByte);