@zxing/library
Version:
TypeScript port of ZXing multi-format 1D/2D barcode image processing library.
548 lines (547 loc) • 24.4 kB
JavaScript
/*
* Copyright 2008 ZXing authors
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
/*namespace com.google.zxing.qrcode.encoder {*/
import EncodeHintType from '../../EncodeHintType';
import BitArray from '../../common/BitArray';
import CharacterSetECI from '../../common/CharacterSetECI';
import GenericGF from '../../common/reedsolomon/GenericGF';
import ReedSolomonEncoder from '../../common/reedsolomon/ReedSolomonEncoder';
import Mode from '../decoder/Mode';
import Version from '../decoder/Version';
import MaskUtil from './MaskUtil';
import ByteMatrix from './ByteMatrix';
import QRCode from './QRCode';
import MatrixUtil from './MatrixUtil';
import StringEncoding from '../../util/StringEncoding';
import BlockPair from './BlockPair';
import WriterException from '../../WriterException';
/*import java.io.UnsupportedEncodingException;*/
/*import java.util.ArrayList;*/
/*import java.util.Collection;*/
/*import java.util.Map;*/
/**
* @author satorux@google.com (Satoru Takabayashi) - creator
* @author dswitkin@google.com (Daniel Switkin) - ported from C++
*/
export default class Encoder {
// TYPESCRIPTPORT: changed to UTF8, the default for js
constructor() { }
// The mask penalty calculation is complicated. See Table 21 of JISX0510:2004 (p.45) for details.
// Basically it applies four rules and summate all penalties.
static calculateMaskPenalty(matrix) {
return MaskUtil.applyMaskPenaltyRule1(matrix)
+ MaskUtil.applyMaskPenaltyRule2(matrix)
+ MaskUtil.applyMaskPenaltyRule3(matrix)
+ MaskUtil.applyMaskPenaltyRule4(matrix);
}
/**
* @param content text to encode
* @param ecLevel error correction level to use
* @return {@link QRCode} representing the encoded QR code
* @throws WriterException if encoding can't succeed, because of for example invalid content
* or configuration
*/
// public static encode(content: string, ecLevel: ErrorCorrectionLevel): QRCode /*throws WriterException*/ {
// return encode(content, ecLevel, null)
// }
static encode(content, ecLevel, hints = null) {
// Determine what character encoding has been specified by the caller, if any
let encoding = Encoder.DEFAULT_BYTE_MODE_ENCODING;
const hasEncodingHint = hints !== null && undefined !== hints.get(EncodeHintType.CHARACTER_SET);
if (hasEncodingHint) {
encoding = hints.get(EncodeHintType.CHARACTER_SET).toString();
}
// Pick an encoding mode appropriate for the content. Note that this will not attempt to use
// multiple modes / segments even if that were more efficient. Twould be nice.
const mode = this.chooseMode(content, encoding);
// This will store the header information, like mode and
// length, as well as "header" segments like an ECI segment.
const headerBits = new BitArray();
// Append ECI segment if applicable
if (mode === Mode.BYTE && (hasEncodingHint || Encoder.DEFAULT_BYTE_MODE_ENCODING !== encoding)) {
const eci = CharacterSetECI.getCharacterSetECIByName(encoding);
if (eci !== undefined) {
this.appendECI(eci, headerBits);
}
}
// (With ECI in place,) Write the mode marker
this.appendModeInfo(mode, headerBits);
// Collect data within the main segment, separately, to count its size if needed. Don't add it to
// main payload yet.
const dataBits = new BitArray();
this.appendBytes(content, mode, dataBits, encoding);
let version;
if (hints !== null && undefined !== hints.get(EncodeHintType.QR_VERSION)) {
const versionNumber = Number.parseInt(hints.get(EncodeHintType.QR_VERSION).toString(), 10);
version = Version.getVersionForNumber(versionNumber);
const bitsNeeded = this.calculateBitsNeeded(mode, headerBits, dataBits, version);
if (!this.willFit(bitsNeeded, version, ecLevel)) {
throw new WriterException('Data too big for requested version');
}
}
else {
version = this.recommendVersion(ecLevel, mode, headerBits, dataBits);
}
const headerAndDataBits = new BitArray();
headerAndDataBits.appendBitArray(headerBits);
// Find "length" of main segment and write it
const numLetters = mode === Mode.BYTE ? dataBits.getSizeInBytes() : content.length;
this.appendLengthInfo(numLetters, version, mode, headerAndDataBits);
// Put data together into the overall payload
headerAndDataBits.appendBitArray(dataBits);
const ecBlocks = version.getECBlocksForLevel(ecLevel);
const numDataBytes = version.getTotalCodewords() - ecBlocks.getTotalECCodewords();
// Terminate the bits properly.
this.terminateBits(numDataBytes, headerAndDataBits);
// Interleave data bits with error correction code.
const finalBits = this.interleaveWithECBytes(headerAndDataBits, version.getTotalCodewords(), numDataBytes, ecBlocks.getNumBlocks());
const qrCode = new QRCode();
qrCode.setECLevel(ecLevel);
qrCode.setMode(mode);
qrCode.setVersion(version);
// Choose the mask pattern and set to "qrCode".
const dimension = version.getDimensionForVersion();
const matrix = new ByteMatrix(dimension, dimension);
const maskPattern = this.chooseMaskPattern(finalBits, ecLevel, version, matrix);
qrCode.setMaskPattern(maskPattern);
// Build the matrix and set it to "qrCode".
MatrixUtil.buildMatrix(finalBits, ecLevel, version, maskPattern, matrix);
qrCode.setMatrix(matrix);
return qrCode;
}
/**
* Decides the smallest version of QR code that will contain all of the provided data.
*
* @throws WriterException if the data cannot fit in any version
*/
static recommendVersion(ecLevel, mode, headerBits, dataBits) {
// Hard part: need to know version to know how many bits length takes. But need to know how many
// bits it takes to know version. First we take a guess at version by assuming version will be
// the minimum, 1:
const provisionalBitsNeeded = this.calculateBitsNeeded(mode, headerBits, dataBits, Version.getVersionForNumber(1));
const provisionalVersion = this.chooseVersion(provisionalBitsNeeded, ecLevel);
// Use that guess to calculate the right version. I am still not sure this works in 100% of cases.
const bitsNeeded = this.calculateBitsNeeded(mode, headerBits, dataBits, provisionalVersion);
return this.chooseVersion(bitsNeeded, ecLevel);
}
static calculateBitsNeeded(mode, headerBits, dataBits, version) {
return headerBits.getSize() + mode.getCharacterCountBits(version) + dataBits.getSize();
}
/**
* @return the code point of the table used in alphanumeric mode or
* -1 if there is no corresponding code in the table.
*/
static getAlphanumericCode(code /*int*/) {
if (code < Encoder.ALPHANUMERIC_TABLE.length) {
return Encoder.ALPHANUMERIC_TABLE[code];
}
return -1;
}
// public static chooseMode(content: string): Mode {
// return chooseMode(content, null);
// }
/**
* Choose the best mode by examining the content. Note that 'encoding' is used as a hint;
* if it is Shift_JIS, and the input is only double-byte Kanji, then we return {@link Mode#KANJI}.
*/
static chooseMode(content, encoding = null) {
if (CharacterSetECI.SJIS.getName() === encoding && this.isOnlyDoubleByteKanji(content)) {
// Choose Kanji mode if all input are double-byte characters
return Mode.KANJI;
}
let hasNumeric = false;
let hasAlphanumeric = false;
for (let i = 0, length = content.length; i < length; ++i) {
const c = content.charAt(i);
if (Encoder.isDigit(c)) {
hasNumeric = true;
}
else if (this.getAlphanumericCode(c.charCodeAt(0)) !== -1) {
hasAlphanumeric = true;
}
else {
return Mode.BYTE;
}
}
if (hasAlphanumeric) {
return Mode.ALPHANUMERIC;
}
if (hasNumeric) {
return Mode.NUMERIC;
}
return Mode.BYTE;
}
static isOnlyDoubleByteKanji(content) {
let bytes;
try {
bytes = StringEncoding.encode(content, CharacterSetECI.SJIS); // content.getBytes("Shift_JIS"))
}
catch (ignored /*: UnsupportedEncodingException*/) {
return false;
}
const length = bytes.length;
if (length % 2 !== 0) {
return false;
}
for (let i = 0; i < length; i += 2) {
const byte1 = bytes[i] & 0xFF;
if ((byte1 < 0x81 || byte1 > 0x9F) && (byte1 < 0xE0 || byte1 > 0xEB)) {
return false;
}
}
return true;
}
static chooseMaskPattern(bits, ecLevel, version, matrix) {
let minPenalty = Number.MAX_SAFE_INTEGER; // Lower penalty is better.
let bestMaskPattern = -1;
// We try all mask patterns to choose the best one.
for (let maskPattern = 0; maskPattern < QRCode.NUM_MASK_PATTERNS; maskPattern++) {
MatrixUtil.buildMatrix(bits, ecLevel, version, maskPattern, matrix);
let penalty = this.calculateMaskPenalty(matrix);
if (penalty < minPenalty) {
minPenalty = penalty;
bestMaskPattern = maskPattern;
}
}
return bestMaskPattern;
}
static chooseVersion(numInputBits /*int*/, ecLevel) {
for (let versionNum = 1; versionNum <= 40; versionNum++) {
const version = Version.getVersionForNumber(versionNum);
if (Encoder.willFit(numInputBits, version, ecLevel)) {
return version;
}
}
throw new WriterException('Data too big');
}
/**
* @return true if the number of input bits will fit in a code with the specified version and
* error correction level.
*/
static willFit(numInputBits /*int*/, version, ecLevel) {
// In the following comments, we use numbers of Version 7-H.
// numBytes = 196
const numBytes = version.getTotalCodewords();
// getNumECBytes = 130
const ecBlocks = version.getECBlocksForLevel(ecLevel);
const numEcBytes = ecBlocks.getTotalECCodewords();
// getNumDataBytes = 196 - 130 = 66
const numDataBytes = numBytes - numEcBytes;
const totalInputBytes = (numInputBits + 7) / 8;
return numDataBytes >= totalInputBytes;
}
/**
* Terminate bits as described in 8.4.8 and 8.4.9 of JISX0510:2004 (p.24).
*/
static terminateBits(numDataBytes /*int*/, bits) {
const capacity = numDataBytes * 8;
if (bits.getSize() > capacity) {
throw new WriterException('data bits cannot fit in the QR Code' + bits.getSize() + ' > ' +
capacity);
}
for (let i = 0; i < 4 && bits.getSize() < capacity; ++i) {
bits.appendBit(false);
}
// Append termination bits. See 8.4.8 of JISX0510:2004 (p.24) for details.
// If the last byte isn't 8-bit aligned, we'll add padding bits.
const numBitsInLastByte = bits.getSize() & 0x07;
if (numBitsInLastByte > 0) {
for (let i = numBitsInLastByte; i < 8; i++) {
bits.appendBit(false);
}
}
// If we have more space, we'll fill the space with padding patterns defined in 8.4.9 (p.24).
const numPaddingBytes = numDataBytes - bits.getSizeInBytes();
for (let i = 0; i < numPaddingBytes; ++i) {
bits.appendBits((i & 0x01) === 0 ? 0xEC : 0x11, 8);
}
if (bits.getSize() !== capacity) {
throw new WriterException('Bits size does not equal capacity');
}
}
/**
* Get number of data bytes and number of error correction bytes for block id "blockID". Store
* the result in "numDataBytesInBlock", and "numECBytesInBlock". See table 12 in 8.5.1 of
* JISX0510:2004 (p.30)
*/
static getNumDataBytesAndNumECBytesForBlockID(numTotalBytes /*int*/, numDataBytes /*int*/, numRSBlocks /*int*/, blockID /*int*/, numDataBytesInBlock, numECBytesInBlock) {
if (blockID >= numRSBlocks) {
throw new WriterException('Block ID too large');
}
// numRsBlocksInGroup2 = 196 % 5 = 1
const numRsBlocksInGroup2 = numTotalBytes % numRSBlocks;
// numRsBlocksInGroup1 = 5 - 1 = 4
const numRsBlocksInGroup1 = numRSBlocks - numRsBlocksInGroup2;
// numTotalBytesInGroup1 = 196 / 5 = 39
const numTotalBytesInGroup1 = Math.floor(numTotalBytes / numRSBlocks);
// numTotalBytesInGroup2 = 39 + 1 = 40
const numTotalBytesInGroup2 = numTotalBytesInGroup1 + 1;
// numDataBytesInGroup1 = 66 / 5 = 13
const numDataBytesInGroup1 = Math.floor(numDataBytes / numRSBlocks);
// numDataBytesInGroup2 = 13 + 1 = 14
const numDataBytesInGroup2 = numDataBytesInGroup1 + 1;
// numEcBytesInGroup1 = 39 - 13 = 26
const numEcBytesInGroup1 = numTotalBytesInGroup1 - numDataBytesInGroup1;
// numEcBytesInGroup2 = 40 - 14 = 26
const numEcBytesInGroup2 = numTotalBytesInGroup2 - numDataBytesInGroup2;
// Sanity checks.
// 26 = 26
if (numEcBytesInGroup1 !== numEcBytesInGroup2) {
throw new WriterException('EC bytes mismatch');
}
// 5 = 4 + 1.
if (numRSBlocks !== numRsBlocksInGroup1 + numRsBlocksInGroup2) {
throw new WriterException('RS blocks mismatch');
}
// 196 = (13 + 26) * 4 + (14 + 26) * 1
if (numTotalBytes !==
((numDataBytesInGroup1 + numEcBytesInGroup1) *
numRsBlocksInGroup1) +
((numDataBytesInGroup2 + numEcBytesInGroup2) *
numRsBlocksInGroup2)) {
throw new WriterException('Total bytes mismatch');
}
if (blockID < numRsBlocksInGroup1) {
numDataBytesInBlock[0] = numDataBytesInGroup1;
numECBytesInBlock[0] = numEcBytesInGroup1;
}
else {
numDataBytesInBlock[0] = numDataBytesInGroup2;
numECBytesInBlock[0] = numEcBytesInGroup2;
}
}
/**
* Interleave "bits" with corresponding error correction bytes. On success, store the result in
* "result". The interleave rule is complicated. See 8.6 of JISX0510:2004 (p.37) for details.
*/
static interleaveWithECBytes(bits, numTotalBytes /*int*/, numDataBytes /*int*/, numRSBlocks /*int*/) {
// "bits" must have "getNumDataBytes" bytes of data.
if (bits.getSizeInBytes() !== numDataBytes) {
throw new WriterException('Number of bits and data bytes does not match');
}
// Step 1. Divide data bytes into blocks and generate error correction bytes for them. We'll
// store the divided data bytes blocks and error correction bytes blocks into "blocks".
let dataBytesOffset = 0;
let maxNumDataBytes = 0;
let maxNumEcBytes = 0;
// Since, we know the number of reedsolmon blocks, we can initialize the vector with the number.
const blocks = new Array(); // new Array<BlockPair>(numRSBlocks)
for (let i = 0; i < numRSBlocks; ++i) {
const numDataBytesInBlock = new Int32Array(1);
const numEcBytesInBlock = new Int32Array(1);
Encoder.getNumDataBytesAndNumECBytesForBlockID(numTotalBytes, numDataBytes, numRSBlocks, i, numDataBytesInBlock, numEcBytesInBlock);
const size = numDataBytesInBlock[0];
const dataBytes = new Uint8Array(size);
bits.toBytes(8 * dataBytesOffset, dataBytes, 0, size);
const ecBytes = Encoder.generateECBytes(dataBytes, numEcBytesInBlock[0]);
blocks.push(new BlockPair(dataBytes, ecBytes));
maxNumDataBytes = Math.max(maxNumDataBytes, size);
maxNumEcBytes = Math.max(maxNumEcBytes, ecBytes.length);
dataBytesOffset += numDataBytesInBlock[0];
}
if (numDataBytes !== dataBytesOffset) {
throw new WriterException('Data bytes does not match offset');
}
const result = new BitArray();
// First, place data blocks.
for (let i = 0; i < maxNumDataBytes; ++i) {
for (const block of blocks) {
const dataBytes = block.getDataBytes();
if (i < dataBytes.length) {
result.appendBits(dataBytes[i], 8);
}
}
}
// Then, place error correction blocks.
for (let i = 0; i < maxNumEcBytes; ++i) {
for (const block of blocks) {
const ecBytes = block.getErrorCorrectionBytes();
if (i < ecBytes.length) {
result.appendBits(ecBytes[i], 8);
}
}
}
if (numTotalBytes !== result.getSizeInBytes()) { // Should be same.
throw new WriterException('Interleaving error: ' + numTotalBytes + ' and ' +
result.getSizeInBytes() + ' differ.');
}
return result;
}
static generateECBytes(dataBytes, numEcBytesInBlock /*int*/) {
const numDataBytes = dataBytes.length;
const toEncode = new Int32Array(numDataBytes + numEcBytesInBlock); // int[numDataBytes + numEcBytesInBlock]
for (let i = 0; i < numDataBytes; i++) {
toEncode[i] = dataBytes[i] & 0xFF;
}
new ReedSolomonEncoder(GenericGF.QR_CODE_FIELD_256).encode(toEncode, numEcBytesInBlock);
const ecBytes = new Uint8Array(numEcBytesInBlock);
for (let i = 0; i < numEcBytesInBlock; i++) {
ecBytes[i] = /*(byte) */ toEncode[numDataBytes + i];
}
return ecBytes;
}
/**
* Append mode info. On success, store the result in "bits".
*/
static appendModeInfo(mode, bits) {
bits.appendBits(mode.getBits(), 4);
}
/**
* Append length info. On success, store the result in "bits".
*/
static appendLengthInfo(numLetters /*int*/, version, mode, bits) {
const numBits = mode.getCharacterCountBits(version);
if (numLetters >= (1 << numBits)) {
throw new WriterException(numLetters + ' is bigger than ' + ((1 << numBits) - 1));
}
bits.appendBits(numLetters, numBits);
}
/**
* Append "bytes" in "mode" mode (encoding) into "bits". On success, store the result in "bits".
*/
static appendBytes(content, mode, bits, encoding) {
switch (mode) {
case Mode.NUMERIC:
Encoder.appendNumericBytes(content, bits);
break;
case Mode.ALPHANUMERIC:
Encoder.appendAlphanumericBytes(content, bits);
break;
case Mode.BYTE:
Encoder.append8BitBytes(content, bits, encoding);
break;
case Mode.KANJI:
Encoder.appendKanjiBytes(content, bits);
break;
default:
throw new WriterException('Invalid mode: ' + mode);
}
}
static getDigit(singleCharacter) {
return singleCharacter.charCodeAt(0) - 48;
}
static isDigit(singleCharacter) {
const cn = Encoder.getDigit(singleCharacter);
return cn >= 0 && cn <= 9;
}
static appendNumericBytes(content, bits) {
const length = content.length;
let i = 0;
while (i < length) {
const num1 = Encoder.getDigit(content.charAt(i));
if (i + 2 < length) {
// Encode three numeric letters in ten bits.
const num2 = Encoder.getDigit(content.charAt(i + 1));
const num3 = Encoder.getDigit(content.charAt(i + 2));
bits.appendBits(num1 * 100 + num2 * 10 + num3, 10);
i += 3;
}
else if (i + 1 < length) {
// Encode two numeric letters in seven bits.
const num2 = Encoder.getDigit(content.charAt(i + 1));
bits.appendBits(num1 * 10 + num2, 7);
i += 2;
}
else {
// Encode one numeric letter in four bits.
bits.appendBits(num1, 4);
i++;
}
}
}
static appendAlphanumericBytes(content, bits) {
const length = content.length;
let i = 0;
while (i < length) {
const code1 = Encoder.getAlphanumericCode(content.charCodeAt(i));
if (code1 === -1) {
throw new WriterException();
}
if (i + 1 < length) {
const code2 = Encoder.getAlphanumericCode(content.charCodeAt(i + 1));
if (code2 === -1) {
throw new WriterException();
}
// Encode two alphanumeric letters in 11 bits.
bits.appendBits(code1 * 45 + code2, 11);
i += 2;
}
else {
// Encode one alphanumeric letter in six bits.
bits.appendBits(code1, 6);
i++;
}
}
}
static append8BitBytes(content, bits, encoding) {
let bytes;
try {
bytes = StringEncoding.encode(content, encoding);
}
catch (uee /*: UnsupportedEncodingException*/) {
throw new WriterException(uee);
}
for (let i = 0, length = bytes.length; i !== length; i++) {
const b = bytes[i];
bits.appendBits(b, 8);
}
}
/**
* @throws WriterException
*/
static appendKanjiBytes(content, bits) {
let bytes;
try {
bytes = StringEncoding.encode(content, CharacterSetECI.SJIS);
}
catch (uee /*: UnsupportedEncodingException*/) {
throw new WriterException(uee);
}
const length = bytes.length;
for (let i = 0; i < length; i += 2) {
const byte1 = bytes[i] & 0xFF;
const byte2 = bytes[i + 1] & 0xFF;
const code = ((byte1 << 8) & 0xFFFFFFFF) | byte2;
let subtracted = -1;
if (code >= 0x8140 && code <= 0x9ffc) {
subtracted = code - 0x8140;
}
else if (code >= 0xe040 && code <= 0xebbf) {
subtracted = code - 0xc140;
}
if (subtracted === -1) {
throw new WriterException('Invalid byte sequence');
}
const encoded = ((subtracted >> 8) * 0xc0) + (subtracted & 0xff);
bits.appendBits(encoded, 13);
}
}
static appendECI(eci, bits) {
bits.appendBits(Mode.ECI.getBits(), 4);
// This is correct for values up to 127, which is all we need now.
bits.appendBits(eci.getValue(), 8);
}
}
// The original table is defined in the table 5 of JISX0510:2004 (p.19).
Encoder.ALPHANUMERIC_TABLE = Int32Array.from([
-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
36, -1, -1, -1, 37, 38, -1, -1, -1, -1, 39, 40, -1, 41, 42, 43,
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 44, -1, -1, -1, -1, -1,
-1, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24,
25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, -1, -1, -1, -1, -1,
]);
Encoder.DEFAULT_BYTE_MODE_ENCODING = CharacterSetECI.UTF8.getName(); // "ISO-8859-1"