123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564 |
-
- // Multibyte codec. In this scheme, a character is represented by 1 or more bytes.
- // Our codec supports UTF-16 surrogates, extensions for GB18030 and unicode sequences.
- // To save memory and loading time, we read table files only when requested.
-
- exports._dbcs = function(options) {
- return new DBCSCodec(options);
- }
-
- var UNASSIGNED = -1,
- GB18030_CODE = -2,
- SEQ_START = -10,
- NODE_START = -1000,
- UNASSIGNED_NODE = new Array(0x100),
- DEF_CHAR = -1;
-
- for (var i = 0; i < 0x100; i++)
- UNASSIGNED_NODE[i] = UNASSIGNED;
-
-
- // Class DBCSCodec reads and initializes mapping tables.
- function DBCSCodec(options) {
- this.options = options;
- if (!options)
- throw new Error("DBCS codec is called without the data.")
- if (!options.table)
- throw new Error("Encoding '" + options.encodingName + "' has no data.");
-
- // Load tables.
- var mappingTable = options.table();
-
-
- // Decode tables: MBCS -> Unicode.
-
- // decodeTables is a trie, encoded as an array of arrays of integers. Internal arrays are trie nodes and all have len = 256.
- // Trie root is decodeTables[0].
- // Values: >= 0 -> unicode character code. can be > 0xFFFF
- // == UNASSIGNED -> unknown/unassigned sequence.
- // == GB18030_CODE -> this is the end of a GB18030 4-byte sequence.
- // <= NODE_START -> index of the next node in our trie to process next byte.
- // <= SEQ_START -> index of the start of a character code sequence, in decodeTableSeq.
- this.decodeTables = [];
- this.decodeTables[0] = UNASSIGNED_NODE.slice(0); // Create root node.
-
- // Sometimes a MBCS char corresponds to a sequence of unicode chars. We store them as arrays of integers here.
- this.decodeTableSeq = [];
-
- // Actual mapping tables consist of chunks. Use them to fill up decode tables.
- for (var i = 0; i < mappingTable.length; i++)
- this._addDecodeChunk(mappingTable[i]);
-
- this.defaultCharUnicode = options.iconv.defaultCharUnicode;
-
-
- // Encode tables: Unicode -> DBCS.
-
- // `encodeTable` is array mapping from unicode char to encoded char. All its values are integers for performance.
- // Because it can be sparse, it is represented as array of buckets by 256 chars each. Bucket can be null.
- // Values: >= 0 -> it is a normal char. Write the value (if <=256 then 1 byte, if <=65536 then 2 bytes, etc.).
- // == UNASSIGNED -> no conversion found. Output a default char.
- // <= SEQ_START -> it's an index in encodeTableSeq, see below. The character starts a sequence.
- this.encodeTable = [];
-
- // `encodeTableSeq` is used when a sequence of unicode characters is encoded as a single code. We use a tree of
- // objects where keys correspond to characters in sequence and leafs are the encoded dbcs values. A special DEF_CHAR key
- // means end of sequence (needed when one sequence is a strict subsequence of another).
- // Objects are kept separately from encodeTable to increase performance.
- this.encodeTableSeq = [];
-
- // Some chars can be decoded, but need not be encoded.
- var skipEncodeChars = {};
- if (options.encodeSkipVals)
- for (var i = 0; i < options.encodeSkipVals.length; i++) {
- var range = options.encodeSkipVals[i];
- for (var j = range.from; j <= range.to; j++)
- skipEncodeChars[j] = true;
- }
-
- // Use decode trie to recursively fill out encode tables.
- this._fillEncodeTable(0, 0, skipEncodeChars);
-
- // Add more encoding pairs when needed.
- for (var uChar in options.encodeAdd || {})
- this._setEncodeChar(uChar.charCodeAt(0), options.encodeAdd[uChar]);
-
- this.defCharSB = this.encodeTable[0][options.iconv.defaultCharSingleByte.charCodeAt(0)];
- if (this.defCharSB === UNASSIGNED) this.defCharSB = this.encodeTable[0]['?'];
- if (this.defCharSB === UNASSIGNED) this.defCharSB = "?".charCodeAt(0);
-
-
- // Load & create GB18030 tables when needed.
- if (typeof options.gb18030 === 'function') {
- this.gb18030 = options.gb18030(); // Load GB18030 ranges.
-
- // Add GB18030 decode tables.
- var thirdByteNodeIdx = this.decodeTables.length;
- var thirdByteNode = this.decodeTables[thirdByteNodeIdx] = UNASSIGNED_NODE.slice(0);
-
- var fourthByteNodeIdx = this.decodeTables.length;
- var fourthByteNode = this.decodeTables[fourthByteNodeIdx] = UNASSIGNED_NODE.slice(0);
-
- for (var i = 0x81; i <= 0xFE; i++) {
- var secondByteNodeIdx = NODE_START - this.decodeTables[0][i];
- var secondByteNode = this.decodeTables[secondByteNodeIdx];
- for (var j = 0x30; j <= 0x39; j++)
- secondByteNode[j] = NODE_START - thirdByteNodeIdx;
- }
- for (var i = 0x81; i <= 0xFE; i++)
- thirdByteNode[i] = NODE_START - fourthByteNodeIdx;
- for (var i = 0x30; i <= 0x39; i++)
- fourthByteNode[i] = GB18030_CODE
- }
- }
-
- // Public interface: create encoder and decoder objects.
- // The methods (write, end) are simple functions to not inhibit optimizations.
- DBCSCodec.prototype.encoder = function encoderDBCS(options) {
- return {
- // Methods
- write: encoderDBCSWrite,
- end: encoderDBCSEnd,
-
- // Encoder state
- leadSurrogate: -1,
- seqObj: undefined,
-
- // Static data
- encodeTable: this.encodeTable,
- encodeTableSeq: this.encodeTableSeq,
- defaultCharSingleByte: this.defCharSB,
- gb18030: this.gb18030,
-
- // Export for testing
- findIdx: findIdx,
- }
- }
-
- DBCSCodec.prototype.decoder = function decoderDBCS(options) {
- return {
- // Methods
- write: decoderDBCSWrite,
- end: decoderDBCSEnd,
-
- // Decoder state
- nodeIdx: 0,
- prevBuf: new Buffer(0),
-
- // Static data
- decodeTables: this.decodeTables,
- decodeTableSeq: this.decodeTableSeq,
- defaultCharUnicode: this.defaultCharUnicode,
- gb18030: this.gb18030,
- }
- }
-
-
-
- // Decoder helpers
- DBCSCodec.prototype._getDecodeTrieNode = function(addr) {
- var bytes = [];
- for (; addr > 0; addr >>= 8)
- bytes.push(addr & 0xFF);
- if (bytes.length == 0)
- bytes.push(0);
-
- var node = this.decodeTables[0];
- for (var i = bytes.length-1; i > 0; i--) { // Traverse nodes deeper into the trie.
- var val = node[bytes[i]];
-
- if (val == UNASSIGNED) { // Create new node.
- node[bytes[i]] = NODE_START - this.decodeTables.length;
- this.decodeTables.push(node = UNASSIGNED_NODE.slice(0));
- }
- else if (val <= NODE_START) { // Existing node.
- node = this.decodeTables[NODE_START - val];
- }
- else
- throw new Error("Overwrite byte in " + this.options.encodingName + ", addr: " + addr.toString(16));
- }
- return node;
- }
-
-
- DBCSCodec.prototype._addDecodeChunk = function(chunk) {
- // First element of chunk is the hex mbcs code where we start.
- var curAddr = parseInt(chunk[0], 16);
-
- // Choose the decoding node where we'll write our chars.
- var writeTable = this._getDecodeTrieNode(curAddr);
- curAddr = curAddr & 0xFF;
-
- // Write all other elements of the chunk to the table.
- for (var k = 1; k < chunk.length; k++) {
- var part = chunk[k];
- if (typeof part === "string") { // String, write as-is.
- for (var l = 0; l < part.length;) {
- var code = part.charCodeAt(l++);
- if (0xD800 <= code && code < 0xDC00) { // Decode surrogate
- var codeTrail = part.charCodeAt(l++);
- if (0xDC00 <= codeTrail && codeTrail < 0xE000)
- writeTable[curAddr++] = 0x10000 + (code - 0xD800) * 0x400 + (codeTrail - 0xDC00);
- else
- throw new Error("Incorrect surrogate pair in " + this.options.encodingName + " at chunk " + chunk[0]);
- }
- else if (0x0FF0 < code && code <= 0x0FFF) { // Character sequence (our own encoding used)
- var len = 0xFFF - code + 2;
- var seq = [];
- for (var m = 0; m < len; m++)
- seq.push(part.charCodeAt(l++)); // Simple variation: don't support surrogates or subsequences in seq.
-
- writeTable[curAddr++] = SEQ_START - this.decodeTableSeq.length;
- this.decodeTableSeq.push(seq);
- }
- else
- writeTable[curAddr++] = code; // Basic char
- }
- }
- else if (typeof part === "number") { // Integer, meaning increasing sequence starting with prev character.
- var charCode = writeTable[curAddr - 1] + 1;
- for (var l = 0; l < part; l++)
- writeTable[curAddr++] = charCode++;
- }
- else
- throw new Error("Incorrect type '" + typeof part + "' given in " + this.options.encodingName + " at chunk " + chunk[0]);
- }
- if (curAddr > 0xFF)
- throw new Error("Incorrect chunk in " + this.options.encodingName + " at addr " + chunk[0] + ": too long" + curAddr);
- }
-
- // Encoder helpers
- DBCSCodec.prototype._getEncodeBucket = function(uCode) {
- var high = uCode >> 8; // This could be > 0xFF because of astral characters.
- if (this.encodeTable[high] === undefined)
- this.encodeTable[high] = UNASSIGNED_NODE.slice(0); // Create bucket on demand.
- return this.encodeTable[high];
- }
-
- DBCSCodec.prototype._setEncodeChar = function(uCode, dbcsCode) {
- var bucket = this._getEncodeBucket(uCode);
- var low = uCode & 0xFF;
- if (bucket[low] <= SEQ_START)
- this.encodeTableSeq[SEQ_START-bucket[low]][DEF_CHAR] = dbcsCode; // There's already a sequence, set a single-char subsequence of it.
- else if (bucket[low] == UNASSIGNED)
- bucket[low] = dbcsCode;
- }
-
- DBCSCodec.prototype._setEncodeSequence = function(seq, dbcsCode) {
-
- // Get the root of character tree according to first character of the sequence.
- var uCode = seq[0];
- var bucket = this._getEncodeBucket(uCode);
- var low = uCode & 0xFF;
-
- var node;
- if (bucket[low] <= SEQ_START) {
- // There's already a sequence with - use it.
- node = this.encodeTableSeq[SEQ_START-bucket[low]];
- }
- else {
- // There was no sequence object - allocate a new one.
- node = {};
- if (bucket[low] !== UNASSIGNED) node[DEF_CHAR] = bucket[low]; // If a char was set before - make it a single-char subsequence.
- bucket[low] = SEQ_START - this.encodeTableSeq.length;
- this.encodeTableSeq.push(node);
- }
-
- // Traverse the character tree, allocating new nodes as needed.
- for (var j = 1; j < seq.length-1; j++) {
- var oldVal = node[uCode];
- if (typeof oldVal === 'object')
- node = oldVal;
- else {
- node = node[uCode] = {}
- if (oldVal !== undefined)
- node[DEF_CHAR] = oldVal
- }
- }
-
- // Set the leaf to given dbcsCode.
- uCode = seq[seq.length-1];
- node[uCode] = dbcsCode;
- }
-
- DBCSCodec.prototype._fillEncodeTable = function(nodeIdx, prefix, skipEncodeChars) {
- var node = this.decodeTables[nodeIdx];
- for (var i = 0; i < 0x100; i++) {
- var uCode = node[i];
- var mbCode = prefix + i;
- if (skipEncodeChars[mbCode])
- continue;
-
- if (uCode >= 0)
- this._setEncodeChar(uCode, mbCode);
- else if (uCode <= NODE_START)
- this._fillEncodeTable(NODE_START - uCode, mbCode << 8, skipEncodeChars);
- else if (uCode <= SEQ_START)
- this._setEncodeSequence(this.decodeTableSeq[SEQ_START - uCode], mbCode);
- }
- }
-
-
-
- // == Actual Encoding ==========================================================
-
-
- function encoderDBCSWrite(str) {
- var newBuf = new Buffer(str.length * (this.gb18030 ? 4 : 3)),
- leadSurrogate = this.leadSurrogate,
- seqObj = this.seqObj, nextChar = -1,
- i = 0, j = 0;
-
- while (true) {
- // 0. Get next character.
- if (nextChar === -1) {
- if (i == str.length) break;
- var uCode = str.charCodeAt(i++);
- }
- else {
- var uCode = nextChar;
- nextChar = -1;
- }
-
- // 1. Handle surrogates.
- if (0xD800 <= uCode && uCode < 0xE000) { // Char is one of surrogates.
- if (uCode < 0xDC00) { // We've got lead surrogate.
- if (leadSurrogate === -1) {
- leadSurrogate = uCode;
- continue;
- } else {
- leadSurrogate = uCode;
- // Double lead surrogate found.
- uCode = UNASSIGNED;
- }
- } else { // We've got trail surrogate.
- if (leadSurrogate !== -1) {
- uCode = 0x10000 + (leadSurrogate - 0xD800) * 0x400 + (uCode - 0xDC00);
- leadSurrogate = -1;
- } else {
- // Incomplete surrogate pair - only trail surrogate found.
- uCode = UNASSIGNED;
- }
-
- }
- }
- else if (leadSurrogate !== -1) {
- // Incomplete surrogate pair - only lead surrogate found.
- nextChar = uCode; uCode = UNASSIGNED; // Write an error, then current char.
- leadSurrogate = -1;
- }
-
- // 2. Convert uCode character.
- var dbcsCode = UNASSIGNED;
- if (seqObj !== undefined && uCode != UNASSIGNED) { // We are in the middle of the sequence
- var resCode = seqObj[uCode];
- if (typeof resCode === 'object') { // Sequence continues.
- seqObj = resCode;
- continue;
-
- } else if (typeof resCode == 'number') { // Sequence finished. Write it.
- dbcsCode = resCode;
-
- } else if (resCode == undefined) { // Current character is not part of the sequence.
-
- // Try default character for this sequence
- resCode = seqObj[DEF_CHAR];
- if (resCode !== undefined) {
- dbcsCode = resCode; // Found. Write it.
- nextChar = uCode; // Current character will be written too in the next iteration.
-
- } else {
- // TODO: What if we have no default? (resCode == undefined)
- // Then, we should write first char of the sequence as-is and try the rest recursively.
- // Didn't do it for now because no encoding has this situation yet.
- // Currently, just skip the sequence and write current char.
- }
- }
- seqObj = undefined;
- }
- else if (uCode >= 0) { // Regular character
- var subtable = this.encodeTable[uCode >> 8];
- if (subtable !== undefined)
- dbcsCode = subtable[uCode & 0xFF];
-
- if (dbcsCode <= SEQ_START) { // Sequence start
- seqObj = this.encodeTableSeq[SEQ_START-dbcsCode];
- continue;
- }
-
- if (dbcsCode == UNASSIGNED && this.gb18030) {
- // Use GB18030 algorithm to find character(s) to write.
- var idx = findIdx(this.gb18030.uChars, uCode);
- if (idx != -1) {
- var dbcsCode = this.gb18030.gbChars[idx] + (uCode - this.gb18030.uChars[idx]);
- newBuf[j++] = 0x81 + Math.floor(dbcsCode / 12600); dbcsCode = dbcsCode % 12600;
- newBuf[j++] = 0x30 + Math.floor(dbcsCode / 1260); dbcsCode = dbcsCode % 1260;
- newBuf[j++] = 0x81 + Math.floor(dbcsCode / 10); dbcsCode = dbcsCode % 10;
- newBuf[j++] = 0x30 + dbcsCode;
- continue;
- }
- }
- }
-
- // 3. Write dbcsCode character.
- if (dbcsCode === UNASSIGNED)
- dbcsCode = this.defaultCharSingleByte;
-
- if (dbcsCode < 0x100) {
- newBuf[j++] = dbcsCode;
- }
- else if (dbcsCode < 0x10000) {
- newBuf[j++] = dbcsCode >> 8; // high byte
- newBuf[j++] = dbcsCode & 0xFF; // low byte
- }
- else {
- newBuf[j++] = dbcsCode >> 16;
- newBuf[j++] = (dbcsCode >> 8) & 0xFF;
- newBuf[j++] = dbcsCode & 0xFF;
- }
- }
-
- this.seqObj = seqObj;
- this.leadSurrogate = leadSurrogate;
- return newBuf.slice(0, j);
- }
-
- function encoderDBCSEnd() {
- if (this.leadSurrogate === -1 && this.seqObj === undefined)
- return; // All clean. Most often case.
-
- var newBuf = new Buffer(10), j = 0;
-
- if (this.seqObj) { // We're in the sequence.
- var dbcsCode = this.seqObj[DEF_CHAR];
- if (dbcsCode !== undefined) { // Write beginning of the sequence.
- if (dbcsCode < 0x100) {
- newBuf[j++] = dbcsCode;
- }
- else {
- newBuf[j++] = dbcsCode >> 8; // high byte
- newBuf[j++] = dbcsCode & 0xFF; // low byte
- }
- } else {
- // See todo above.
- }
- this.seqObj = undefined;
- }
-
- if (this.leadSurrogate !== -1) {
- // Incomplete surrogate pair - only lead surrogate found.
- newBuf[j++] = this.defaultCharSingleByte;
- this.leadSurrogate = -1;
- }
-
- return newBuf.slice(0, j);
- }
-
-
- // == Actual Decoding ==========================================================
-
-
- function decoderDBCSWrite(buf) {
- var newBuf = new Buffer(buf.length*2),
- nodeIdx = this.nodeIdx,
- prevBuf = this.prevBuf, prevBufOffset = this.prevBuf.length,
- seqStart = -this.prevBuf.length, // idx of the start of current parsed sequence.
- uCode;
-
- if (prevBufOffset > 0) // Make prev buf overlap a little to make it easier to slice later.
- prevBuf = Buffer.concat([prevBuf, buf.slice(0, 10)]);
-
- for (var i = 0, j = 0; i < buf.length; i++) {
- var curByte = (i >= 0) ? buf[i] : prevBuf[i + prevBufOffset];
-
- // Lookup in current trie node.
- var uCode = this.decodeTables[nodeIdx][curByte];
-
- if (uCode >= 0) {
- // Normal character, just use it.
- }
- else if (uCode === UNASSIGNED) { // Unknown char.
- // TODO: Callback with seq.
- //var curSeq = (seqStart >= 0) ? buf.slice(seqStart, i+1) : prevBuf.slice(seqStart + prevBufOffset, i+1 + prevBufOffset);
- i = seqStart; // Try to parse again, after skipping first byte of the sequence ('i' will be incremented by 'for' cycle).
- uCode = this.defaultCharUnicode.charCodeAt(0);
- }
- else if (uCode === GB18030_CODE) {
- var curSeq = (seqStart >= 0) ? buf.slice(seqStart, i+1) : prevBuf.slice(seqStart + prevBufOffset, i+1 + prevBufOffset);
- var ptr = (curSeq[0]-0x81)*12600 + (curSeq[1]-0x30)*1260 + (curSeq[2]-0x81)*10 + (curSeq[3]-0x30);
- var idx = findIdx(this.gb18030.gbChars, ptr);
- uCode = this.gb18030.uChars[idx] + ptr - this.gb18030.gbChars[idx];
- }
- else if (uCode <= NODE_START) { // Go to next trie node.
- nodeIdx = NODE_START - uCode;
- continue;
- }
- else if (uCode <= SEQ_START) { // Output a sequence of chars.
- var seq = this.decodeTableSeq[SEQ_START - uCode];
- for (var k = 0; k < seq.length - 1; k++) {
- uCode = seq[k];
- newBuf[j++] = uCode & 0xFF;
- newBuf[j++] = uCode >> 8;
- }
- uCode = seq[seq.length-1];
- }
- else
- throw new Error("Unknown table value when decoding: " + val);
-
- // Write the character to buffer, handling higher planes using surrogate pair.
- if (uCode > 0xFFFF) {
- uCode -= 0x10000;
- var uCodeLead = 0xD800 + Math.floor(uCode / 0x400);
- newBuf[j++] = uCodeLead & 0xFF;
- newBuf[j++] = uCodeLead >> 8;
-
- uCode = 0xDC00 + uCode % 0x400;
- }
- newBuf[j++] = uCode & 0xFF;
- newBuf[j++] = uCode >> 8;
-
- // Reset trie node.
- nodeIdx = 0; seqStart = i+1;
- }
-
- this.nodeIdx = nodeIdx;
- this.prevBuf = (seqStart >= 0) ? buf.slice(seqStart) : prevBuf.slice(seqStart + prevBufOffset);
- return newBuf.slice(0, j).toString('ucs2');
- }
-
- function decoderDBCSEnd() {
- var ret = '';
-
- // Try to parse all remaining chars.
- while (this.prevBuf.length > 0) {
- // Skip 1 character in the buffer.
- ret += this.defaultCharUnicode;
- var buf = this.prevBuf.slice(1);
-
- // Parse remaining as usual.
- this.prevBuf = new Buffer(0);
- this.nodeIdx = 0;
- if (buf.length > 0)
- ret += decoderDBCSWrite.call(this, buf);
- }
-
- this.nodeIdx = 0;
- return ret;
- }
-
- // Binary search for GB18030. Returns largest i such that table[i] <= val.
- function findIdx(table, val) {
- if (table[0] > val)
- return -1;
-
- var l = 0, r = table.length;
- while (l < r-1) { // always table[l] <= val < table[r]
- var mid = l + Math.floor((r-l+1)/2);
- if (table[mid] <= val)
- l = mid;
- else
- r = mid;
- }
- return l;
- }
|