ACIL FM
Dark
Refresh
Current DIR:
/lib/node_modules/npm/node_modules/iconv-lite/encodings
/
lib
node_modules
npm
node_modules
iconv-lite
encodings
Upload
Zip Selected
Delete Selected
Pilih semua
Nama
Ukuran
Permission
Aksi
tables
-
chmod
Open
Rename
Delete
dbcs-codec.js
22.52 MB
chmod
View
DL
Edit
Rename
Delete
dbcs-data.js
9.17 MB
chmod
View
DL
Edit
Rename
Delete
index.js
733 B
chmod
View
DL
Edit
Rename
Delete
internal.js
6.16 MB
chmod
View
DL
Edit
Rename
Delete
sbcs-codec.js
2.14 MB
chmod
View
DL
Edit
Rename
Delete
sbcs-data-generated.js
31.28 MB
chmod
View
DL
Edit
Rename
Delete
sbcs-data.js
5 MB
chmod
View
DL
Edit
Rename
Delete
utf7.js
9.07 MB
chmod
View
DL
Edit
Rename
Delete
utf16.js
5.37 MB
chmod
View
DL
Edit
Rename
Delete
utf32.js
9.75 MB
chmod
View
DL
Edit
Rename
Delete
Edit file: /lib/node_modules/npm/node_modules/iconv-lite/encodings/utf32.js
'use strict'; var Buffer = require('safer-buffer').Buffer; // == UTF32-LE/BE codec. ========================================================== exports._utf32 = Utf32Codec; function Utf32Codec(codecOptions, iconv) { this.iconv = iconv; this.bomAware = true; this.isLE = codecOptions.isLE; } exports.utf32le = { type: '_utf32', isLE: true }; exports.utf32be = { type: '_utf32', isLE: false }; // Aliases exports.ucs4le = 'utf32le'; exports.ucs4be = 'utf32be'; Utf32Codec.prototype.encoder = Utf32Encoder; Utf32Codec.prototype.decoder = Utf32Decoder; // -- Encoding function Utf32Encoder(options, codec) { this.isLE = codec.isLE; this.highSurrogate = 0; } Utf32Encoder.prototype.write = function(str) { var src = Buffer.from(str, 'ucs2'); var dst = Buffer.alloc(src.length * 2); var write32 = this.isLE ? dst.writeUInt32LE : dst.writeUInt32BE; var offset = 0; for (var i = 0; i < src.length; i += 2) { var code = src.readUInt16LE(i); var isHighSurrogate = (0xD800 <= code && code < 0xDC00); var isLowSurrogate = (0xDC00 <= code && code < 0xE000); if (this.highSurrogate) { if (isHighSurrogate || !isLowSurrogate) { // There shouldn't be two high surrogates in a row, nor a high surrogate which isn't followed by a low // surrogate. If this happens, keep the pending high surrogate as a stand-alone semi-invalid character // (technically wrong, but expected by some applications, like Windows file names). write32.call(dst, this.highSurrogate, offset); offset += 4; } else { // Create 32-bit value from high and low surrogates; var codepoint = (((this.highSurrogate - 0xD800) << 10) | (code - 0xDC00)) + 0x10000; write32.call(dst, codepoint, offset); offset += 4; this.highSurrogate = 0; continue; } } if (isHighSurrogate) this.highSurrogate = code; else { // Even if the current character is a low surrogate, with no previous high surrogate, we'll // encode it as a semi-invalid stand-alone character for the same reasons expressed above for // unpaired high surrogates. write32.call(dst, code, offset); offset += 4; this.highSurrogate = 0; } } if (offset < dst.length) dst = dst.slice(0, offset); return dst; }; Utf32Encoder.prototype.end = function() { // Treat any leftover high surrogate as a semi-valid independent character. if (!this.highSurrogate) return; var buf = Buffer.alloc(4); if (this.isLE) buf.writeUInt32LE(this.highSurrogate, 0); else buf.writeUInt32BE(this.highSurrogate, 0); this.highSurrogate = 0; return buf; }; // -- Decoding function Utf32Decoder(options, codec) { this.isLE = codec.isLE; this.badChar = codec.iconv.defaultCharUnicode.charCodeAt(0); this.overflow = []; } Utf32Decoder.prototype.write = function(src) { if (src.length === 0) return ''; var i = 0; var codepoint = 0; var dst = Buffer.alloc(src.length + 4); var offset = 0; var isLE = this.isLE; var overflow = this.overflow; var badChar = this.badChar; if (overflow.length > 0) { for (; i < src.length && overflow.length < 4; i++) overflow.push(src[i]); if (overflow.length === 4) { // NOTE: codepoint is a signed int32 and can be negative. // NOTE: We copied this block from below to help V8 optimize it (it works with array, not buffer). if (isLE) { codepoint = overflow[i] | (overflow[i+1] << 8) | (overflow[i+2] << 16) | (overflow[i+3] << 24); } else { codepoint = overflow[i+3] | (overflow[i+2] << 8) | (overflow[i+1] << 16) | (overflow[i] << 24); } overflow.length = 0; offset = _writeCodepoint(dst, offset, codepoint, badChar); } } // Main loop. Should be as optimized as possible. for (; i < src.length - 3; i += 4) { // NOTE: codepoint is a signed int32 and can be negative. if (isLE) { codepoint = src[i] | (src[i+1] << 8) | (src[i+2] << 16) | (src[i+3] << 24); } else { codepoint = src[i+3] | (src[i+2] << 8) | (src[i+1] << 16) | (src[i] << 24); } offset = _writeCodepoint(dst, offset, codepoint, badChar); } // Keep overflowing bytes. for (; i < src.length; i++) { overflow.push(src[i]); } return dst.slice(0, offset).toString('ucs2'); }; function _writeCodepoint(dst, offset, codepoint, badChar) { // NOTE: codepoint is signed int32 and can be negative. We keep it that way to help V8 with optimizations. if (codepoint < 0 || codepoint > 0x10FFFF) { // Not a valid Unicode codepoint codepoint = badChar; } // Ephemeral Planes: Write high surrogate. if (codepoint >= 0x10000) { codepoint -= 0x10000; var high = 0xD800 | (codepoint >> 10); dst[offset++] = high & 0xff; dst[offset++] = high >> 8; // Low surrogate is written below. var codepoint = 0xDC00 | (codepoint & 0x3FF); } // Write BMP char or low surrogate. dst[offset++] = codepoint & 0xff; dst[offset++] = codepoint >> 8; return offset; }; Utf32Decoder.prototype.end = function() { this.overflow.length = 0; }; // == UTF-32 Auto codec ============================================================= // Decoder chooses automatically from UTF-32LE and UTF-32BE using BOM and space-based heuristic. // Defaults to UTF-32LE. http://en.wikipedia.org/wiki/UTF-32 // Encoder/decoder default can be changed: iconv.decode(buf, 'utf32', {defaultEncoding: 'utf-32be'}); // Encoder prepends BOM (which can be overridden with (addBOM: false}). exports.utf32 = Utf32AutoCodec; exports.ucs4 = 'utf32'; function Utf32AutoCodec(options, iconv) { this.iconv = iconv; } Utf32AutoCodec.prototype.encoder = Utf32AutoEncoder; Utf32AutoCodec.prototype.decoder = Utf32AutoDecoder; // -- Encoding function Utf32AutoEncoder(options, codec) { options = options || {}; if (options.addBOM === undefined) options.addBOM = true; this.encoder = codec.iconv.getEncoder(options.defaultEncoding || 'utf-32le', options); } Utf32AutoEncoder.prototype.write = function(str) { return this.encoder.write(str); }; Utf32AutoEncoder.prototype.end = function() { return this.encoder.end(); }; // -- Decoding function Utf32AutoDecoder(options, codec) { this.decoder = null; this.initialBufs = []; this.initialBufsLen = 0; this.options = options || {}; this.iconv = codec.iconv; } Utf32AutoDecoder.prototype.write = function(buf) { if (!this.decoder) { // Codec is not chosen yet. Accumulate initial bytes. this.initialBufs.push(buf); this.initialBufsLen += buf.length; if (this.initialBufsLen < 32) // We need more bytes to use space heuristic (see below) return ''; // We have enough bytes -> detect endianness. var encoding = detectEncoding(this.initialBufs, this.options.defaultEncoding); this.decoder = this.iconv.getDecoder(encoding, this.options); var resStr = ''; for (var i = 0; i < this.initialBufs.length; i++) resStr += this.decoder.write(this.initialBufs[i]); this.initialBufs.length = this.initialBufsLen = 0; return resStr; } return this.decoder.write(buf); }; Utf32AutoDecoder.prototype.end = function() { if (!this.decoder) { var encoding = detectEncoding(this.initialBufs, this.options.defaultEncoding); this.decoder = this.iconv.getDecoder(encoding, this.options); var resStr = ''; for (var i = 0; i < this.initialBufs.length; i++) resStr += this.decoder.write(this.initialBufs[i]); var trail = this.decoder.end(); if (trail) resStr += trail; this.initialBufs.length = this.initialBufsLen = 0; return resStr; } return this.decoder.end(); }; function detectEncoding(bufs, defaultEncoding) { var b = []; var charsProcessed = 0; var invalidLE = 0, invalidBE = 0; // Number of invalid chars when decoded as LE or BE. var bmpCharsLE = 0, bmpCharsBE = 0; // Number of BMP chars when decoded as LE or BE. outer_loop: for (var i = 0; i < bufs.length; i++) { var buf = bufs[i]; for (var j = 0; j < buf.length; j++) { b.push(buf[j]); if (b.length === 4) { if (charsProcessed === 0) { // Check BOM first. if (b[0] === 0xFF && b[1] === 0xFE && b[2] === 0 && b[3] === 0) { return 'utf-32le'; } if (b[0] === 0 && b[1] === 0 && b[2] === 0xFE && b[3] === 0xFF) { return 'utf-32be'; } } if (b[0] !== 0 || b[1] > 0x10) invalidBE++; if (b[3] !== 0 || b[2] > 0x10) invalidLE++; if (b[0] === 0 && b[1] === 0 && (b[2] !== 0 || b[3] !== 0)) bmpCharsBE++; if ((b[0] !== 0 || b[1] !== 0) && b[2] === 0 && b[3] === 0) bmpCharsLE++; b.length = 0; charsProcessed++; if (charsProcessed >= 100) { break outer_loop; } } } } // Make decisions. if (bmpCharsBE - invalidBE > bmpCharsLE - invalidLE) return 'utf-32be'; if (bmpCharsBE - invalidBE < bmpCharsLE - invalidLE) return 'utf-32le'; // Couldn't decide (likely all zeros or not enough data). return defaultEncoding || 'utf-32le'; }
Simpan
Batal
Isi Zip:
Unzip
Create
Buat Folder
Buat File
Terminal / Execute
Run
Chmod Bulk
All File
All Folder
All File dan Folder
Apply