Source file src/pkg/compress/flate/huffman_bit_writer.go

     1	// Copyright 2009 The Go Authors. All rights reserved.
     2	// Use of this source code is governed by a BSD-style
     3	// license that can be found in the LICENSE file.
     4	
     5	package flate
     6	
     7	import (
     8		"io"
     9	)
    10	
    11	const (
    12		// The largest offset code.
    13		offsetCodeCount = 30
    14	
    15		// The special code used to mark the end of a block.
    16		endBlockMarker = 256
    17	
    18		// The first length code.
    19		lengthCodesStart = 257
    20	
    21		// The number of codegen codes.
    22		codegenCodeCount = 19
    23		badCode          = 255
    24	
    25		// bufferFlushSize indicates the buffer size
    26		// after which bytes are flushed to the writer.
    27		// Should preferably be a multiple of 6, since
    28		// we accumulate 6 bytes between writes to the buffer.
    29		bufferFlushSize = 240
    30	
    31		// bufferSize is the actual output byte buffer size.
    32		// It must have additional headroom for a flush
    33		// which can contain up to 8 bytes.
    34		bufferSize = bufferFlushSize + 8
    35	)
    36	
    37	// The number of extra bits needed by length code X - LENGTH_CODES_START.
    38	var lengthExtraBits = []int8{
    39		/* 257 */ 0, 0, 0,
    40		/* 260 */ 0, 0, 0, 0, 0, 1, 1, 1, 1, 2,
    41		/* 270 */ 2, 2, 2, 3, 3, 3, 3, 4, 4, 4,
    42		/* 280 */ 4, 5, 5, 5, 5, 0,
    43	}
    44	
    45	// The length indicated by length code X - LENGTH_CODES_START.
    46	var lengthBase = []uint32{
    47		0, 1, 2, 3, 4, 5, 6, 7, 8, 10,
    48		12, 14, 16, 20, 24, 28, 32, 40, 48, 56,
    49		64, 80, 96, 112, 128, 160, 192, 224, 255,
    50	}
    51	
    52	// offset code word extra bits.
    53	var offsetExtraBits = []int8{
    54		0, 0, 0, 0, 1, 1, 2, 2, 3, 3,
    55		4, 4, 5, 5, 6, 6, 7, 7, 8, 8,
    56		9, 9, 10, 10, 11, 11, 12, 12, 13, 13,
    57	}
    58	
    59	var offsetBase = []uint32{
    60		0x000000, 0x000001, 0x000002, 0x000003, 0x000004,
    61		0x000006, 0x000008, 0x00000c, 0x000010, 0x000018,
    62		0x000020, 0x000030, 0x000040, 0x000060, 0x000080,
    63		0x0000c0, 0x000100, 0x000180, 0x000200, 0x000300,
    64		0x000400, 0x000600, 0x000800, 0x000c00, 0x001000,
    65		0x001800, 0x002000, 0x003000, 0x004000, 0x006000,
    66	}
    67	
    68	// The odd order in which the codegen code sizes are written.
    69	var codegenOrder = []uint32{16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15}
    70	
    71	type huffmanBitWriter struct {
    72		// writer is the underlying writer.
    73		// Do not use it directly; use the write method, which ensures
    74		// that Write errors are sticky.
    75		writer io.Writer
    76	
    77		// Data waiting to be written is bytes[0:nbytes]
    78		// and then the low nbits of bits.
    79		bits            uint64
    80		nbits           uint
    81		bytes           [bufferSize]byte
    82		codegenFreq     [codegenCodeCount]int32
    83		nbytes          int
    84		literalFreq     []int32
    85		offsetFreq      []int32
    86		codegen         []uint8
    87		literalEncoding *huffmanEncoder
    88		offsetEncoding  *huffmanEncoder
    89		codegenEncoding *huffmanEncoder
    90		err             error
    91	}
    92	
    93	func newHuffmanBitWriter(w io.Writer) *huffmanBitWriter {
    94		return &huffmanBitWriter{
    95			writer:          w,
    96			literalFreq:     make([]int32, maxNumLit),
    97			offsetFreq:      make([]int32, offsetCodeCount),
    98			codegen:         make([]uint8, maxNumLit+offsetCodeCount+1),
    99			literalEncoding: newHuffmanEncoder(maxNumLit),
   100			codegenEncoding: newHuffmanEncoder(codegenCodeCount),
   101			offsetEncoding:  newHuffmanEncoder(offsetCodeCount),
   102		}
   103	}
   104	
   105	func (w *huffmanBitWriter) reset(writer io.Writer) {
   106		w.writer = writer
   107		w.bits, w.nbits, w.nbytes, w.err = 0, 0, 0, nil
   108		w.bytes = [bufferSize]byte{}
   109	}
   110	
   111	func (w *huffmanBitWriter) flush() {
   112		if w.err != nil {
   113			w.nbits = 0
   114			return
   115		}
   116		n := w.nbytes
   117		for w.nbits != 0 {
   118			w.bytes[n] = byte(w.bits)
   119			w.bits >>= 8
   120			if w.nbits > 8 { // Avoid underflow
   121				w.nbits -= 8
   122			} else {
   123				w.nbits = 0
   124			}
   125			n++
   126		}
   127		w.bits = 0
   128		w.write(w.bytes[:n])
   129		w.nbytes = 0
   130	}
   131	
   132	func (w *huffmanBitWriter) write(b []byte) {
   133		if w.err != nil {
   134			return
   135		}
   136		_, w.err = w.writer.Write(b)
   137	}
   138	
   139	func (w *huffmanBitWriter) writeBits(b int32, nb uint) {
   140		if w.err != nil {
   141			return
   142		}
   143		w.bits |= uint64(b) << w.nbits
   144		w.nbits += nb
   145		if w.nbits >= 48 {
   146			bits := w.bits
   147			w.bits >>= 48
   148			w.nbits -= 48
   149			n := w.nbytes
   150			bytes := w.bytes[n : n+6]
   151			bytes[0] = byte(bits)
   152			bytes[1] = byte(bits >> 8)
   153			bytes[2] = byte(bits >> 16)
   154			bytes[3] = byte(bits >> 24)
   155			bytes[4] = byte(bits >> 32)
   156			bytes[5] = byte(bits >> 40)
   157			n += 6
   158			if n >= bufferFlushSize {
   159				w.write(w.bytes[:n])
   160				n = 0
   161			}
   162			w.nbytes = n
   163		}
   164	}
   165	
   166	func (w *huffmanBitWriter) writeBytes(bytes []byte) {
   167		if w.err != nil {
   168			return
   169		}
   170		n := w.nbytes
   171		if w.nbits&7 != 0 {
   172			w.err = InternalError("writeBytes with unfinished bits")
   173			return
   174		}
   175		for w.nbits != 0 {
   176			w.bytes[n] = byte(w.bits)
   177			w.bits >>= 8
   178			w.nbits -= 8
   179			n++
   180		}
   181		if n != 0 {
   182			w.write(w.bytes[:n])
   183		}
   184		w.nbytes = 0
   185		w.write(bytes)
   186	}
   187	
   188	// RFC 1951 3.2.7 specifies a special run-length encoding for specifying
   189	// the literal and offset lengths arrays (which are concatenated into a single
   190	// array).  This method generates that run-length encoding.
   191	//
   192	// The result is written into the codegen array, and the frequencies
   193	// of each code is written into the codegenFreq array.
   194	// Codes 0-15 are single byte codes. Codes 16-18 are followed by additional
   195	// information. Code badCode is an end marker
   196	//
   197	//  numLiterals      The number of literals in literalEncoding
   198	//  numOffsets       The number of offsets in offsetEncoding
   199	//  litenc, offenc   The literal and offset encoder to use
   200	func (w *huffmanBitWriter) generateCodegen(numLiterals int, numOffsets int, litEnc, offEnc *huffmanEncoder) {
   201		for i := range w.codegenFreq {
   202			w.codegenFreq[i] = 0
   203		}
   204		// Note that we are using codegen both as a temporary variable for holding
   205		// a copy of the frequencies, and as the place where we put the result.
   206		// This is fine because the output is always shorter than the input used
   207		// so far.
   208		codegen := w.codegen // cache
   209		// Copy the concatenated code sizes to codegen. Put a marker at the end.
   210		cgnl := codegen[:numLiterals]
   211		for i := range cgnl {
   212			cgnl[i] = uint8(litEnc.codes[i].len)
   213		}
   214	
   215		cgnl = codegen[numLiterals : numLiterals+numOffsets]
   216		for i := range cgnl {
   217			cgnl[i] = uint8(offEnc.codes[i].len)
   218		}
   219		codegen[numLiterals+numOffsets] = badCode
   220	
   221		size := codegen[0]
   222		count := 1
   223		outIndex := 0
   224		for inIndex := 1; size != badCode; inIndex++ {
   225			// INVARIANT: We have seen "count" copies of size that have not yet
   226			// had output generated for them.
   227			nextSize := codegen[inIndex]
   228			if nextSize == size {
   229				count++
   230				continue
   231			}
   232			// We need to generate codegen indicating "count" of size.
   233			if size != 0 {
   234				codegen[outIndex] = size
   235				outIndex++
   236				w.codegenFreq[size]++
   237				count--
   238				for count >= 3 {
   239					n := 6
   240					if n > count {
   241						n = count
   242					}
   243					codegen[outIndex] = 16
   244					outIndex++
   245					codegen[outIndex] = uint8(n - 3)
   246					outIndex++
   247					w.codegenFreq[16]++
   248					count -= n
   249				}
   250			} else {
   251				for count >= 11 {
   252					n := 138
   253					if n > count {
   254						n = count
   255					}
   256					codegen[outIndex] = 18
   257					outIndex++
   258					codegen[outIndex] = uint8(n - 11)
   259					outIndex++
   260					w.codegenFreq[18]++
   261					count -= n
   262				}
   263				if count >= 3 {
   264					// count >= 3 && count <= 10
   265					codegen[outIndex] = 17
   266					outIndex++
   267					codegen[outIndex] = uint8(count - 3)
   268					outIndex++
   269					w.codegenFreq[17]++
   270					count = 0
   271				}
   272			}
   273			count--
   274			for ; count >= 0; count-- {
   275				codegen[outIndex] = size
   276				outIndex++
   277				w.codegenFreq[size]++
   278			}
   279			// Set up invariant for next time through the loop.
   280			size = nextSize
   281			count = 1
   282		}
   283		// Marker indicating the end of the codegen.
   284		codegen[outIndex] = badCode
   285	}
   286	
   287	// dynamicSize returns the size of dynamically encoded data in bits.
   288	func (w *huffmanBitWriter) dynamicSize(litEnc, offEnc *huffmanEncoder, extraBits int) (size, numCodegens int) {
   289		numCodegens = len(w.codegenFreq)
   290		for numCodegens > 4 && w.codegenFreq[codegenOrder[numCodegens-1]] == 0 {
   291			numCodegens--
   292		}
   293		header := 3 + 5 + 5 + 4 + (3 * numCodegens) +
   294			w.codegenEncoding.bitLength(w.codegenFreq[:]) +
   295			int(w.codegenFreq[16])*2 +
   296			int(w.codegenFreq[17])*3 +
   297			int(w.codegenFreq[18])*7
   298		size = header +
   299			litEnc.bitLength(w.literalFreq) +
   300			offEnc.bitLength(w.offsetFreq) +
   301			extraBits
   302	
   303		return size, numCodegens
   304	}
   305	
   306	// fixedSize returns the size of dynamically encoded data in bits.
   307	func (w *huffmanBitWriter) fixedSize(extraBits int) int {
   308		return 3 +
   309			fixedLiteralEncoding.bitLength(w.literalFreq) +
   310			fixedOffsetEncoding.bitLength(w.offsetFreq) +
   311			extraBits
   312	}
   313	
   314	// storedSize calculates the stored size, including header.
   315	// The function returns the size in bits and whether the block
   316	// fits inside a single block.
   317	func (w *huffmanBitWriter) storedSize(in []byte) (int, bool) {
   318		if in == nil {
   319			return 0, false
   320		}
   321		if len(in) <= maxStoreBlockSize {
   322			return (len(in) + 5) * 8, true
   323		}
   324		return 0, false
   325	}
   326	
   327	func (w *huffmanBitWriter) writeCode(c hcode) {
   328		if w.err != nil {
   329			return
   330		}
   331		w.bits |= uint64(c.code) << w.nbits
   332		w.nbits += uint(c.len)
   333		if w.nbits >= 48 {
   334			bits := w.bits
   335			w.bits >>= 48
   336			w.nbits -= 48
   337			n := w.nbytes
   338			bytes := w.bytes[n : n+6]
   339			bytes[0] = byte(bits)
   340			bytes[1] = byte(bits >> 8)
   341			bytes[2] = byte(bits >> 16)
   342			bytes[3] = byte(bits >> 24)
   343			bytes[4] = byte(bits >> 32)
   344			bytes[5] = byte(bits >> 40)
   345			n += 6
   346			if n >= bufferFlushSize {
   347				w.write(w.bytes[:n])
   348				n = 0
   349			}
   350			w.nbytes = n
   351		}
   352	}
   353	
   354	// Write the header of a dynamic Huffman block to the output stream.
   355	//
   356	//  numLiterals  The number of literals specified in codegen
   357	//  numOffsets   The number of offsets specified in codegen
   358	//  numCodegens  The number of codegens used in codegen
   359	func (w *huffmanBitWriter) writeDynamicHeader(numLiterals int, numOffsets int, numCodegens int, isEof bool) {
   360		if w.err != nil {
   361			return
   362		}
   363		var firstBits int32 = 4
   364		if isEof {
   365			firstBits = 5
   366		}
   367		w.writeBits(firstBits, 3)
   368		w.writeBits(int32(numLiterals-257), 5)
   369		w.writeBits(int32(numOffsets-1), 5)
   370		w.writeBits(int32(numCodegens-4), 4)
   371	
   372		for i := 0; i < numCodegens; i++ {
   373			value := uint(w.codegenEncoding.codes[codegenOrder[i]].len)
   374			w.writeBits(int32(value), 3)
   375		}
   376	
   377		i := 0
   378		for {
   379			var codeWord int = int(w.codegen[i])
   380			i++
   381			if codeWord == badCode {
   382				break
   383			}
   384			w.writeCode(w.codegenEncoding.codes[uint32(codeWord)])
   385	
   386			switch codeWord {
   387			case 16:
   388				w.writeBits(int32(w.codegen[i]), 2)
   389				i++
   390				break
   391			case 17:
   392				w.writeBits(int32(w.codegen[i]), 3)
   393				i++
   394				break
   395			case 18:
   396				w.writeBits(int32(w.codegen[i]), 7)
   397				i++
   398				break
   399			}
   400		}
   401	}
   402	
   403	func (w *huffmanBitWriter) writeStoredHeader(length int, isEof bool) {
   404		if w.err != nil {
   405			return
   406		}
   407		var flag int32
   408		if isEof {
   409			flag = 1
   410		}
   411		w.writeBits(flag, 3)
   412		w.flush()
   413		w.writeBits(int32(length), 16)
   414		w.writeBits(int32(^uint16(length)), 16)
   415	}
   416	
   417	func (w *huffmanBitWriter) writeFixedHeader(isEof bool) {
   418		if w.err != nil {
   419			return
   420		}
   421		// Indicate that we are a fixed Huffman block
   422		var value int32 = 2
   423		if isEof {
   424			value = 3
   425		}
   426		w.writeBits(value, 3)
   427	}
   428	
   429	// writeBlock will write a block of tokens with the smallest encoding.
   430	// The original input can be supplied, and if the huffman encoded data
   431	// is larger than the original bytes, the data will be written as a
   432	// stored block.
   433	// If the input is nil, the tokens will always be Huffman encoded.
   434	func (w *huffmanBitWriter) writeBlock(tokens []token, eof bool, input []byte) {
   435		if w.err != nil {
   436			return
   437		}
   438	
   439		tokens = append(tokens, endBlockMarker)
   440		numLiterals, numOffsets := w.indexTokens(tokens)
   441	
   442		var extraBits int
   443		storedSize, storable := w.storedSize(input)
   444		if storable {
   445			// We only bother calculating the costs of the extra bits required by
   446			// the length of offset fields (which will be the same for both fixed
   447			// and dynamic encoding), if we need to compare those two encodings
   448			// against stored encoding.
   449			for lengthCode := lengthCodesStart + 8; lengthCode < numLiterals; lengthCode++ {
   450				// First eight length codes have extra size = 0.
   451				extraBits += int(w.literalFreq[lengthCode]) * int(lengthExtraBits[lengthCode-lengthCodesStart])
   452			}
   453			for offsetCode := 4; offsetCode < numOffsets; offsetCode++ {
   454				// First four offset codes have extra size = 0.
   455				extraBits += int(w.offsetFreq[offsetCode]) * int(offsetExtraBits[offsetCode])
   456			}
   457		}
   458	
   459		// Figure out smallest code.
   460		// Fixed Huffman baseline.
   461		var literalEncoding = fixedLiteralEncoding
   462		var offsetEncoding = fixedOffsetEncoding
   463		var size = w.fixedSize(extraBits)
   464	
   465		// Dynamic Huffman?
   466		var numCodegens int
   467	
   468		// Generate codegen and codegenFrequencies, which indicates how to encode
   469		// the literalEncoding and the offsetEncoding.
   470		w.generateCodegen(numLiterals, numOffsets, w.literalEncoding, w.offsetEncoding)
   471		w.codegenEncoding.generate(w.codegenFreq[:], 7)
   472		dynamicSize, numCodegens := w.dynamicSize(w.literalEncoding, w.offsetEncoding, extraBits)
   473	
   474		if dynamicSize < size {
   475			size = dynamicSize
   476			literalEncoding = w.literalEncoding
   477			offsetEncoding = w.offsetEncoding
   478		}
   479	
   480		// Stored bytes?
   481		if storable && storedSize < size {
   482			w.writeStoredHeader(len(input), eof)
   483			w.writeBytes(input)
   484			return
   485		}
   486	
   487		// Huffman.
   488		if literalEncoding == fixedLiteralEncoding {
   489			w.writeFixedHeader(eof)
   490		} else {
   491			w.writeDynamicHeader(numLiterals, numOffsets, numCodegens, eof)
   492		}
   493	
   494		// Write the tokens.
   495		w.writeTokens(tokens, literalEncoding.codes, offsetEncoding.codes)
   496	}
   497	
   498	// writeBlockDynamic encodes a block using a dynamic Huffman table.
   499	// This should be used if the symbols used have a disproportionate
   500	// histogram distribution.
   501	// If input is supplied and the compression savings are below 1/16th of the
   502	// input size the block is stored.
   503	func (w *huffmanBitWriter) writeBlockDynamic(tokens []token, eof bool, input []byte) {
   504		if w.err != nil {
   505			return
   506		}
   507	
   508		tokens = append(tokens, endBlockMarker)
   509		numLiterals, numOffsets := w.indexTokens(tokens)
   510	
   511		// Generate codegen and codegenFrequencies, which indicates how to encode
   512		// the literalEncoding and the offsetEncoding.
   513		w.generateCodegen(numLiterals, numOffsets, w.literalEncoding, w.offsetEncoding)
   514		w.codegenEncoding.generate(w.codegenFreq[:], 7)
   515		size, numCodegens := w.dynamicSize(w.literalEncoding, w.offsetEncoding, 0)
   516	
   517		// Store bytes, if we don't get a reasonable improvement.
   518		if ssize, storable := w.storedSize(input); storable && ssize < (size+size>>4) {
   519			w.writeStoredHeader(len(input), eof)
   520			w.writeBytes(input)
   521			return
   522		}
   523	
   524		// Write Huffman table.
   525		w.writeDynamicHeader(numLiterals, numOffsets, numCodegens, eof)
   526	
   527		// Write the tokens.
   528		w.writeTokens(tokens, w.literalEncoding.codes, w.offsetEncoding.codes)
   529	}
   530	
   531	// indexTokens indexes a slice of tokens, and updates
   532	// literalFreq and offsetFreq, and generates literalEncoding
   533	// and offsetEncoding.
   534	// The number of literal and offset tokens is returned.
   535	func (w *huffmanBitWriter) indexTokens(tokens []token) (numLiterals, numOffsets int) {
   536		for i := range w.literalFreq {
   537			w.literalFreq[i] = 0
   538		}
   539		for i := range w.offsetFreq {
   540			w.offsetFreq[i] = 0
   541		}
   542	
   543		for _, t := range tokens {
   544			if t < matchType {
   545				w.literalFreq[t.literal()]++
   546				continue
   547			}
   548			length := t.length()
   549			offset := t.offset()
   550			w.literalFreq[lengthCodesStart+lengthCode(length)]++
   551			w.offsetFreq[offsetCode(offset)]++
   552		}
   553	
   554		// get the number of literals
   555		numLiterals = len(w.literalFreq)
   556		for w.literalFreq[numLiterals-1] == 0 {
   557			numLiterals--
   558		}
   559		// get the number of offsets
   560		numOffsets = len(w.offsetFreq)
   561		for numOffsets > 0 && w.offsetFreq[numOffsets-1] == 0 {
   562			numOffsets--
   563		}
   564		if numOffsets == 0 {
   565			// We haven't found a single match. If we want to go with the dynamic encoding,
   566			// we should count at least one offset to be sure that the offset huffman tree could be encoded.
   567			w.offsetFreq[0] = 1
   568			numOffsets = 1
   569		}
   570		w.literalEncoding.generate(w.literalFreq, 15)
   571		w.offsetEncoding.generate(w.offsetFreq, 15)
   572		return
   573	}
   574	
   575	// writeTokens writes a slice of tokens to the output.
   576	// codes for literal and offset encoding must be supplied.
   577	func (w *huffmanBitWriter) writeTokens(tokens []token, leCodes, oeCodes []hcode) {
   578		if w.err != nil {
   579			return
   580		}
   581		for _, t := range tokens {
   582			if t < matchType {
   583				w.writeCode(leCodes[t.literal()])
   584				continue
   585			}
   586			// Write the length
   587			length := t.length()
   588			lengthCode := lengthCode(length)
   589			w.writeCode(leCodes[lengthCode+lengthCodesStart])
   590			extraLengthBits := uint(lengthExtraBits[lengthCode])
   591			if extraLengthBits > 0 {
   592				extraLength := int32(length - lengthBase[lengthCode])
   593				w.writeBits(extraLength, extraLengthBits)
   594			}
   595			// Write the offset
   596			offset := t.offset()
   597			offsetCode := offsetCode(offset)
   598			w.writeCode(oeCodes[offsetCode])
   599			extraOffsetBits := uint(offsetExtraBits[offsetCode])
   600			if extraOffsetBits > 0 {
   601				extraOffset := int32(offset - offsetBase[offsetCode])
   602				w.writeBits(extraOffset, extraOffsetBits)
   603			}
   604		}
   605	}
   606	
   607	// huffOffset is a static offset encoder used for huffman only encoding.
   608	// It can be reused since we will not be encoding offset values.
   609	var huffOffset *huffmanEncoder
   610	
   611	func init() {
   612		offsetFreq := make([]int32, offsetCodeCount)
   613		offsetFreq[0] = 1
   614		huffOffset = newHuffmanEncoder(offsetCodeCount)
   615		huffOffset.generate(offsetFreq, 15)
   616	}
   617	
   618	// writeBlockHuff encodes a block of bytes as either
   619	// Huffman encoded literals or uncompressed bytes if the
   620	// results only gains very little from compression.
   621	func (w *huffmanBitWriter) writeBlockHuff(eof bool, input []byte) {
   622		if w.err != nil {
   623			return
   624		}
   625	
   626		// Clear histogram
   627		for i := range w.literalFreq {
   628			w.literalFreq[i] = 0
   629		}
   630	
   631		// Add everything as literals
   632		histogram(input, w.literalFreq)
   633	
   634		w.literalFreq[endBlockMarker] = 1
   635	
   636		const numLiterals = endBlockMarker + 1
   637		const numOffsets = 1
   638	
   639		w.literalEncoding.generate(w.literalFreq, 15)
   640	
   641		// Figure out smallest code.
   642		// Always use dynamic Huffman or Store
   643		var numCodegens int
   644	
   645		// Generate codegen and codegenFrequencies, which indicates how to encode
   646		// the literalEncoding and the offsetEncoding.
   647		w.generateCodegen(numLiterals, numOffsets, w.literalEncoding, huffOffset)
   648		w.codegenEncoding.generate(w.codegenFreq[:], 7)
   649		size, numCodegens := w.dynamicSize(w.literalEncoding, huffOffset, 0)
   650	
   651		// Store bytes, if we don't get a reasonable improvement.
   652		if ssize, storable := w.storedSize(input); storable && ssize < (size+size>>4) {
   653			w.writeStoredHeader(len(input), eof)
   654			w.writeBytes(input)
   655			return
   656		}
   657	
   658		// Huffman.
   659		w.writeDynamicHeader(numLiterals, numOffsets, numCodegens, eof)
   660		encoding := w.literalEncoding.codes[:257]
   661		n := w.nbytes
   662		for _, t := range input {
   663			// Bitwriting inlined, ~30% speedup
   664			c := encoding[t]
   665			w.bits |= uint64(c.code) << w.nbits
   666			w.nbits += uint(c.len)
   667			if w.nbits < 48 {
   668				continue
   669			}
   670			// Store 6 bytes
   671			bits := w.bits
   672			w.bits >>= 48
   673			w.nbits -= 48
   674			bytes := w.bytes[n : n+6]
   675			bytes[0] = byte(bits)
   676			bytes[1] = byte(bits >> 8)
   677			bytes[2] = byte(bits >> 16)
   678			bytes[3] = byte(bits >> 24)
   679			bytes[4] = byte(bits >> 32)
   680			bytes[5] = byte(bits >> 40)
   681			n += 6
   682			if n < bufferFlushSize {
   683				continue
   684			}
   685			w.write(w.bytes[:n])
   686			if w.err != nil {
   687				return // Return early in the event of write failures
   688			}
   689			n = 0
   690		}
   691		w.nbytes = n
   692		w.writeCode(encoding[endBlockMarker])
   693	}
   694	
   695	// histogram accumulates a histogram of b in h.
   696	//
   697	// len(h) must be >= 256, and h's elements must be all zeroes.
   698	func histogram(b []byte, h []int32) {
   699		h = h[:256]
   700		for _, t := range b {
   701			h[t]++
   702		}
   703	}
   704	

View as plain text