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.  Data is always written
    79  	// sequentially into the bytes array.
    80  	bits            uint64
    81  	nbits           uint
    82  	bytes           [bufferSize]byte
    83  	codegenFreq     [codegenCodeCount]int32
    84  	nbytes          int
    85  	literalFreq     []int32
    86  	offsetFreq      []int32
    87  	codegen         []uint8
    88  	literalEncoding *huffmanEncoder
    89  	offsetEncoding  *huffmanEncoder
    90  	codegenEncoding *huffmanEncoder
    91  	err             error
    92  }
    93  
    94  func newHuffmanBitWriter(w io.Writer) *huffmanBitWriter {
    95  	return &huffmanBitWriter{
    96  		writer:          w,
    97  		literalFreq:     make([]int32, maxNumLit),
    98  		offsetFreq:      make([]int32, offsetCodeCount),
    99  		codegen:         make([]uint8, maxNumLit+offsetCodeCount+1),
   100  		literalEncoding: newHuffmanEncoder(maxNumLit),
   101  		codegenEncoding: newHuffmanEncoder(codegenCodeCount),
   102  		offsetEncoding:  newHuffmanEncoder(offsetCodeCount),
   103  	}
   104  }
   105  
   106  func (w *huffmanBitWriter) reset(writer io.Writer) {
   107  	w.writer = writer
   108  	w.bits, w.nbits, w.nbytes, w.err = 0, 0, 0, nil
   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  	clear(w.codegenFreq[:])
   202  	// Note that we are using codegen both as a temporary variable for holding
   203  	// a copy of the frequencies, and as the place where we put the result.
   204  	// This is fine because the output is always shorter than the input used
   205  	// so far.
   206  	codegen := w.codegen // cache
   207  	// Copy the concatenated code sizes to codegen. Put a marker at the end.
   208  	cgnl := codegen[:numLiterals]
   209  	for i := range cgnl {
   210  		cgnl[i] = uint8(litEnc.codes[i].len)
   211  	}
   212  
   213  	cgnl = codegen[numLiterals : numLiterals+numOffsets]
   214  	for i := range cgnl {
   215  		cgnl[i] = uint8(offEnc.codes[i].len)
   216  	}
   217  	codegen[numLiterals+numOffsets] = badCode
   218  
   219  	size := codegen[0]
   220  	count := 1
   221  	outIndex := 0
   222  	for inIndex := 1; size != badCode; inIndex++ {
   223  		// INVARIANT: We have seen "count" copies of size that have not yet
   224  		// had output generated for them.
   225  		nextSize := codegen[inIndex]
   226  		if nextSize == size {
   227  			count++
   228  			continue
   229  		}
   230  		// We need to generate codegen indicating "count" of size.
   231  		if size != 0 {
   232  			codegen[outIndex] = size
   233  			outIndex++
   234  			w.codegenFreq[size]++
   235  			count--
   236  			for count >= 3 {
   237  				n := 6
   238  				if n > count {
   239  					n = count
   240  				}
   241  				codegen[outIndex] = 16
   242  				outIndex++
   243  				codegen[outIndex] = uint8(n - 3)
   244  				outIndex++
   245  				w.codegenFreq[16]++
   246  				count -= n
   247  			}
   248  		} else {
   249  			for count >= 11 {
   250  				n := 138
   251  				if n > count {
   252  					n = count
   253  				}
   254  				codegen[outIndex] = 18
   255  				outIndex++
   256  				codegen[outIndex] = uint8(n - 11)
   257  				outIndex++
   258  				w.codegenFreq[18]++
   259  				count -= n
   260  			}
   261  			if count >= 3 {
   262  				// count >= 3 && count <= 10
   263  				codegen[outIndex] = 17
   264  				outIndex++
   265  				codegen[outIndex] = uint8(count - 3)
   266  				outIndex++
   267  				w.codegenFreq[17]++
   268  				count = 0
   269  			}
   270  		}
   271  		count--
   272  		for ; count >= 0; count-- {
   273  			codegen[outIndex] = size
   274  			outIndex++
   275  			w.codegenFreq[size]++
   276  		}
   277  		// Set up invariant for next time through the loop.
   278  		size = nextSize
   279  		count = 1
   280  	}
   281  	// Marker indicating the end of the codegen.
   282  	codegen[outIndex] = badCode
   283  }
   284  
   285  // dynamicSize returns the size of dynamically encoded data in bits.
   286  func (w *huffmanBitWriter) dynamicSize(litEnc, offEnc *huffmanEncoder, extraBits int) (size, numCodegens int) {
   287  	numCodegens = len(w.codegenFreq)
   288  	for numCodegens > 4 && w.codegenFreq[codegenOrder[numCodegens-1]] == 0 {
   289  		numCodegens--
   290  	}
   291  	header := 3 + 5 + 5 + 4 + (3 * numCodegens) +
   292  		w.codegenEncoding.bitLength(w.codegenFreq[:]) +
   293  		int(w.codegenFreq[16])*2 +
   294  		int(w.codegenFreq[17])*3 +
   295  		int(w.codegenFreq[18])*7
   296  	size = header +
   297  		litEnc.bitLength(w.literalFreq) +
   298  		offEnc.bitLength(w.offsetFreq) +
   299  		extraBits
   300  
   301  	return size, numCodegens
   302  }
   303  
   304  // fixedSize returns the size of dynamically encoded data in bits.
   305  func (w *huffmanBitWriter) fixedSize(extraBits int) int {
   306  	return 3 +
   307  		fixedLiteralEncoding.bitLength(w.literalFreq) +
   308  		fixedOffsetEncoding.bitLength(w.offsetFreq) +
   309  		extraBits
   310  }
   311  
   312  // storedSize calculates the stored size, including header.
   313  // The function returns the size in bits and whether the block
   314  // fits inside a single block.
   315  func (w *huffmanBitWriter) storedSize(in []byte) (int, bool) {
   316  	if in == nil {
   317  		return 0, false
   318  	}
   319  	if len(in) <= maxStoreBlockSize {
   320  		return (len(in) + 5) * 8, true
   321  	}
   322  	return 0, false
   323  }
   324  
   325  func (w *huffmanBitWriter) writeCode(c hcode) {
   326  	if w.err != nil {
   327  		return
   328  	}
   329  	w.bits |= uint64(c.code) << w.nbits
   330  	w.nbits += uint(c.len)
   331  	if w.nbits >= 48 {
   332  		bits := w.bits
   333  		w.bits >>= 48
   334  		w.nbits -= 48
   335  		n := w.nbytes
   336  		bytes := w.bytes[n : n+6]
   337  		bytes[0] = byte(bits)
   338  		bytes[1] = byte(bits >> 8)
   339  		bytes[2] = byte(bits >> 16)
   340  		bytes[3] = byte(bits >> 24)
   341  		bytes[4] = byte(bits >> 32)
   342  		bytes[5] = byte(bits >> 40)
   343  		n += 6
   344  		if n >= bufferFlushSize {
   345  			w.write(w.bytes[:n])
   346  			n = 0
   347  		}
   348  		w.nbytes = n
   349  	}
   350  }
   351  
   352  // Write the header of a dynamic Huffman block to the output stream.
   353  //
   354  //	numLiterals  The number of literals specified in codegen
   355  //	numOffsets   The number of offsets specified in codegen
   356  //	numCodegens  The number of codegens used in codegen
   357  func (w *huffmanBitWriter) writeDynamicHeader(numLiterals int, numOffsets int, numCodegens int, isEof bool) {
   358  	if w.err != nil {
   359  		return
   360  	}
   361  	var firstBits int32 = 4
   362  	if isEof {
   363  		firstBits = 5
   364  	}
   365  	w.writeBits(firstBits, 3)
   366  	w.writeBits(int32(numLiterals-257), 5)
   367  	w.writeBits(int32(numOffsets-1), 5)
   368  	w.writeBits(int32(numCodegens-4), 4)
   369  
   370  	for i := 0; i < numCodegens; i++ {
   371  		value := uint(w.codegenEncoding.codes[codegenOrder[i]].len)
   372  		w.writeBits(int32(value), 3)
   373  	}
   374  
   375  	i := 0
   376  	for {
   377  		var codeWord int = int(w.codegen[i])
   378  		i++
   379  		if codeWord == badCode {
   380  			break
   381  		}
   382  		w.writeCode(w.codegenEncoding.codes[uint32(codeWord)])
   383  
   384  		switch codeWord {
   385  		case 16:
   386  			w.writeBits(int32(w.codegen[i]), 2)
   387  			i++
   388  		case 17:
   389  			w.writeBits(int32(w.codegen[i]), 3)
   390  			i++
   391  		case 18:
   392  			w.writeBits(int32(w.codegen[i]), 7)
   393  			i++
   394  		}
   395  	}
   396  }
   397  
   398  func (w *huffmanBitWriter) writeStoredHeader(length int, isEof bool) {
   399  	if w.err != nil {
   400  		return
   401  	}
   402  	var flag int32
   403  	if isEof {
   404  		flag = 1
   405  	}
   406  	w.writeBits(flag, 3)
   407  	w.flush()
   408  	w.writeBits(int32(length), 16)
   409  	w.writeBits(int32(^uint16(length)), 16)
   410  }
   411  
   412  func (w *huffmanBitWriter) writeFixedHeader(isEof bool) {
   413  	if w.err != nil {
   414  		return
   415  	}
   416  	// Indicate that we are a fixed Huffman block
   417  	var value int32 = 2
   418  	if isEof {
   419  		value = 3
   420  	}
   421  	w.writeBits(value, 3)
   422  }
   423  
   424  // writeBlock will write a block of tokens with the smallest encoding.
   425  // The original input can be supplied, and if the huffman encoded data
   426  // is larger than the original bytes, the data will be written as a
   427  // stored block.
   428  // If the input is nil, the tokens will always be Huffman encoded.
   429  func (w *huffmanBitWriter) writeBlock(tokens []token, eof bool, input []byte) {
   430  	if w.err != nil {
   431  		return
   432  	}
   433  
   434  	tokens = append(tokens, endBlockMarker)
   435  	numLiterals, numOffsets := w.indexTokens(tokens)
   436  
   437  	var extraBits int
   438  	storedSize, storable := w.storedSize(input)
   439  	if storable {
   440  		// We only bother calculating the costs of the extra bits required by
   441  		// the length of offset fields (which will be the same for both fixed
   442  		// and dynamic encoding), if we need to compare those two encodings
   443  		// against stored encoding.
   444  		for lengthCode := lengthCodesStart + 8; lengthCode < numLiterals; lengthCode++ {
   445  			// First eight length codes have extra size = 0.
   446  			extraBits += int(w.literalFreq[lengthCode]) * int(lengthExtraBits[lengthCode-lengthCodesStart])
   447  		}
   448  		for offsetCode := 4; offsetCode < numOffsets; offsetCode++ {
   449  			// First four offset codes have extra size = 0.
   450  			extraBits += int(w.offsetFreq[offsetCode]) * int(offsetExtraBits[offsetCode])
   451  		}
   452  	}
   453  
   454  	// Figure out smallest code.
   455  	// Fixed Huffman baseline.
   456  	var literalEncoding = fixedLiteralEncoding
   457  	var offsetEncoding = fixedOffsetEncoding
   458  	var size = w.fixedSize(extraBits)
   459  
   460  	// Dynamic Huffman?
   461  	var numCodegens int
   462  
   463  	// Generate codegen and codegenFrequencies, which indicates how to encode
   464  	// the literalEncoding and the offsetEncoding.
   465  	w.generateCodegen(numLiterals, numOffsets, w.literalEncoding, w.offsetEncoding)
   466  	w.codegenEncoding.generate(w.codegenFreq[:], 7)
   467  	dynamicSize, numCodegens := w.dynamicSize(w.literalEncoding, w.offsetEncoding, extraBits)
   468  
   469  	if dynamicSize < size {
   470  		size = dynamicSize
   471  		literalEncoding = w.literalEncoding
   472  		offsetEncoding = w.offsetEncoding
   473  	}
   474  
   475  	// Stored bytes?
   476  	if storable && storedSize < size {
   477  		w.writeStoredHeader(len(input), eof)
   478  		w.writeBytes(input)
   479  		return
   480  	}
   481  
   482  	// Huffman.
   483  	if literalEncoding == fixedLiteralEncoding {
   484  		w.writeFixedHeader(eof)
   485  	} else {
   486  		w.writeDynamicHeader(numLiterals, numOffsets, numCodegens, eof)
   487  	}
   488  
   489  	// Write the tokens.
   490  	w.writeTokens(tokens, literalEncoding.codes, offsetEncoding.codes)
   491  }
   492  
   493  // writeBlockDynamic encodes a block using a dynamic Huffman table.
   494  // This should be used if the symbols used have a disproportionate
   495  // histogram distribution.
   496  // If input is supplied and the compression savings are below 1/16th of the
   497  // input size the block is stored.
   498  func (w *huffmanBitWriter) writeBlockDynamic(tokens []token, eof bool, input []byte) {
   499  	if w.err != nil {
   500  		return
   501  	}
   502  
   503  	tokens = append(tokens, endBlockMarker)
   504  	numLiterals, numOffsets := w.indexTokens(tokens)
   505  
   506  	// Generate codegen and codegenFrequencies, which indicates how to encode
   507  	// the literalEncoding and the offsetEncoding.
   508  	w.generateCodegen(numLiterals, numOffsets, w.literalEncoding, w.offsetEncoding)
   509  	w.codegenEncoding.generate(w.codegenFreq[:], 7)
   510  	size, numCodegens := w.dynamicSize(w.literalEncoding, w.offsetEncoding, 0)
   511  
   512  	// Store bytes, if we don't get a reasonable improvement.
   513  	if ssize, storable := w.storedSize(input); storable && ssize < (size+size>>4) {
   514  		w.writeStoredHeader(len(input), eof)
   515  		w.writeBytes(input)
   516  		return
   517  	}
   518  
   519  	// Write Huffman table.
   520  	w.writeDynamicHeader(numLiterals, numOffsets, numCodegens, eof)
   521  
   522  	// Write the tokens.
   523  	w.writeTokens(tokens, w.literalEncoding.codes, w.offsetEncoding.codes)
   524  }
   525  
   526  // indexTokens indexes a slice of tokens, and updates
   527  // literalFreq and offsetFreq, and generates literalEncoding
   528  // and offsetEncoding.
   529  // The number of literal and offset tokens is returned.
   530  func (w *huffmanBitWriter) indexTokens(tokens []token) (numLiterals, numOffsets int) {
   531  	clear(w.literalFreq)
   532  	clear(w.offsetFreq)
   533  
   534  	for _, t := range tokens {
   535  		if t < matchType {
   536  			w.literalFreq[t.literal()]++
   537  			continue
   538  		}
   539  		length := t.length()
   540  		offset := t.offset()
   541  		w.literalFreq[lengthCodesStart+lengthCode(length)]++
   542  		w.offsetFreq[offsetCode(offset)]++
   543  	}
   544  
   545  	// get the number of literals
   546  	numLiterals = len(w.literalFreq)
   547  	for w.literalFreq[numLiterals-1] == 0 {
   548  		numLiterals--
   549  	}
   550  	// get the number of offsets
   551  	numOffsets = len(w.offsetFreq)
   552  	for numOffsets > 0 && w.offsetFreq[numOffsets-1] == 0 {
   553  		numOffsets--
   554  	}
   555  	if numOffsets == 0 {
   556  		// We haven't found a single match. If we want to go with the dynamic encoding,
   557  		// we should count at least one offset to be sure that the offset huffman tree could be encoded.
   558  		w.offsetFreq[0] = 1
   559  		numOffsets = 1
   560  	}
   561  	w.literalEncoding.generate(w.literalFreq, 15)
   562  	w.offsetEncoding.generate(w.offsetFreq, 15)
   563  	return
   564  }
   565  
   566  // writeTokens writes a slice of tokens to the output.
   567  // codes for literal and offset encoding must be supplied.
   568  func (w *huffmanBitWriter) writeTokens(tokens []token, leCodes, oeCodes []hcode) {
   569  	if w.err != nil {
   570  		return
   571  	}
   572  	for _, t := range tokens {
   573  		if t < matchType {
   574  			w.writeCode(leCodes[t.literal()])
   575  			continue
   576  		}
   577  		// Write the length
   578  		length := t.length()
   579  		lengthCode := lengthCode(length)
   580  		w.writeCode(leCodes[lengthCode+lengthCodesStart])
   581  		extraLengthBits := uint(lengthExtraBits[lengthCode])
   582  		if extraLengthBits > 0 {
   583  			extraLength := int32(length - lengthBase[lengthCode])
   584  			w.writeBits(extraLength, extraLengthBits)
   585  		}
   586  		// Write the offset
   587  		offset := t.offset()
   588  		offsetCode := offsetCode(offset)
   589  		w.writeCode(oeCodes[offsetCode])
   590  		extraOffsetBits := uint(offsetExtraBits[offsetCode])
   591  		if extraOffsetBits > 0 {
   592  			extraOffset := int32(offset - offsetBase[offsetCode])
   593  			w.writeBits(extraOffset, extraOffsetBits)
   594  		}
   595  	}
   596  }
   597  
   598  // huffOffset is a static offset encoder used for huffman only encoding.
   599  // It can be reused since we will not be encoding offset values.
   600  var huffOffset *huffmanEncoder
   601  
   602  func init() {
   603  	offsetFreq := make([]int32, offsetCodeCount)
   604  	offsetFreq[0] = 1
   605  	huffOffset = newHuffmanEncoder(offsetCodeCount)
   606  	huffOffset.generate(offsetFreq, 15)
   607  }
   608  
   609  // writeBlockHuff encodes a block of bytes as either
   610  // Huffman encoded literals or uncompressed bytes if the
   611  // results only gains very little from compression.
   612  func (w *huffmanBitWriter) writeBlockHuff(eof bool, input []byte) {
   613  	if w.err != nil {
   614  		return
   615  	}
   616  
   617  	// Clear histogram
   618  	clear(w.literalFreq)
   619  
   620  	// Add everything as literals
   621  	histogram(input, w.literalFreq)
   622  
   623  	w.literalFreq[endBlockMarker] = 1
   624  
   625  	const numLiterals = endBlockMarker + 1
   626  	w.offsetFreq[0] = 1
   627  	const numOffsets = 1
   628  
   629  	w.literalEncoding.generate(w.literalFreq, 15)
   630  
   631  	// Figure out smallest code.
   632  	// Always use dynamic Huffman or Store
   633  	var numCodegens int
   634  
   635  	// Generate codegen and codegenFrequencies, which indicates how to encode
   636  	// the literalEncoding and the offsetEncoding.
   637  	w.generateCodegen(numLiterals, numOffsets, w.literalEncoding, huffOffset)
   638  	w.codegenEncoding.generate(w.codegenFreq[:], 7)
   639  	size, numCodegens := w.dynamicSize(w.literalEncoding, huffOffset, 0)
   640  
   641  	// Store bytes, if we don't get a reasonable improvement.
   642  	if ssize, storable := w.storedSize(input); storable && ssize < (size+size>>4) {
   643  		w.writeStoredHeader(len(input), eof)
   644  		w.writeBytes(input)
   645  		return
   646  	}
   647  
   648  	// Huffman.
   649  	w.writeDynamicHeader(numLiterals, numOffsets, numCodegens, eof)
   650  	encoding := w.literalEncoding.codes[:257]
   651  	n := w.nbytes
   652  	for _, t := range input {
   653  		// Bitwriting inlined, ~30% speedup
   654  		c := encoding[t]
   655  		w.bits |= uint64(c.code) << w.nbits
   656  		w.nbits += uint(c.len)
   657  		if w.nbits < 48 {
   658  			continue
   659  		}
   660  		// Store 6 bytes
   661  		bits := w.bits
   662  		w.bits >>= 48
   663  		w.nbits -= 48
   664  		bytes := w.bytes[n : n+6]
   665  		bytes[0] = byte(bits)
   666  		bytes[1] = byte(bits >> 8)
   667  		bytes[2] = byte(bits >> 16)
   668  		bytes[3] = byte(bits >> 24)
   669  		bytes[4] = byte(bits >> 32)
   670  		bytes[5] = byte(bits >> 40)
   671  		n += 6
   672  		if n < bufferFlushSize {
   673  			continue
   674  		}
   675  		w.write(w.bytes[:n])
   676  		if w.err != nil {
   677  			return // Return early in the event of write failures
   678  		}
   679  		n = 0
   680  	}
   681  	w.nbytes = n
   682  	w.writeCode(encoding[endBlockMarker])
   683  }
   684  
   685  // histogram accumulates a histogram of b in h.
   686  //
   687  // len(h) must be >= 256, and h's elements must be all zeroes.
   688  func histogram(b []byte, h []int32) {
   689  	h = h[:256]
   690  	for _, t := range b {
   691  		h[t]++
   692  	}
   693  }
   694
View as plain text