Source file src/pkg/compress/flate/deflate.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		"fmt"
     9		"io"
    10		"math"
    11	)
    12	
    13	const (
    14		NoCompression      = 0
    15		BestSpeed          = 1
    16		BestCompression    = 9
    17		DefaultCompression = -1
    18	
    19		// HuffmanOnly disables Lempel-Ziv match searching and only performs Huffman
    20		// entropy encoding. This mode is useful in compressing data that has
    21		// already been compressed with an LZ style algorithm (e.g. Snappy or LZ4)
    22		// that lacks an entropy encoder. Compression gains are achieved when
    23		// certain bytes in the input stream occur more frequently than others.
    24		//
    25		// Note that HuffmanOnly produces a compressed output that is
    26		// RFC 1951 compliant. That is, any valid DEFLATE decompressor will
    27		// continue to be able to decompress this output.
    28		HuffmanOnly = -2
    29	)
    30	
    31	const (
    32		logWindowSize = 15
    33		windowSize    = 1 << logWindowSize
    34		windowMask    = windowSize - 1
    35	
    36		// The LZ77 step produces a sequence of literal tokens and <length, offset>
    37		// pair tokens. The offset is also known as distance. The underlying wire
    38		// format limits the range of lengths and offsets. For example, there are
    39		// 256 legitimate lengths: those in the range [3, 258]. This package's
    40		// compressor uses a higher minimum match length, enabling optimizations
    41		// such as finding matches via 32-bit loads and compares.
    42		baseMatchLength = 3       // The smallest match length per the RFC section 3.2.5
    43		minMatchLength  = 4       // The smallest match length that the compressor actually emits
    44		maxMatchLength  = 258     // The largest match length
    45		baseMatchOffset = 1       // The smallest match offset
    46		maxMatchOffset  = 1 << 15 // The largest match offset
    47	
    48		// The maximum number of tokens we put into a single flate block, just to
    49		// stop things from getting too large.
    50		maxFlateBlockTokens = 1 << 14
    51		maxStoreBlockSize   = 65535
    52		hashBits            = 17 // After 17 performance degrades
    53		hashSize            = 1 << hashBits
    54		hashMask            = (1 << hashBits) - 1
    55		maxHashOffset       = 1 << 24
    56	
    57		skipNever = math.MaxInt32
    58	)
    59	
    60	type compressionLevel struct {
    61		level, good, lazy, nice, chain, fastSkipHashing int
    62	}
    63	
    64	var levels = []compressionLevel{
    65		{0, 0, 0, 0, 0, 0}, // NoCompression.
    66		{1, 0, 0, 0, 0, 0}, // BestSpeed uses a custom algorithm; see deflatefast.go.
    67		// For levels 2-3 we don't bother trying with lazy matches.
    68		{2, 4, 0, 16, 8, 5},
    69		{3, 4, 0, 32, 32, 6},
    70		// Levels 4-9 use increasingly more lazy matching
    71		// and increasingly stringent conditions for "good enough".
    72		{4, 4, 4, 16, 16, skipNever},
    73		{5, 8, 16, 32, 32, skipNever},
    74		{6, 8, 16, 128, 128, skipNever},
    75		{7, 8, 32, 128, 256, skipNever},
    76		{8, 32, 128, 258, 1024, skipNever},
    77		{9, 32, 258, 258, 4096, skipNever},
    78	}
    79	
    80	type compressor struct {
    81		compressionLevel
    82	
    83		w          *huffmanBitWriter
    84		bulkHasher func([]byte, []uint32)
    85	
    86		// compression algorithm
    87		fill      func(*compressor, []byte) int // copy data to window
    88		step      func(*compressor)             // process window
    89		sync      bool                          // requesting flush
    90		bestSpeed *deflateFast                  // Encoder for BestSpeed
    91	
    92		// Input hash chains
    93		// hashHead[hashValue] contains the largest inputIndex with the specified hash value
    94		// If hashHead[hashValue] is within the current window, then
    95		// hashPrev[hashHead[hashValue] & windowMask] contains the previous index
    96		// with the same hash value.
    97		chainHead  int
    98		hashHead   [hashSize]uint32
    99		hashPrev   [windowSize]uint32
   100		hashOffset int
   101	
   102		// input window: unprocessed data is window[index:windowEnd]
   103		index         int
   104		window        []byte
   105		windowEnd     int
   106		blockStart    int  // window index where current tokens start
   107		byteAvailable bool // if true, still need to process window[index-1].
   108	
   109		// queued output tokens
   110		tokens []token
   111	
   112		// deflate state
   113		length         int
   114		offset         int
   115		hash           uint32
   116		maxInsertIndex int
   117		err            error
   118	
   119		// hashMatch must be able to contain hashes for the maximum match length.
   120		hashMatch [maxMatchLength - 1]uint32
   121	}
   122	
   123	func (d *compressor) fillDeflate(b []byte) int {
   124		if d.index >= 2*windowSize-(minMatchLength+maxMatchLength) {
   125			// shift the window by windowSize
   126			copy(d.window, d.window[windowSize:2*windowSize])
   127			d.index -= windowSize
   128			d.windowEnd -= windowSize
   129			if d.blockStart >= windowSize {
   130				d.blockStart -= windowSize
   131			} else {
   132				d.blockStart = math.MaxInt32
   133			}
   134			d.hashOffset += windowSize
   135			if d.hashOffset > maxHashOffset {
   136				delta := d.hashOffset - 1
   137				d.hashOffset -= delta
   138				d.chainHead -= delta
   139	
   140				// Iterate over slices instead of arrays to avoid copying
   141				// the entire table onto the stack (Issue #18625).
   142				for i, v := range d.hashPrev[:] {
   143					if int(v) > delta {
   144						d.hashPrev[i] = uint32(int(v) - delta)
   145					} else {
   146						d.hashPrev[i] = 0
   147					}
   148				}
   149				for i, v := range d.hashHead[:] {
   150					if int(v) > delta {
   151						d.hashHead[i] = uint32(int(v) - delta)
   152					} else {
   153						d.hashHead[i] = 0
   154					}
   155				}
   156			}
   157		}
   158		n := copy(d.window[d.windowEnd:], b)
   159		d.windowEnd += n
   160		return n
   161	}
   162	
   163	func (d *compressor) writeBlock(tokens []token, index int) error {
   164		if index > 0 {
   165			var window []byte
   166			if d.blockStart <= index {
   167				window = d.window[d.blockStart:index]
   168			}
   169			d.blockStart = index
   170			d.w.writeBlock(tokens, false, window)
   171			return d.w.err
   172		}
   173		return nil
   174	}
   175	
   176	// fillWindow will fill the current window with the supplied
   177	// dictionary and calculate all hashes.
   178	// This is much faster than doing a full encode.
   179	// Should only be used after a reset.
   180	func (d *compressor) fillWindow(b []byte) {
   181		// Do not fill window if we are in store-only mode.
   182		if d.compressionLevel.level < 2 {
   183			return
   184		}
   185		if d.index != 0 || d.windowEnd != 0 {
   186			panic("internal error: fillWindow called with stale data")
   187		}
   188	
   189		// If we are given too much, cut it.
   190		if len(b) > windowSize {
   191			b = b[len(b)-windowSize:]
   192		}
   193		// Add all to window.
   194		n := copy(d.window, b)
   195	
   196		// Calculate 256 hashes at the time (more L1 cache hits)
   197		loops := (n + 256 - minMatchLength) / 256
   198		for j := 0; j < loops; j++ {
   199			index := j * 256
   200			end := index + 256 + minMatchLength - 1
   201			if end > n {
   202				end = n
   203			}
   204			toCheck := d.window[index:end]
   205			dstSize := len(toCheck) - minMatchLength + 1
   206	
   207			if dstSize <= 0 {
   208				continue
   209			}
   210	
   211			dst := d.hashMatch[:dstSize]
   212			d.bulkHasher(toCheck, dst)
   213			var newH uint32
   214			for i, val := range dst {
   215				di := i + index
   216				newH = val
   217				hh := &d.hashHead[newH&hashMask]
   218				// Get previous value with the same hash.
   219				// Our chain should point to the previous value.
   220				d.hashPrev[di&windowMask] = *hh
   221				// Set the head of the hash chain to us.
   222				*hh = uint32(di + d.hashOffset)
   223			}
   224			d.hash = newH
   225		}
   226		// Update window information.
   227		d.windowEnd = n
   228		d.index = n
   229	}
   230	
   231	// Try to find a match starting at index whose length is greater than prevSize.
   232	// We only look at chainCount possibilities before giving up.
   233	func (d *compressor) findMatch(pos int, prevHead int, prevLength int, lookahead int) (length, offset int, ok bool) {
   234		minMatchLook := maxMatchLength
   235		if lookahead < minMatchLook {
   236			minMatchLook = lookahead
   237		}
   238	
   239		win := d.window[0 : pos+minMatchLook]
   240	
   241		// We quit when we get a match that's at least nice long
   242		nice := len(win) - pos
   243		if d.nice < nice {
   244			nice = d.nice
   245		}
   246	
   247		// If we've got a match that's good enough, only look in 1/4 the chain.
   248		tries := d.chain
   249		length = prevLength
   250		if length >= d.good {
   251			tries >>= 2
   252		}
   253	
   254		wEnd := win[pos+length]
   255		wPos := win[pos:]
   256		minIndex := pos - windowSize
   257	
   258		for i := prevHead; tries > 0; tries-- {
   259			if wEnd == win[i+length] {
   260				n := matchLen(win[i:], wPos, minMatchLook)
   261	
   262				if n > length && (n > minMatchLength || pos-i <= 4096) {
   263					length = n
   264					offset = pos - i
   265					ok = true
   266					if n >= nice {
   267						// The match is good enough that we don't try to find a better one.
   268						break
   269					}
   270					wEnd = win[pos+n]
   271				}
   272			}
   273			if i == minIndex {
   274				// hashPrev[i & windowMask] has already been overwritten, so stop now.
   275				break
   276			}
   277			i = int(d.hashPrev[i&windowMask]) - d.hashOffset
   278			if i < minIndex || i < 0 {
   279				break
   280			}
   281		}
   282		return
   283	}
   284	
   285	func (d *compressor) writeStoredBlock(buf []byte) error {
   286		if d.w.writeStoredHeader(len(buf), false); d.w.err != nil {
   287			return d.w.err
   288		}
   289		d.w.writeBytes(buf)
   290		return d.w.err
   291	}
   292	
   293	const hashmul = 0x1e35a7bd
   294	
   295	// hash4 returns a hash representation of the first 4 bytes
   296	// of the supplied slice.
   297	// The caller must ensure that len(b) >= 4.
   298	func hash4(b []byte) uint32 {
   299		return ((uint32(b[3]) | uint32(b[2])<<8 | uint32(b[1])<<16 | uint32(b[0])<<24) * hashmul) >> (32 - hashBits)
   300	}
   301	
   302	// bulkHash4 will compute hashes using the same
   303	// algorithm as hash4
   304	func bulkHash4(b []byte, dst []uint32) {
   305		if len(b) < minMatchLength {
   306			return
   307		}
   308		hb := uint32(b[3]) | uint32(b[2])<<8 | uint32(b[1])<<16 | uint32(b[0])<<24
   309		dst[0] = (hb * hashmul) >> (32 - hashBits)
   310		end := len(b) - minMatchLength + 1
   311		for i := 1; i < end; i++ {
   312			hb = (hb << 8) | uint32(b[i+3])
   313			dst[i] = (hb * hashmul) >> (32 - hashBits)
   314		}
   315	}
   316	
   317	// matchLen returns the number of matching bytes in a and b
   318	// up to length 'max'. Both slices must be at least 'max'
   319	// bytes in size.
   320	func matchLen(a, b []byte, max int) int {
   321		a = a[:max]
   322		b = b[:len(a)]
   323		for i, av := range a {
   324			if b[i] != av {
   325				return i
   326			}
   327		}
   328		return max
   329	}
   330	
   331	// encSpeed will compress and store the currently added data,
   332	// if enough has been accumulated or we at the end of the stream.
   333	// Any error that occurred will be in d.err
   334	func (d *compressor) encSpeed() {
   335		// We only compress if we have maxStoreBlockSize.
   336		if d.windowEnd < maxStoreBlockSize {
   337			if !d.sync {
   338				return
   339			}
   340	
   341			// Handle small sizes.
   342			if d.windowEnd < 128 {
   343				switch {
   344				case d.windowEnd == 0:
   345					return
   346				case d.windowEnd <= 16:
   347					d.err = d.writeStoredBlock(d.window[:d.windowEnd])
   348				default:
   349					d.w.writeBlockHuff(false, d.window[:d.windowEnd])
   350					d.err = d.w.err
   351				}
   352				d.windowEnd = 0
   353				d.bestSpeed.reset()
   354				return
   355			}
   356	
   357		}
   358		// Encode the block.
   359		d.tokens = d.bestSpeed.encode(d.tokens[:0], d.window[:d.windowEnd])
   360	
   361		// If we removed less than 1/16th, Huffman compress the block.
   362		if len(d.tokens) > d.windowEnd-(d.windowEnd>>4) {
   363			d.w.writeBlockHuff(false, d.window[:d.windowEnd])
   364		} else {
   365			d.w.writeBlockDynamic(d.tokens, false, d.window[:d.windowEnd])
   366		}
   367		d.err = d.w.err
   368		d.windowEnd = 0
   369	}
   370	
   371	func (d *compressor) initDeflate() {
   372		d.window = make([]byte, 2*windowSize)
   373		d.hashOffset = 1
   374		d.tokens = make([]token, 0, maxFlateBlockTokens+1)
   375		d.length = minMatchLength - 1
   376		d.offset = 0
   377		d.byteAvailable = false
   378		d.index = 0
   379		d.hash = 0
   380		d.chainHead = -1
   381		d.bulkHasher = bulkHash4
   382	}
   383	
   384	func (d *compressor) deflate() {
   385		if d.windowEnd-d.index < minMatchLength+maxMatchLength && !d.sync {
   386			return
   387		}
   388	
   389		d.maxInsertIndex = d.windowEnd - (minMatchLength - 1)
   390		if d.index < d.maxInsertIndex {
   391			d.hash = hash4(d.window[d.index : d.index+minMatchLength])
   392		}
   393	
   394	Loop:
   395		for {
   396			if d.index > d.windowEnd {
   397				panic("index > windowEnd")
   398			}
   399			lookahead := d.windowEnd - d.index
   400			if lookahead < minMatchLength+maxMatchLength {
   401				if !d.sync {
   402					break Loop
   403				}
   404				if d.index > d.windowEnd {
   405					panic("index > windowEnd")
   406				}
   407				if lookahead == 0 {
   408					// Flush current output block if any.
   409					if d.byteAvailable {
   410						// There is still one pending token that needs to be flushed
   411						d.tokens = append(d.tokens, literalToken(uint32(d.window[d.index-1])))
   412						d.byteAvailable = false
   413					}
   414					if len(d.tokens) > 0 {
   415						if d.err = d.writeBlock(d.tokens, d.index); d.err != nil {
   416							return
   417						}
   418						d.tokens = d.tokens[:0]
   419					}
   420					break Loop
   421				}
   422			}
   423			if d.index < d.maxInsertIndex {
   424				// Update the hash
   425				d.hash = hash4(d.window[d.index : d.index+minMatchLength])
   426				hh := &d.hashHead[d.hash&hashMask]
   427				d.chainHead = int(*hh)
   428				d.hashPrev[d.index&windowMask] = uint32(d.chainHead)
   429				*hh = uint32(d.index + d.hashOffset)
   430			}
   431			prevLength := d.length
   432			prevOffset := d.offset
   433			d.length = minMatchLength - 1
   434			d.offset = 0
   435			minIndex := d.index - windowSize
   436			if minIndex < 0 {
   437				minIndex = 0
   438			}
   439	
   440			if d.chainHead-d.hashOffset >= minIndex &&
   441				(d.fastSkipHashing != skipNever && lookahead > minMatchLength-1 ||
   442					d.fastSkipHashing == skipNever && lookahead > prevLength && prevLength < d.lazy) {
   443				if newLength, newOffset, ok := d.findMatch(d.index, d.chainHead-d.hashOffset, minMatchLength-1, lookahead); ok {
   444					d.length = newLength
   445					d.offset = newOffset
   446				}
   447			}
   448			if d.fastSkipHashing != skipNever && d.length >= minMatchLength ||
   449				d.fastSkipHashing == skipNever && prevLength >= minMatchLength && d.length <= prevLength {
   450				// There was a match at the previous step, and the current match is
   451				// not better. Output the previous match.
   452				if d.fastSkipHashing != skipNever {
   453					d.tokens = append(d.tokens, matchToken(uint32(d.length-baseMatchLength), uint32(d.offset-baseMatchOffset)))
   454				} else {
   455					d.tokens = append(d.tokens, matchToken(uint32(prevLength-baseMatchLength), uint32(prevOffset-baseMatchOffset)))
   456				}
   457				// Insert in the hash table all strings up to the end of the match.
   458				// index and index-1 are already inserted. If there is not enough
   459				// lookahead, the last two strings are not inserted into the hash
   460				// table.
   461				if d.length <= d.fastSkipHashing {
   462					var newIndex int
   463					if d.fastSkipHashing != skipNever {
   464						newIndex = d.index + d.length
   465					} else {
   466						newIndex = d.index + prevLength - 1
   467					}
   468					for d.index++; d.index < newIndex; d.index++ {
   469						if d.index < d.maxInsertIndex {
   470							d.hash = hash4(d.window[d.index : d.index+minMatchLength])
   471							// Get previous value with the same hash.
   472							// Our chain should point to the previous value.
   473							hh := &d.hashHead[d.hash&hashMask]
   474							d.hashPrev[d.index&windowMask] = *hh
   475							// Set the head of the hash chain to us.
   476							*hh = uint32(d.index + d.hashOffset)
   477						}
   478					}
   479					if d.fastSkipHashing == skipNever {
   480						d.byteAvailable = false
   481						d.length = minMatchLength - 1
   482					}
   483				} else {
   484					// For matches this long, we don't bother inserting each individual
   485					// item into the table.
   486					d.index += d.length
   487					if d.index < d.maxInsertIndex {
   488						d.hash = hash4(d.window[d.index : d.index+minMatchLength])
   489					}
   490				}
   491				if len(d.tokens) == maxFlateBlockTokens {
   492					// The block includes the current character
   493					if d.err = d.writeBlock(d.tokens, d.index); d.err != nil {
   494						return
   495					}
   496					d.tokens = d.tokens[:0]
   497				}
   498			} else {
   499				if d.fastSkipHashing != skipNever || d.byteAvailable {
   500					i := d.index - 1
   501					if d.fastSkipHashing != skipNever {
   502						i = d.index
   503					}
   504					d.tokens = append(d.tokens, literalToken(uint32(d.window[i])))
   505					if len(d.tokens) == maxFlateBlockTokens {
   506						if d.err = d.writeBlock(d.tokens, i+1); d.err != nil {
   507							return
   508						}
   509						d.tokens = d.tokens[:0]
   510					}
   511				}
   512				d.index++
   513				if d.fastSkipHashing == skipNever {
   514					d.byteAvailable = true
   515				}
   516			}
   517		}
   518	}
   519	
   520	func (d *compressor) fillStore(b []byte) int {
   521		n := copy(d.window[d.windowEnd:], b)
   522		d.windowEnd += n
   523		return n
   524	}
   525	
   526	func (d *compressor) store() {
   527		if d.windowEnd > 0 && (d.windowEnd == maxStoreBlockSize || d.sync) {
   528			d.err = d.writeStoredBlock(d.window[:d.windowEnd])
   529			d.windowEnd = 0
   530		}
   531	}
   532	
   533	// storeHuff compresses and stores the currently added data
   534	// when the d.window is full or we are at the end of the stream.
   535	// Any error that occurred will be in d.err
   536	func (d *compressor) storeHuff() {
   537		if d.windowEnd < len(d.window) && !d.sync || d.windowEnd == 0 {
   538			return
   539		}
   540		d.w.writeBlockHuff(false, d.window[:d.windowEnd])
   541		d.err = d.w.err
   542		d.windowEnd = 0
   543	}
   544	
   545	func (d *compressor) write(b []byte) (n int, err error) {
   546		if d.err != nil {
   547			return 0, d.err
   548		}
   549		n = len(b)
   550		for len(b) > 0 {
   551			d.step(d)
   552			b = b[d.fill(d, b):]
   553			if d.err != nil {
   554				return 0, d.err
   555			}
   556		}
   557		return n, nil
   558	}
   559	
   560	func (d *compressor) syncFlush() error {
   561		if d.err != nil {
   562			return d.err
   563		}
   564		d.sync = true
   565		d.step(d)
   566		if d.err == nil {
   567			d.w.writeStoredHeader(0, false)
   568			d.w.flush()
   569			d.err = d.w.err
   570		}
   571		d.sync = false
   572		return d.err
   573	}
   574	
   575	func (d *compressor) init(w io.Writer, level int) (err error) {
   576		d.w = newHuffmanBitWriter(w)
   577	
   578		switch {
   579		case level == NoCompression:
   580			d.window = make([]byte, maxStoreBlockSize)
   581			d.fill = (*compressor).fillStore
   582			d.step = (*compressor).store
   583		case level == HuffmanOnly:
   584			d.window = make([]byte, maxStoreBlockSize)
   585			d.fill = (*compressor).fillStore
   586			d.step = (*compressor).storeHuff
   587		case level == BestSpeed:
   588			d.compressionLevel = levels[level]
   589			d.window = make([]byte, maxStoreBlockSize)
   590			d.fill = (*compressor).fillStore
   591			d.step = (*compressor).encSpeed
   592			d.bestSpeed = newDeflateFast()
   593			d.tokens = make([]token, maxStoreBlockSize)
   594		case level == DefaultCompression:
   595			level = 6
   596			fallthrough
   597		case 2 <= level && level <= 9:
   598			d.compressionLevel = levels[level]
   599			d.initDeflate()
   600			d.fill = (*compressor).fillDeflate
   601			d.step = (*compressor).deflate
   602		default:
   603			return fmt.Errorf("flate: invalid compression level %d: want value in range [-2, 9]", level)
   604		}
   605		return nil
   606	}
   607	
   608	func (d *compressor) reset(w io.Writer) {
   609		d.w.reset(w)
   610		d.sync = false
   611		d.err = nil
   612		switch d.compressionLevel.level {
   613		case NoCompression:
   614			d.windowEnd = 0
   615		case BestSpeed:
   616			d.windowEnd = 0
   617			d.tokens = d.tokens[:0]
   618			d.bestSpeed.reset()
   619		default:
   620			d.chainHead = -1
   621			for i := range d.hashHead {
   622				d.hashHead[i] = 0
   623			}
   624			for i := range d.hashPrev {
   625				d.hashPrev[i] = 0
   626			}
   627			d.hashOffset = 1
   628			d.index, d.windowEnd = 0, 0
   629			d.blockStart, d.byteAvailable = 0, false
   630			d.tokens = d.tokens[:0]
   631			d.length = minMatchLength - 1
   632			d.offset = 0
   633			d.hash = 0
   634			d.maxInsertIndex = 0
   635		}
   636	}
   637	
   638	func (d *compressor) close() error {
   639		if d.err != nil {
   640			return d.err
   641		}
   642		d.sync = true
   643		d.step(d)
   644		if d.err != nil {
   645			return d.err
   646		}
   647		if d.w.writeStoredHeader(0, true); d.w.err != nil {
   648			return d.w.err
   649		}
   650		d.w.flush()
   651		return d.w.err
   652	}
   653	
   654	// NewWriter returns a new Writer compressing data at the given level.
   655	// Following zlib, levels range from 1 (BestSpeed) to 9 (BestCompression);
   656	// higher levels typically run slower but compress more. Level 0
   657	// (NoCompression) does not attempt any compression; it only adds the
   658	// necessary DEFLATE framing.
   659	// Level -1 (DefaultCompression) uses the default compression level.
   660	// Level -2 (HuffmanOnly) will use Huffman compression only, giving
   661	// a very fast compression for all types of input, but sacrificing considerable
   662	// compression efficiency.
   663	//
   664	// If level is in the range [-2, 9] then the error returned will be nil.
   665	// Otherwise the error returned will be non-nil.
   666	func NewWriter(w io.Writer, level int) (*Writer, error) {
   667		var dw Writer
   668		if err := dw.d.init(w, level); err != nil {
   669			return nil, err
   670		}
   671		return &dw, nil
   672	}
   673	
   674	// NewWriterDict is like NewWriter but initializes the new
   675	// Writer with a preset dictionary. The returned Writer behaves
   676	// as if the dictionary had been written to it without producing
   677	// any compressed output. The compressed data written to w
   678	// can only be decompressed by a Reader initialized with the
   679	// same dictionary.
   680	func NewWriterDict(w io.Writer, level int, dict []byte) (*Writer, error) {
   681		dw := &dictWriter{w}
   682		zw, err := NewWriter(dw, level)
   683		if err != nil {
   684			return nil, err
   685		}
   686		zw.d.fillWindow(dict)
   687		zw.dict = append(zw.dict, dict...) // duplicate dictionary for Reset method.
   688		return zw, err
   689	}
   690	
   691	type dictWriter struct {
   692		w io.Writer
   693	}
   694	
   695	func (w *dictWriter) Write(b []byte) (n int, err error) {
   696		return w.w.Write(b)
   697	}
   698	
   699	// A Writer takes data written to it and writes the compressed
   700	// form of that data to an underlying writer (see NewWriter).
   701	type Writer struct {
   702		d    compressor
   703		dict []byte
   704	}
   705	
   706	// Write writes data to w, which will eventually write the
   707	// compressed form of data to its underlying writer.
   708	func (w *Writer) Write(data []byte) (n int, err error) {
   709		return w.d.write(data)
   710	}
   711	
   712	// Flush flushes any pending data to the underlying writer.
   713	// It is useful mainly in compressed network protocols, to ensure that
   714	// a remote reader has enough data to reconstruct a packet.
   715	// Flush does not return until the data has been written.
   716	// Calling Flush when there is no pending data still causes the Writer
   717	// to emit a sync marker of at least 4 bytes.
   718	// If the underlying writer returns an error, Flush returns that error.
   719	//
   720	// In the terminology of the zlib library, Flush is equivalent to Z_SYNC_FLUSH.
   721	func (w *Writer) Flush() error {
   722		// For more about flushing:
   723		// https://www.bolet.org/~pornin/deflate-flush.html
   724		return w.d.syncFlush()
   725	}
   726	
   727	// Close flushes and closes the writer.
   728	func (w *Writer) Close() error {
   729		return w.d.close()
   730	}
   731	
   732	// Reset discards the writer's state and makes it equivalent to
   733	// the result of NewWriter or NewWriterDict called with dst
   734	// and w's level and dictionary.
   735	func (w *Writer) Reset(dst io.Writer) {
   736		if dw, ok := w.d.w.writer.(*dictWriter); ok {
   737			// w was created with NewWriterDict
   738			dw.w = dst
   739			w.d.reset(dw)
   740			w.d.fillWindow(w.dict)
   741		} else {
   742			// w was created with NewWriter
   743			w.d.reset(dst)
   744		}
   745	}
   746	

View as plain text