/vendor/github.com/klauspost/compress/flate/huffman_bit_writer.go at ba9e4e25705b04fbcd6ebe61c8dd0aa075dc3b0e

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034

// Copyright 2009 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.

package flate

import (
	"encoding/binary"
	"fmt"
	"io"
)

const (
	// The largest offset code.
	offsetCodeCount = 30

	// The special code used to mark the end of a block.
	endBlockMarker = 256

	// The first length code.
	lengthCodesStart = 257

	// The number of codegen codes.
	codegenCodeCount = 19
	badCode          = 255

	// bufferFlushSize indicates the buffer size
	// after which bytes are flushed to the writer.
	// Should preferably be a multiple of 6, since
	// we accumulate 6 bytes between writes to the buffer.
	bufferFlushSize = 246

	// bufferSize is the actual output byte buffer size.
	// It must have additional headroom for a flush
	// which can contain up to 8 bytes.
	bufferSize = bufferFlushSize + 8
)

// The number of extra bits needed by length code X - LENGTH_CODES_START.
var lengthExtraBits = [32]int8{
	/* 257 */ 0, 0, 0,
	/* 260 */ 0, 0, 0, 0, 0, 1, 1, 1, 1, 2,
	/* 270 */ 2, 2, 2, 3, 3, 3, 3, 4, 4, 4,
	/* 280 */ 4, 5, 5, 5, 5, 0,
}

// The length indicated by length code X - LENGTH_CODES_START.
var lengthBase = [32]uint8{
	0, 1, 2, 3, 4, 5, 6, 7, 8, 10,
	12, 14, 16, 20, 24, 28, 32, 40, 48, 56,
	64, 80, 96, 112, 128, 160, 192, 224, 255,
}

// offset code word extra bits.
var offsetExtraBits = [64]int8{
	0, 0, 0, 0, 1, 1, 2, 2, 3, 3,
	4, 4, 5, 5, 6, 6, 7, 7, 8, 8,
	9, 9, 10, 10, 11, 11, 12, 12, 13, 13,
	/* extended window */
	14, 14, 15, 15, 16, 16, 17, 17, 18, 18, 19, 19, 20, 20,
}

var offsetCombined = [32]uint32{}

func init() {
	var offsetBase = [64]uint32{
		/* normal deflate */
		0x000000, 0x000001, 0x000002, 0x000003, 0x000004,
		0x000006, 0x000008, 0x00000c, 0x000010, 0x000018,
		0x000020, 0x000030, 0x000040, 0x000060, 0x000080,
		0x0000c0, 0x000100, 0x000180, 0x000200, 0x000300,
		0x000400, 0x000600, 0x000800, 0x000c00, 0x001000,
		0x001800, 0x002000, 0x003000, 0x004000, 0x006000,

		/* extended window */
		0x008000, 0x00c000, 0x010000, 0x018000, 0x020000,
		0x030000, 0x040000, 0x060000, 0x080000, 0x0c0000,
		0x100000, 0x180000, 0x200000, 0x300000,
	}

	for i := range offsetCombined[:] {
		// Don't use extended window values...
		if offsetBase[i] > 0x006000 {
			continue
		}
		offsetCombined[i] = uint32(offsetExtraBits[i])<<16 | (offsetBase[i])
	}
}

// The odd order in which the codegen code sizes are written.
var codegenOrder = []uint32{16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15}

type huffmanBitWriter struct {
	// writer is the underlying writer.
	// Do not use it directly; use the write method, which ensures
	// that Write errors are sticky.
	writer io.Writer

	// Data waiting to be written is bytes[0:nbytes]
	// and then the low nbits of bits.
	bits            uint64
	nbits           uint16
	nbytes          uint8
	lastHuffMan     bool
	literalEncoding *huffmanEncoder
	tmpLitEncoding  *huffmanEncoder
	offsetEncoding  *huffmanEncoder
	codegenEncoding *huffmanEncoder
	err             error
	lastHeader      int
	// Set between 0 (reused block can be up to 2x the size)
	logNewTablePenalty uint
	bytes              [256 + 8]byte
	literalFreq        [lengthCodesStart + 32]uint16
	offsetFreq         [32]uint16
	codegenFreq        [codegenCodeCount]uint16

	// codegen must have an extra space for the final symbol.
	codegen [literalCount + offsetCodeCount + 1]uint8
}

// Huffman reuse.
//
// The huffmanBitWriter supports reusing huffman tables and thereby combining block sections.
//
// This is controlled by several variables:
//
// If lastHeader is non-zero the Huffman table can be reused.
// This also indicates that a Huffman table has been generated that can output all
// possible symbols.
// It also indicates that an EOB has not yet been emitted, so if a new tabel is generated
// an EOB with the previous table must be written.
//
// If lastHuffMan is set, a table for outputting literals has been generated and offsets are invalid.
//
// An incoming block estimates the output size of a new table using a 'fresh' by calculating the
// optimal size and adding a penalty in 'logNewTablePenalty'.
// A Huffman table is not optimal, which is why we add a penalty, and generating a new table
// is slower both for compression and decompression.

func newHuffmanBitWriter(w io.Writer) *huffmanBitWriter {
	return &huffmanBitWriter{
		writer:          w,
		literalEncoding: newHuffmanEncoder(literalCount),
		tmpLitEncoding:  newHuffmanEncoder(literalCount),
		codegenEncoding: newHuffmanEncoder(codegenCodeCount),
		offsetEncoding:  newHuffmanEncoder(offsetCodeCount),
	}
}

func (w *huffmanBitWriter) reset(writer io.Writer) {
	w.writer = writer
	w.bits, w.nbits, w.nbytes, w.err = 0, 0, 0, nil
	w.lastHeader = 0
	w.lastHuffMan = false
}

func (w *huffmanBitWriter) canReuse(t *tokens) (offsets, lits bool) {
	offsets, lits = true, true
	a := t.offHist[:offsetCodeCount]
	b := w.offsetFreq[:len(a)]
	for i := range a {
		if b[i] == 0 && a[i] != 0 {
			offsets = false
			break
		}
	}

	a = t.extraHist[:literalCount-256]
	b = w.literalFreq[256:literalCount]
	b = b[:len(a)]
	for i := range a {
		if b[i] == 0 && a[i] != 0 {
			lits = false
			break
		}
	}
	if lits {
		a = t.litHist[:]
		b = w.literalFreq[:len(a)]
		for i := range a {
			if b[i] == 0 && a[i] != 0 {
				lits = false
				break
			}
		}
	}
	return
}

func (w *huffmanBitWriter) flush() {
	if w.err != nil {
		w.nbits = 0
		return
	}
	if w.lastHeader > 0 {
		// We owe an EOB
		w.writeCode(w.literalEncoding.codes[endBlockMarker])
		w.lastHeader = 0
	}
	n := w.nbytes
	for w.nbits != 0 {
		w.bytes[n] = byte(w.bits)
		w.bits >>= 8
		if w.nbits > 8 { // Avoid underflow
			w.nbits -= 8
		} else {
			w.nbits = 0
		}
		n++
	}
	w.bits = 0
	w.write(w.bytes[:n])
	w.nbytes = 0
}

func (w *huffmanBitWriter) write(b []byte) {
	if w.err != nil {
		return
	}
	_, w.err = w.writer.Write(b)
}

func (w *huffmanBitWriter) writeBits(b int32, nb uint16) {
	w.bits |= uint64(b) << w.nbits
	w.nbits += nb
	if w.nbits >= 48 {
		w.writeOutBits()
	}
}

func (w *huffmanBitWriter) writeBytes(bytes []byte) {
	if w.err != nil {
		return
	}
	n := w.nbytes
	if w.nbits&7 != 0 {
		w.err = InternalError("writeBytes with unfinished bits")
		return
	}
	for w.nbits != 0 {
		w.bytes[n] = byte(w.bits)
		w.bits >>= 8
		w.nbits -= 8
		n++
	}
	if n != 0 {
		w.write(w.bytes[:n])
	}
	w.nbytes = 0
	w.write(bytes)
}

// RFC 1951 3.2.7 specifies a special run-length encoding for specifying
// the literal and offset lengths arrays (which are concatenated into a single
// array).  This method generates that run-length encoding.
//
// The result is written into the codegen array, and the frequencies
// of each code is written into the codegenFreq array.
// Codes 0-15 are single byte codes. Codes 16-18 are followed by additional
// information. Code badCode is an end marker
//
//  numLiterals      The number of literals in literalEncoding
//  numOffsets       The number of offsets in offsetEncoding
//  litenc, offenc   The literal and offset encoder to use
func (w *huffmanBitWriter) generateCodegen(numLiterals int, numOffsets int, litEnc, offEnc *huffmanEncoder) {
	for i := range w.codegenFreq {
		w.codegenFreq[i] = 0
	}
	// Note that we are using codegen both as a temporary variable for holding
	// a copy of the frequencies, and as the place where we put the result.
	// This is fine because the output is always shorter than the input used
	// so far.
	codegen := w.codegen[:] // cache
	// Copy the concatenated code sizes to codegen. Put a marker at the end.
	cgnl := codegen[:numLiterals]
	for i := range cgnl {
		cgnl[i] = uint8(litEnc.codes[i].len)
	}

	cgnl = codegen[numLiterals : numLiterals+numOffsets]
	for i := range cgnl {
		cgnl[i] = uint8(offEnc.codes[i].len)
	}
	codegen[numLiterals+numOffsets] = badCode

	size := codegen[0]
	count := 1
	outIndex := 0
	for inIndex := 1; size != badCode; inIndex++ {
		// INVARIANT: We have seen "count" copies of size that have not yet
		// had output generated for them.
		nextSize := codegen[inIndex]
		if nextSize == size {
			count++
			continue
		}
		// We need to generate codegen indicating "count" of size.
		if size != 0 {
			codegen[outIndex] = size
			outIndex++
			w.codegenFreq[size]++
			count--
			for count >= 3 {
				n := 6
				if n > count {
					n = count
				}
				codegen[outIndex] = 16
				outIndex++
				codegen[outIndex] = uint8(n - 3)
				outIndex++
				w.codegenFreq[16]++
				count -= n
			}
		} else {
			for count >= 11 {
				n := 138
				if n > count {
					n = count
				}
				codegen[outIndex] = 18
				outIndex++
				codegen[outIndex] = uint8(n - 11)
				outIndex++
				w.codegenFreq[18]++
				count -= n
			}
			if count >= 3 {
				// count >= 3 && count <= 10
				codegen[outIndex] = 17
				outIndex++
				codegen[outIndex] = uint8(count - 3)
				outIndex++
				w.codegenFreq[17]++
				count = 0
			}
		}
		count--
		for ; count >= 0; count-- {
			codegen[outIndex] = size
			outIndex++
			w.codegenFreq[size]++
		}
		// Set up invariant for next time through the loop.
		size = nextSize
		count = 1
	}
	// Marker indicating the end of the codegen.
	codegen[outIndex] = badCode
}

func (w *huffmanBitWriter) codegens() int {
	numCodegens := len(w.codegenFreq)
	for numCodegens > 4 && w.codegenFreq[codegenOrder[numCodegens-1]] == 0 {
		numCodegens--
	}
	return numCodegens
}

func (w *huffmanBitWriter) headerSize() (size, numCodegens int) {
	numCodegens = len(w.codegenFreq)
	for numCodegens > 4 && w.codegenFreq[codegenOrder[numCodegens-1]] == 0 {
		numCodegens--
	}
	return 3 + 5 + 5 + 4 + (3 * numCodegens) +
		w.codegenEncoding.bitLength(w.codegenFreq[:]) +
		int(w.codegenFreq[16])*2 +
		int(w.codegenFreq[17])*3 +
		int(w.codegenFreq[18])*7, numCodegens
}

// dynamicSize returns the size of dynamically encoded data in bits.
func (w *huffmanBitWriter) dynamicReuseSize(litEnc, offEnc *huffmanEncoder) (size int) {
	size = litEnc.bitLength(w.literalFreq[:]) +
		offEnc.bitLength(w.offsetFreq[:])
	return size
}

// dynamicSize returns the size of dynamically encoded data in bits.
func (w *huffmanBitWriter) dynamicSize(litEnc, offEnc *huffmanEncoder, extraBits int) (size, numCodegens int) {
	header, numCodegens := w.headerSize()
	size = header +
		litEnc.bitLength(w.literalFreq[:]) +
		offEnc.bitLength(w.offsetFreq[:]) +
		extraBits
	return size, numCodegens
}

// extraBitSize will return the number of bits that will be written
// as "extra" bits on matches.
func (w *huffmanBitWriter) extraBitSize() int {
	total := 0
	for i, n := range w.literalFreq[257:literalCount] {
		total += int(n) * int(lengthExtraBits[i&31])
	}
	for i, n := range w.offsetFreq[:offsetCodeCount] {
		total += int(n) * int(offsetExtraBits[i&31])
	}
	return total
}

// fixedSize returns the size of dynamically encoded data in bits.
func (w *huffmanBitWriter) fixedSize(extraBits int) int {
	return 3 +
		fixedLiteralEncoding.bitLength(w.literalFreq[:]) +
		fixedOffsetEncoding.bitLength(w.offsetFreq[:]) +
		extraBits
}

// storedSize calculates the stored size, including header.
// The function returns the size in bits and whether the block
// fits inside a single block.
func (w *huffmanBitWriter) storedSize(in []byte) (int, bool) {
	if in == nil {
		return 0, false
	}
	if len(in) <= maxStoreBlockSize {
		return (len(in) + 5) * 8, true
	}
	return 0, false
}

func (w *huffmanBitWriter) writeCode(c hcode) {
	// The function does not get inlined if we "& 63" the shift.
	w.bits |= uint64(c.code) << w.nbits
	w.nbits += c.len
	if w.nbits >= 48 {
		w.writeOutBits()
	}
}

// writeOutBits will write bits to the buffer.
func (w *huffmanBitWriter) writeOutBits() {
	bits := w.bits
	w.bits >>= 48
	w.nbits -= 48
	n := w.nbytes

	// We over-write, but faster...
	binary.LittleEndian.PutUint64(w.bytes[n:], bits)
	n += 6

	if n >= bufferFlushSize {
		if w.err != nil {
			n = 0
			return
		}
		w.write(w.bytes[:n])
		n = 0
	}

	w.nbytes = n
}

// Write the header of a dynamic Huffman block to the output stream.
//
//  numLiterals  The number of literals specified in codegen
//  numOffsets   The number of offsets specified in codegen
//  numCodegens  The number of codegens used in codegen
func (w *huffmanBitWriter) writeDynamicHeader(numLiterals int, numOffsets int, numCodegens int, isEof bool) {
	if w.err != nil {
		return
	}
	var firstBits int32 = 4
	if isEof {
		firstBits = 5
	}
	w.writeBits(firstBits, 3)
	w.writeBits(int32(numLiterals-257), 5)
	w.writeBits(int32(numOffsets-1), 5)
	w.writeBits(int32(numCodegens-4), 4)

	for i := 0; i < numCodegens; i++ {
		value := uint(w.codegenEncoding.codes[codegenOrder[i]].len)
		w.writeBits(int32(value), 3)
	}

	i := 0
	for {
		var codeWord = uint32(w.codegen[i])
		i++
		if codeWord == badCode {
			break
		}
		w.writeCode(w.codegenEncoding.codes[codeWord])

		switch codeWord {
		case 16:
			w.writeBits(int32(w.codegen[i]), 2)
			i++
		case 17:
			w.writeBits(int32(w.codegen[i]), 3)
			i++
		case 18:
			w.writeBits(int32(w.codegen[i]), 7)
			i++
		}
	}
}

// writeStoredHeader will write a stored header.
// If the stored block is only used for EOF,
// it is replaced with a fixed huffman block.
func (w *huffmanBitWriter) writeStoredHeader(length int, isEof bool) {
	if w.err != nil {
		return
	}
	if w.lastHeader > 0 {
		// We owe an EOB
		w.writeCode(w.literalEncoding.codes[endBlockMarker])
		w.lastHeader = 0
	}

	// To write EOF, use a fixed encoding block. 10 bits instead of 5 bytes.
	if length == 0 && isEof {
		w.writeFixedHeader(isEof)
		// EOB: 7 bits, value: 0
		w.writeBits(0, 7)
		w.flush()
		return
	}

	var flag int32
	if isEof {
		flag = 1
	}
	w.writeBits(flag, 3)
	w.flush()
	w.writeBits(int32(length), 16)
	w.writeBits(int32(^uint16(length)), 16)
}

func (w *huffmanBitWriter) writeFixedHeader(isEof bool) {
	if w.err != nil {
		return
	}
	if w.lastHeader > 0 {
		// We owe an EOB
		w.writeCode(w.literalEncoding.codes[endBlockMarker])
		w.lastHeader = 0
	}

	// Indicate that we are a fixed Huffman block
	var value int32 = 2
	if isEof {
		value = 3
	}
	w.writeBits(value, 3)
}

// writeBlock will write a block of tokens with the smallest encoding.
// The original input can be supplied, and if the huffman encoded data
// is larger than the original bytes, the data will be written as a
// stored block.
// If the input is nil, the tokens will always be Huffman encoded.
func (w *huffmanBitWriter) writeBlock(tokens *tokens, eof bool, input []byte) {
	if w.err != nil {
		return
	}

	tokens.AddEOB()
	if w.lastHeader > 0 {
		// We owe an EOB
		w.writeCode(w.literalEncoding.codes[endBlockMarker])
		w.lastHeader = 0
	}
	numLiterals, numOffsets := w.indexTokens(tokens, false)
	w.generate(tokens)
	var extraBits int
	storedSize, storable := w.storedSize(input)
	if storable {
		extraBits = w.extraBitSize()
	}

	// Figure out smallest code.
	// Fixed Huffman baseline.
	var literalEncoding = fixedLiteralEncoding
	var offsetEncoding = fixedOffsetEncoding
	var size = w.fixedSize(extraBits)

	// Dynamic Huffman?
	var numCodegens int

	// Generate codegen and codegenFrequencies, which indicates how to encode
	// the literalEncoding and the offsetEncoding.
	w.generateCodegen(numLiterals, numOffsets, w.literalEncoding, w.offsetEncoding)
	w.codegenEncoding.generate(w.codegenFreq[:], 7)
	dynamicSize, numCodegens := w.dynamicSize(w.literalEncoding, w.offsetEncoding, extraBits)

	if dynamicSize < size {
		size = dynamicSize
		literalEncoding = w.literalEncoding
		offsetEncoding = w.offsetEncoding
	}

	// Stored bytes?
	if storable && storedSize < size {
		w.writeStoredHeader(len(input), eof)
		w.writeBytes(input)
		return
	}

	// Huffman.
	if literalEncoding == fixedLiteralEncoding {
		w.writeFixedHeader(eof)
	} else {
		w.writeDynamicHeader(numLiterals, numOffsets, numCodegens, eof)
	}

	// Write the tokens.
	w.writeTokens(tokens.Slice(), literalEncoding.codes, offsetEncoding.codes)
}

// writeBlockDynamic encodes a block using a dynamic Huffman table.
// This should be used if the symbols used have a disproportionate
// histogram distribution.
// If input is supplied and the compression savings are below 1/16th of the
// input size the block is stored.
func (w *huffmanBitWriter) writeBlockDynamic(tokens *tokens, eof bool, input []byte, sync bool) {
	if w.err != nil {
		return
	}

	sync = sync || eof
	if sync {
		tokens.AddEOB()
	}

	// We cannot reuse pure huffman table, and must mark as EOF.
	if (w.lastHuffMan || eof) && w.lastHeader > 0 {
		// We will not try to reuse.
		w.writeCode(w.literalEncoding.codes[endBlockMarker])
		w.lastHeader = 0
		w.lastHuffMan = false
	}
	if !sync {
		tokens.Fill()
	}
	numLiterals, numOffsets := w.indexTokens(tokens, !sync)

	var size int
	// Check if we should reuse.
	if w.lastHeader > 0 {
		// Estimate size for using a new table.
		// Use the previous header size as the best estimate.
		newSize := w.lastHeader + tokens.EstimatedBits()
		newSize += newSize >> w.logNewTablePenalty

		// The estimated size is calculated as an optimal table.
		// We add a penalty to make it more realistic and re-use a bit more.
		reuseSize := w.dynamicReuseSize(w.literalEncoding, w.offsetEncoding) + w.extraBitSize()

		// Check if a new table is better.
		if newSize < reuseSize {
			// Write the EOB we owe.
			w.writeCode(w.literalEncoding.codes[endBlockMarker])
			size = newSize
			w.lastHeader = 0
		} else {
			size = reuseSize
		}
		// Check if we get a reasonable size decrease.
		if ssize, storable := w.storedSize(input); storable && ssize < (size+size>>4) {
			w.writeStoredHeader(len(input), eof)
			w.writeBytes(input)
			w.lastHeader = 0
			return
		}
	}

	// We want a new block/table
	if w.lastHeader == 0 {
		w.generate(tokens)
		// Generate codegen and codegenFrequencies, which indicates how to encode
		// the literalEncoding and the offsetEncoding.
		w.generateCodegen(numLiterals, numOffsets, w.literalEncoding, w.offsetEncoding)
		w.codegenEncoding.generate(w.codegenFreq[:], 7)
		var numCodegens int
		size, numCodegens = w.dynamicSize(w.literalEncoding, w.offsetEncoding, w.extraBitSize())
		// Store bytes, if we don't get a reasonable improvement.
		if ssize, storable := w.storedSize(input); storable && ssize < (size+size>>4) {
			w.writeStoredHeader(len(input), eof)
			w.writeBytes(input)
			w.lastHeader = 0
			return
		}

		// Write Huffman table.
		w.writeDynamicHeader(numLiterals, numOffsets, numCodegens, eof)
		w.lastHeader, _ = w.headerSize()
		w.lastHuffMan = false
	}

	if sync {
		w.lastHeader = 0
	}
	// Write the tokens.
	w.writeTokens(tokens.Slice(), w.literalEncoding.codes, w.offsetEncoding.codes)
}

// indexTokens indexes a slice of tokens, and updates
// literalFreq and offsetFreq, and generates literalEncoding
// and offsetEncoding.
// The number of literal and offset tokens is returned.
func (w *huffmanBitWriter) indexTokens(t *tokens, filled bool) (numLiterals, numOffsets int) {
	copy(w.literalFreq[:], t.litHist[:])
	copy(w.literalFreq[256:], t.extraHist[:])
	copy(w.offsetFreq[:], t.offHist[:offsetCodeCount])

	if t.n == 0 {
		return
	}
	if filled {
		return maxNumLit, maxNumDist
	}
	// get the number of literals
	numLiterals = len(w.literalFreq)
	for w.literalFreq[numLiterals-1] == 0 {
		numLiterals--
	}
	// get the number of offsets
	numOffsets = len(w.offsetFreq)
	for numOffsets > 0 && w.offsetFreq[numOffsets-1] == 0 {
		numOffsets--
	}
	if numOffsets == 0 {
		// We haven't found a single match. If we want to go with the dynamic encoding,
		// we should count at least one offset to be sure that the offset huffman tree could be encoded.
		w.offsetFreq[0] = 1
		numOffsets = 1
	}
	return
}

func (w *huffmanBitWriter) generate(t *tokens) {
	w.literalEncoding.generate(w.literalFreq[:literalCount], 15)
	w.offsetEncoding.generate(w.offsetFreq[:offsetCodeCount], 15)
}

// writeTokens writes a slice of tokens to the output.
// codes for literal and offset encoding must be supplied.
func (w *huffmanBitWriter) writeTokens(tokens []token, leCodes, oeCodes []hcode) {
	if w.err != nil {
		return
	}
	if len(tokens) == 0 {
		return
	}

	// Only last token should be endBlockMarker.
	var deferEOB bool
	if tokens[len(tokens)-1] == endBlockMarker {
		tokens = tokens[:len(tokens)-1]
		deferEOB = true
	}

	// Create slices up to the next power of two to avoid bounds checks.
	lits := leCodes[:256]
	offs := oeCodes[:32]
	lengths := leCodes[lengthCodesStart:]
	lengths = lengths[:32]

	// Go 1.16 LOVES having these on stack.
	bits, nbits, nbytes := w.bits, w.nbits, w.nbytes

	for _, t := range tokens {
		if t < matchType {
			//w.writeCode(lits[t.literal()])
			c := lits[t.literal()]
			bits |= uint64(c.code) << nbits
			nbits += c.len
			if nbits >= 48 {
				binary.LittleEndian.PutUint64(w.bytes[nbytes:], bits)
				//*(*uint64)(unsafe.Pointer(&w.bytes[nbytes])) = bits
				bits >>= 48
				nbits -= 48
				nbytes += 6
				if nbytes >= bufferFlushSize {
					if w.err != nil {
						nbytes = 0
						return
					}
					_, w.err = w.writer.Write(w.bytes[:nbytes])
					nbytes = 0
				}
			}
			continue
		}

		// Write the length
		length := t.length()
		lengthCode := lengthCode(length)
		if false {
			w.writeCode(lengths[lengthCode&31])
		} else {
			// inlined
			c := lengths[lengthCode&31]
			bits |= uint64(c.code) << nbits
			nbits += c.len
			if nbits >= 48 {
				binary.LittleEndian.PutUint64(w.bytes[nbytes:], bits)
				//*(*uint64)(unsafe.Pointer(&w.bytes[nbytes])) = bits
				bits >>= 48
				nbits -= 48
				nbytes += 6
				if nbytes >= bufferFlushSize {
					if w.err != nil {
						nbytes = 0
						return
					}
					_, w.err = w.writer.Write(w.bytes[:nbytes])
					nbytes = 0
				}
			}
		}

		extraLengthBits := uint16(lengthExtraBits[lengthCode&31])
		if extraLengthBits > 0 {
			//w.writeBits(extraLength, extraLengthBits)
			extraLength := int32(length - lengthBase[lengthCode&31])
			bits |= uint64(extraLength) << nbits
			nbits += extraLengthBits
			if nbits >= 48 {
				binary.LittleEndian.PutUint64(w.bytes[nbytes:], bits)
				//*(*uint64)(unsafe.Pointer(&w.bytes[nbytes])) = bits
				bits >>= 48
				nbits -= 48
				nbytes += 6
				if nbytes >= bufferFlushSize {
					if w.err != nil {
						nbytes = 0
						return
					}
					_, w.err = w.writer.Write(w.bytes[:nbytes])
					nbytes = 0
				}
			}
		}
		// Write the offset
		offset := t.offset()
		offsetCode := offset >> 16
		offset &= matchOffsetOnlyMask
		if false {
			w.writeCode(offs[offsetCode&31])
		} else {
			// inlined
			c := offs[offsetCode]
			bits |= uint64(c.code) << nbits
			nbits += c.len
			if nbits >= 48 {
				binary.LittleEndian.PutUint64(w.bytes[nbytes:], bits)
				//*(*uint64)(unsafe.Pointer(&w.bytes[nbytes])) = bits
				bits >>= 48
				nbits -= 48
				nbytes += 6
				if nbytes >= bufferFlushSize {
					if w.err != nil {
						nbytes = 0
						return
					}
					_, w.err = w.writer.Write(w.bytes[:nbytes])
					nbytes = 0
				}
			}
		}
		offsetComb := offsetCombined[offsetCode]
		if offsetComb > 1<<16 {
			//w.writeBits(extraOffset, extraOffsetBits)
			bits |= uint64(offset&matchOffsetOnlyMask-(offsetComb&0xffff)) << nbits
			nbits += uint16(offsetComb >> 16)
			if nbits >= 48 {
				binary.LittleEndian.PutUint64(w.bytes[nbytes:], bits)
				//*(*uint64)(unsafe.Pointer(&w.bytes[nbytes])) = bits
				bits >>= 48
				nbits -= 48
				nbytes += 6
				if nbytes >= bufferFlushSize {
					if w.err != nil {
						nbytes = 0
						return
					}
					_, w.err = w.writer.Write(w.bytes[:nbytes])
					nbytes = 0
				}
			}
		}
	}
	// Restore...
	w.bits, w.nbits, w.nbytes = bits, nbits, nbytes

	if deferEOB {
		w.writeCode(leCodes[endBlockMarker])
	}
}

// huffOffset is a static offset encoder used for huffman only encoding.
// It can be reused since we will not be encoding offset values.
var huffOffset *huffmanEncoder

func init() {
	w := newHuffmanBitWriter(nil)
	w.offsetFreq[0] = 1
	huffOffset = newHuffmanEncoder(offsetCodeCount)
	huffOffset.generate(w.offsetFreq[:offsetCodeCount], 15)
}

// writeBlockHuff encodes a block of bytes as either
// Huffman encoded literals or uncompressed bytes if the
// results only gains very little from compression.
func (w *huffmanBitWriter) writeBlockHuff(eof bool, input []byte, sync bool) {
	if w.err != nil {
		return
	}

	// Clear histogram
	for i := range w.literalFreq[:] {
		w.literalFreq[i] = 0
	}
	if !w.lastHuffMan {
		for i := range w.offsetFreq[:] {
			w.offsetFreq[i] = 0
		}
	}

	// Fill is rarely better...
	const fill = false
	const numLiterals = endBlockMarker + 1
	const numOffsets = 1

	// Add everything as literals
	// We have to estimate the header size.
	// Assume header is around 70 bytes:
	// https://stackoverflow.com/a/25454430
	const guessHeaderSizeBits = 70 * 8
	histogram(input, w.literalFreq[:numLiterals], fill)
	w.literalFreq[endBlockMarker] = 1
	w.tmpLitEncoding.generate(w.literalFreq[:numLiterals], 15)
	if fill {
		// Clear fill...
		for i := range w.literalFreq[:numLiterals] {
			w.literalFreq[i] = 0
		}
		histogram(input, w.literalFreq[:numLiterals], false)
	}
	estBits := w.tmpLitEncoding.canReuseBits(w.literalFreq[:numLiterals])
	estBits += w.lastHeader
	if w.lastHeader == 0 {
		estBits += guessHeaderSizeBits
	}
	estBits += estBits >> w.logNewTablePenalty

	// Store bytes, if we don't get a reasonable improvement.
	ssize, storable := w.storedSize(input)
	if storable && ssize <= estBits {
		w.writeStoredHeader(len(input), eof)
		w.writeBytes(input)
		return
	}

	if w.lastHeader > 0 {
		reuseSize := w.literalEncoding.canReuseBits(w.literalFreq[:256])

		if estBits < reuseSize {
			if debugDeflate {
				//fmt.Println("not reusing, reuse:", reuseSize/8, "> new:", estBits/8, "- header est:", w.lastHeader/8)
			}
			// We owe an EOB
			w.writeCode(w.literalEncoding.codes[endBlockMarker])
			w.lastHeader = 0
		} else if debugDeflate {
			fmt.Println("reusing, reuse:", reuseSize/8, "> new:", estBits/8, "- header est:", w.lastHeader/8)
		}
	}

	count := 0
	if w.lastHeader == 0 {
		// Use the temp encoding, so swap.
		w.literalEncoding, w.tmpLitEncoding = w.tmpLitEncoding, w.literalEncoding
		// Generate codegen and codegenFrequencies, which indicates how to encode
		// the literalEncoding and the offsetEncoding.
		w.generateCodegen(numLiterals, numOffsets, w.literalEncoding, huffOffset)
		w.codegenEncoding.generate(w.codegenFreq[:], 7)
		numCodegens := w.codegens()

		// Huffman.
		w.writeDynamicHeader(numLiterals, numOffsets, numCodegens, eof)
		w.lastHuffMan = true
		w.lastHeader, _ = w.headerSize()
		if debugDeflate {
			count += w.lastHeader
			fmt.Println("header:", count/8)
		}
	}

	encoding := w.literalEncoding.codes[:256]
	// Go 1.16 LOVES having these on stack. At least 1.5x the speed.
	bits, nbits, nbytes := w.bits, w.nbits, w.nbytes
	for _, t := range input {
		// Bitwriting inlined, ~30% speedup
		c := encoding[t]
		bits |= uint64(c.code) << nbits
		nbits += c.len
		if debugDeflate {
			count += int(c.len)
		}
		if nbits >= 48 {
			binary.LittleEndian.PutUint64(w.bytes[nbytes:], bits)
			//*(*uint64)(unsafe.Pointer(&w.bytes[nbytes])) = bits
			bits >>= 48
			nbits -= 48
			nbytes += 6
			if nbytes >= bufferFlushSize {
				if w.err != nil {
					nbytes = 0
					return
				}
				_, w.err = w.writer.Write(w.bytes[:nbytes])
				nbytes = 0
			}
		}
	}
	// Restore...
	w.bits, w.nbits, w.nbytes = bits, nbits, nbytes

	if debugDeflate {
		fmt.Println("wrote", count/8, "bytes")
	}
	if eof || sync {
		w.writeCode(w.literalEncoding.codes[endBlockMarker])
		w.lastHeader = 0
		w.lastHuffMan = false
	}
}