aboutsummaryrefslogtreecommitdiff
path: root/vendor/github.com/ulikunitz/xz/lzma/rangecodec.go
blob: 7b299abfefd9715a20b9a9665246e200fd751a58 (plain)
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
// Copyright 2014-2019 Ulrich Kunitz. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.

package lzma

import (
	"errors"
	"io"
)

// rangeEncoder implements range encoding of single bits. The low value can
// overflow therefore we need uint64. The cache value is used to handle
// overflows.
type rangeEncoder struct {
	lbw      *LimitedByteWriter
	nrange   uint32
	low      uint64
	cacheLen int64
	cache    byte
}

// maxInt64 provides the  maximal value of the int64 type
const maxInt64 = 1<<63 - 1

// newRangeEncoder creates a new range encoder.
func newRangeEncoder(bw io.ByteWriter) (re *rangeEncoder, err error) {
	lbw, ok := bw.(*LimitedByteWriter)
	if !ok {
		lbw = &LimitedByteWriter{BW: bw, N: maxInt64}
	}
	return &rangeEncoder{
		lbw:      lbw,
		nrange:   0xffffffff,
		cacheLen: 1}, nil
}

// Available returns the number of bytes that still can be written. The
// method takes the bytes that will be currently written by Close into
// account.
func (e *rangeEncoder) Available() int64 {
	return e.lbw.N - (e.cacheLen + 4)
}

// writeByte writes a single byte to the underlying writer. An error is
// returned if the limit is reached. The written byte will be counted if
// the underlying writer doesn't return an error.
func (e *rangeEncoder) writeByte(c byte) error {
	if e.Available() < 1 {
		return ErrLimit
	}
	return e.lbw.WriteByte(c)
}

// DirectEncodeBit encodes the least-significant bit of b with probability 1/2.
func (e *rangeEncoder) DirectEncodeBit(b uint32) error {
	e.nrange >>= 1
	e.low += uint64(e.nrange) & (0 - (uint64(b) & 1))

	// normalize
	const top = 1 << 24
	if e.nrange >= top {
		return nil
	}
	e.nrange <<= 8
	return e.shiftLow()
}

// EncodeBit encodes the least significant bit of b. The p value will be
// updated by the function depending on the bit encoded.
func (e *rangeEncoder) EncodeBit(b uint32, p *prob) error {
	bound := p.bound(e.nrange)
	if b&1 == 0 {
		e.nrange = bound
		p.inc()
	} else {
		e.low += uint64(bound)
		e.nrange -= bound
		p.dec()
	}

	// normalize
	const top = 1 << 24
	if e.nrange >= top {
		return nil
	}
	e.nrange <<= 8
	return e.shiftLow()
}

// Close writes a complete copy of the low value.
func (e *rangeEncoder) Close() error {
	for i := 0; i < 5; i++ {
		if err := e.shiftLow(); err != nil {
			return err
		}
	}
	return nil
}

// shiftLow shifts the low value for 8 bit. The shifted byte is written into
// the byte writer. The cache value is used to handle overflows.
func (e *rangeEncoder) shiftLow() error {
	if uint32(e.low) < 0xff000000 || (e.low>>32) != 0 {
		tmp := e.cache
		for {
			err := e.writeByte(tmp + byte(e.low>>32))
			if err != nil {
				return err
			}
			tmp = 0xff
			e.cacheLen--
			if e.cacheLen <= 0 {
				if e.cacheLen < 0 {
					panic("negative cacheLen")
				}
				break
			}
		}
		e.cache = byte(uint32(e.low) >> 24)
	}
	e.cacheLen++
	e.low = uint64(uint32(e.low) << 8)
	return nil
}

// rangeDecoder decodes single bits of the range encoding stream.
type rangeDecoder struct {
	br     io.ByteReader
	nrange uint32
	code   uint32
}

// newRangeDecoder initializes a range decoder. It reads five bytes from the
// reader and therefore may return an error.
func newRangeDecoder(br io.ByteReader) (d *rangeDecoder, err error) {
	d = &rangeDecoder{br: br, nrange: 0xffffffff}

	b, err := d.br.ReadByte()
	if err != nil {
		return nil, err
	}
	if b != 0 {
		return nil, errors.New("newRangeDecoder: first byte not zero")
	}

	for i := 0; i < 4; i++ {
		if err = d.updateCode(); err != nil {
			return nil, err
		}
	}

	if d.code >= d.nrange {
		return nil, errors.New("newRangeDecoder: d.code >= d.nrange")
	}

	return d, nil
}

// possiblyAtEnd checks whether the decoder may be at the end of the stream.
func (d *rangeDecoder) possiblyAtEnd() bool {
	return d.code == 0
}

// DirectDecodeBit decodes a bit with probability 1/2. The return value b will
// contain the bit at the least-significant position. All other bits will be
// zero.
func (d *rangeDecoder) DirectDecodeBit() (b uint32, err error) {
	d.nrange >>= 1
	d.code -= d.nrange
	t := 0 - (d.code >> 31)
	d.code += d.nrange & t
	b = (t + 1) & 1

	// d.code will stay less then d.nrange

	// normalize
	// assume d.code < d.nrange
	const top = 1 << 24
	if d.nrange >= top {
		return b, nil
	}
	d.nrange <<= 8
	// d.code < d.nrange will be maintained
	return b, d.updateCode()
}

// decodeBit decodes a single bit. The bit will be returned at the
// least-significant position. All other bits will be zero. The probability
// value will be updated.
func (d *rangeDecoder) DecodeBit(p *prob) (b uint32, err error) {
	bound := p.bound(d.nrange)
	if d.code < bound {
		d.nrange = bound
		p.inc()
		b = 0
	} else {
		d.code -= bound
		d.nrange -= bound
		p.dec()
		b = 1
	}
	// normalize
	// assume d.code < d.nrange
	const top = 1 << 24
	if d.nrange >= top {
		return b, nil
	}
	d.nrange <<= 8
	// d.code < d.nrange will be maintained
	return b, d.updateCode()
}

// updateCode reads a new byte into the code.
func (d *rangeDecoder) updateCode() error {
	b, err := d.br.ReadByte()
	if err != nil {
		return err
	}
	d.code = (d.code << 8) | uint32(b)
	return nil
}