-
Notifications
You must be signed in to change notification settings - Fork 0
/
rdiff.go
229 lines (200 loc) · 6.39 KB
/
rdiff.go
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
package rdiff
import (
"bytes"
"hash"
"io"
"slices"
)
// OpType represents a block operation/instruction, useful to recompute the target, based on source.
type OpType byte
const (
// OpBlockKeep means a block is unchanged and should be kept as it is.
OpBlockKeep OpType = iota
// OpBlockUpdate means there is a match in the target, and it contains extra data(literal) compared to the initial block
OpBlockUpdate
// OpBlockRemove means there is no match for a target block in the source
OpBlockRemove
// OpBlockNew means there is a literal block in the source that doesn't have any match in the target - new data
OpBlockNew
)
// Block represents a chunk of data(bytes) used by the target to split its data.
type Block struct {
StrongHash []byte
WeakHash uint32
}
// Operation represents an instruction given by the source to the target, in order to allow the target to update its content.
type Operation struct {
Type OpType
// the index of the block from the target, for OpBlockNew -1 is used to enforce that the BlockIndex
// is not important in this case
BlockIndex int
// additional literal data, if the block was modified, or a new block if the Block was not matched (BlockIndex == 0)
Data []byte
}
// blockData is used to compute the block search list(map[uint32][]blockData)
type blockData struct {
strongHash []byte
blockIndex int
}
type rDiff struct {
blockSize int
weakHasher *adler32RollingHash
strongHasher hash.Hash
}
func newRDiff(blockSize int, weakHasher *adler32RollingHash, strongHasher hash.Hash) *rDiff {
return &rDiff{
blockSize: blockSize,
weakHasher: weakHasher,
strongHasher: strongHasher,
}
}
// ComputeSignature computes the signature of a target and returns a []Block, based on the blockSize.
// Every Block contains the weak hash and strong hash.
// It returns a non-nil error in case target encounters a reading error, other than io.EOF.
func (r *rDiff) ComputeSignature(target io.Reader) ([]Block, error) {
var output []Block
block := make([]byte, r.blockSize)
// it's enough a single Reset call, as the WriteAll method acts like a Reset and Write.
r.weakHasher.Reset()
for {
n, err := target.Read(block)
if n == 0 && err == io.EOF {
break
}
if err != nil && err != io.EOF {
return output, err
}
block = block[:n]
r.strongHasher.Reset()
_, _ = r.strongHasher.Write(block)
// it doesn't need reset, as it's always rewriting the digest
r.weakHasher.WriteAll(block)
bl := Block{
StrongHash: r.strongHasher.Sum(nil),
WeakHash: r.weakHasher.Sum32(),
}
output = append(output, bl)
}
return output, nil
}
// ComputeDelta computes the instruction list(operations list) based on the target's blockList
// to be able to update its content to match the source.
func (r *rDiff) ComputeDelta(source io.Reader, blockList []Block) ([]Operation, error) {
tempDelta := make(map[int]Operation, len(blockList))
searchList := computeSearchList(blockList)
block := make([]byte, r.blockSize)
var literal []byte
// it's enough a single Reset call, as the WriteAll method acts like a Reset and Write.
r.weakHasher.Reset()
rolling := false
for {
n, err := r.read(source, block, rolling)
if n == 0 && err == io.EOF {
break
}
if err != nil && err != io.EOF {
return nil, err
}
block = block[:n]
if !rolling {
r.weakHasher.WriteAll(block)
} else {
oldest := r.weakHasher.Roll(block[0])
literal = append(literal, oldest)
}
if blIdx := r.searchBlock(searchList, r.weakHasher.Sum32()); blIdx != -1 {
rolling = false
tempDelta[blIdx] = createOperation(blIdx, literal)
literal = literal[:0]
continue
}
rolling = true
}
r.updateDeltaWithLiteralBlockOperation(tempDelta, rolling, literal)
return computeFinalDelta(blockList, tempDelta), nil
}
func (r *rDiff) updateDeltaWithLiteralBlockOperation(delta map[int]Operation, rolling bool, literal []byte) {
// the last read block will not be added to the delta if it was not matched in the target,
// so we need to add it to the literal collection
if rolling {
l := r.weakHasher.GetWindowContent()
literal = append(literal, l...)
}
// collecting leftovers literals into a single new block
if len(literal) > 0 {
op := Operation{
Type: OpBlockNew,
BlockIndex: -1,
}
op.Data = append(op.Data, literal...)
delta[-1] = op
}
}
func (r *rDiff) read(reader io.Reader, block []byte, rolling bool) (int, error) {
// adjusting reading size to block of bytes or single byte
// after a found match in the target, we need to read up to a full block
// if rolling is in place, then we read up to a single byte
if !rolling {
block = block[:r.blockSize]
} else {
block = block[:1]
}
return reader.Read(block)
}
func (r *rDiff) searchBlock(searchList map[uint32][]blockData, weakHash uint32) int {
if bl, found := searchList[weakHash]; found {
r.strongHasher.Reset()
currBlockContent := r.weakHasher.GetWindowContent()
// nolint
r.strongHasher.Write(currBlockContent)
strongHash := r.strongHasher.Sum(nil)
blFoundIdx := slices.IndexFunc(bl, func(el blockData) bool { return bytes.Equal(el.strongHash, strongHash) })
if blFoundIdx != -1 {
blockIndex := bl[blFoundIdx].blockIndex
//remove the strong hash from the list, because if we have identical blocks in the target,
//then we'll always match the same block
searchList[weakHash] = slices.Delete(bl, blFoundIdx, blFoundIdx+1)
return blockIndex
}
}
return -1
}
func createOperation(index int, lit []byte) Operation {
opType := OpBlockKeep
if len(lit) > 0 {
opType = OpBlockUpdate
}
op := Operation{
Type: opType,
BlockIndex: index,
}
op.Data = append(op.Data, lit...)
return op
}
func computeSearchList(blockList []Block) map[uint32][]blockData {
sl := make(map[uint32][]blockData, len(blockList))
for i, block := range blockList {
sl[block.WeakHash] = append(sl[block.WeakHash], blockData{strongHash: block.StrongHash, blockIndex: i})
}
return sl
}
func computeFinalDelta(target []Block, delta map[int]Operation) []Operation {
// len(target)+1 is used to cover the max possible size: all target blocks + 1 extra literal block(if any)
output := make([]Operation, 0, len(target)+1)
for i := range target {
op, ok := delta[i]
if !ok {
removed := Operation{
Type: OpBlockRemove,
BlockIndex: i,
}
output = append(output, removed)
continue
}
output = append(output, op)
}
if extra, ok := delta[-1]; ok {
output = append(output, extra)
}
return output
}