Check-in [6d10a1d28f]
Overview
Comment:Integrate the decompressor2 branch into trunk now that it is faster.
Downloads: Tarball | ZIP archive | SQL archive
Timelines: family | ancestors | descendants | both | trunk
Files: files | file ages | folders
SHA1: 6d10a1d28facd4fe7120f66a985041bd2fe38f0e
User & Date: spaskalev on 2014-12-22 16:41:58
Other Links: manifest | tags
Context
2014-12-22
17:15
Calculate the decompressed block length outside of the predictor loop check-in: 27ecac81d3 user: spaskalev tags: trunk
16:41
Integrate the decompressor2 branch into trunk now that it is faster. check-in: 6d10a1d28f user: spaskalev tags: trunk
16:35
Decompressor - try to fill as much as possible in the output buffer in a single pass. Closed-Leaf check-in: 42ba1f458d user: spaskalev tags: decompressor2
2014-12-21
23:24
The decompressor now tries to fill in the whole output buffer in a single call. check-in: 1f73d90f65 user: spaskalev tags: trunk
Changes

Modified src/0dev.org/predictor/predictor.go from [9dfe9174f8] to [71e92568a2].

1
2
3
4
5

6
7
8
9
10
11
12
// Package predictor implements the predictor compression/decompression algorithm
// as specified by RFC1978 - PPP Predictor Compression Protocol
package predictor

import (

	"io"
)

type context struct {
	table [1 << 16]byte
	input []byte
	hash  uint16





>







1
2
3
4
5
6
7
8
9
10
11
12
13
// Package predictor implements the predictor compression/decompression algorithm
// as specified by RFC1978 - PPP Predictor Compression Protocol
package predictor

import (
	bits "0dev.org/bits"
	"io"
)

type context struct {
	table [1 << 16]byte
	input []byte
	hash  uint16
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
// and decompresses data according to the predictor algorithm
func Decompressor(reader io.Reader) io.Reader {
	var ctx context
	ctx.input = make([]byte, 0, 8)

	return decompressor(func(output []byte) (int, error) {
		var (
			err          error
			flags        byte
			i, rc, total int
		)

		// Sanity check for space to read into
		if len(output) == 0 {
			return 0, nil
		}

		// Check whether we have leftover data in the buffer
		if len(ctx.input) > 0 {
			rc = copy(output, ctx.input)

			// Check whether we still have leftover data in the buffer :)
			if rc < len(ctx.input) {
				ctx.input = ctx.input[:copy(ctx.input, ctx.input[rc:])]
			}
			return rc, nil
		}

	loop:
		// Read the flags

		rc, err = reader.Read(ctx.input[:1])

		if err != nil && err != io.EOF {
			return 0, err
		}
		if rc == 0 {
			return total, err
		}



		ctx.input = ctx.input[:8]
		flags = ctx.input[0]



		for i = 0; i < 8; i++ {
			if flags&(1<<uint(i)) > 0 {
				// Guess was right
				ctx.input[i] = ctx.table[ctx.hash]
			} else {
				rc, err = reader.Read(ctx.input[i:(i + 1)])




				if err == io.EOF {

					break

				}



				if err != nil {
					return total, err
				}

				if rc == 0 { // treat as EoF
					break
				}



				ctx.table[ctx.hash] = ctx.input[i]
			}

			ctx.hash = (ctx.hash << 4) ^ uint16(ctx.input[i])
		}



		rc = copy(output, ctx.input[:i])

		total += rc

		// Place any remaining bytes in the buffer
		if rc < i {
			ctx.input = ctx.input[:copy(ctx.input, ctx.input[rc:i])]
		} else {
			// Clear the buffer
			ctx.input = ctx.input[:0]

			// Advance the output buffer ...
			output = output[i:]
			// ... and decompress the next block if there is any space left
			if len(output) > 0 && err != io.EOF {
				goto loop
			}
		}

		return total, err
	})
}







|
|
|


















<
|
>

>

|
<
|



>
>


>
>

<
<
<
<
|
|
|
>
>
>
|
>
|
>
|
|
>
>
|
<
<
|
|
|
|
|
>
>


|



>
>
|
>
|

|
|
|




<
|
<
|
|






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
// and decompresses data according to the predictor algorithm
func Decompressor(reader io.Reader) io.Reader {
	var ctx context
	ctx.input = make([]byte, 0, 8)

	return decompressor(func(output []byte) (int, error) {
		var (
			err                             error
			flags                           byte
			rc, available, predicted, total int
		)

		// Sanity check for space to read into
		if len(output) == 0 {
			return 0, nil
		}

		// Check whether we have leftover data in the buffer
		if len(ctx.input) > 0 {
			rc = copy(output, ctx.input)

			// Check whether we still have leftover data in the buffer :)
			if rc < len(ctx.input) {
				ctx.input = ctx.input[:copy(ctx.input, ctx.input[rc:])]
			}
			return rc, nil
		}


		// Read the next prediction header
	readHeader:
		rc, err = reader.Read(ctx.input[:1])
		// Fail on error unless it is EOF
		if err != nil && err != io.EOF {
			return total, err

		} else if rc == 0 {
			return total, err
		}

		// Extend the buffer, copy the prediction header
		//  and calculate the number of subsequent bytes to read
		ctx.input = ctx.input[:8]
		flags = ctx.input[0]
		predicted = int(bits.Hamming(flags))
		available = 8 - predicted





		// Read the non-predicted bytes and place them in the end of the buffer
		rc, err = reader.Read(ctx.input[predicted:])
	retryData:
		if rc < int(available) && err == nil {
			// Retry the read if we have fewer bytes than what the prediction header indicates
			var r int
			r, err = reader.Read(ctx.input[predicted+rc:])
			rc += r
			goto retryData
		} // Continue on any error, try to decompress and return it along the result

		// Walk the buffer, filling in the predicted blanks,
		// relocating read bytes and and updating the guess table
		for i, a := uint(0), predicted; i < 8; i++ {
			if (flags & (1 << i)) > 0 {


				// Guess succeeded, fill in from the table
				ctx.input[i] = ctx.table[ctx.hash]
				rc++
			} else {
				// Relocate a read byte
				ctx.input[i], a = ctx.input[a], a+1
				// Guess failed, update the table
				ctx.table[ctx.hash] = ctx.input[i]
			}
			// Update the hash
			ctx.hash = (ctx.hash << 4) ^ uint16(ctx.input[i])
		}

		// rc now contains the precise amount of populated data
		ctx.input = ctx.input[:rc]
		available = copy(output, ctx.input)

		total += available

		// Check for remaining bytes that dont fit in the output buffer
		if available < rc {
			ctx.input = ctx.input[:copy(ctx.input, ctx.input[available:])]
		} else {
			// Clear the buffer
			ctx.input = ctx.input[:0]


			output = output[available:]

			if len(output) > 0 && err == nil {
				goto readHeader
			}
		}

		return total, err
	})
}