Merge pull request #14427 from zsfelfoldi/compress
common/bitutil: added data compression algorithm
This commit is contained in:
Péter Szilágyi
GitHub
commit
2e4d23a793
170
common/bitutil/compress.go
Normal file
170
common/bitutil/compress.go
Normal file
@ -0,0 +1,170 @@
|
||||
// Copyright 2017 The go-ethereum Authors
|
||||
// This file is part of the go-ethereum library.
|
||||
//
|
||||
// The go-ethereum library is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Lesser General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
//
|
||||
// The go-ethereum library is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Lesser General Public License for more details.
|
||||
//
|
||||
// You should have received a copy of the GNU Lesser General Public License
|
||||
// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package bitutil
|
||||
|
||||
import "errors"
|
||||
|
||||
var (
|
||||
// errMissingData is returned from decompression if the byte referenced by
|
||||
// the bitset header overflows the input data.
|
||||
errMissingData = errors.New("missing bytes on input")
|
||||
|
||||
// errUnreferencedData is returned from decompression if not all bytes were used
|
||||
// up from the input data after decompressing it.
|
||||
errUnreferencedData = errors.New("extra bytes on input")
|
||||
|
||||
// errExceededTarget is returned from decompression if the bitset header has
|
||||
// more bits defined than the number of target buffer space available.
|
||||
errExceededTarget = errors.New("target data size exceeded")
|
||||
|
||||
// errZeroContent is returned from decompression if a data byte referenced in
|
||||
// the bitset header is actually a zero byte.
|
||||
errZeroContent = errors.New("zero byte in input content")
|
||||
)
|
||||
|
||||
// The compression algorithm implemented by CompressBytes and DecompressBytes is
|
||||
// optimized for sparse input data which contains a lot of zero bytes. Decompression
|
||||
// requires knowledge of the decompressed data length.
|
||||
//
|
||||
// Compression works as follows:
|
||||
//
|
||||
// if data only contains zeroes,
|
||||
// CompressBytes(data) == nil
|
||||
// otherwise if len(data) <= 1,
|
||||
// CompressBytes(data) == data
|
||||
// otherwise:
|
||||
// CompressBytes(data) == append(CompressBytes(nonZeroBitset(data)), nonZeroBytes(data)...)
|
||||
// where
|
||||
// nonZeroBitset(data) is a bit vector with len(data) bits (MSB first):
|
||||
// nonZeroBitset(data)[i/8] && (1 << (7-i%8)) != 0 if data[i] != 0
|
||||
// len(nonZeroBitset(data)) == (len(data)+7)/8
|
||||
// nonZeroBytes(data) contains the non-zero bytes of data in the same order
|
||||
|
||||
// CompressBytes compresses the input byte slice according to the sparse bitset
|
||||
// representation algorithm. If the result is bigger than the original input, no
|
||||
// compression is done.
|
||||
func CompressBytes(data []byte) []byte {
|
||||
if out := bitsetEncodeBytes(data); len(out) < len(data) {
|
||||
return out
|
||||
}
|
||||
cpy := make([]byte, len(data))
|
||||
copy(cpy, data)
|
||||
return cpy
|
||||
}
|
||||
|
||||
// bitsetEncodeBytes compresses the input byte slice according to the sparse
|
||||
// bitset representation algorithm.
|
||||
func bitsetEncodeBytes(data []byte) []byte {
|
||||
// Empty slices get compressed to nil
|
||||
if len(data) == 0 {
|
||||
return nil
|
||||
}
|
||||
// One byte slices compress to nil or retain the single byte
|
||||
if len(data) == 1 {
|
||||
if data[0] == 0 {
|
||||
return nil
|
||||
}
|
||||
return data
|
||||
}
|
||||
// Calculate the bitset of set bytes, and gather the non-zero bytes
|
||||
nonZeroBitset := make([]byte, (len(data)+7)/8)
|
||||
nonZeroBytes := make([]byte, 0, len(data))
|
||||
|
||||
for i, b := range data {
|
||||
if b != 0 {
|
||||
nonZeroBytes = append(nonZeroBytes, b)
|
||||
nonZeroBitset[i/8] |= 1 << byte(7-i%8)
|
||||
}
|
||||
}
|
||||
if len(nonZeroBytes) == 0 {
|
||||
return nil
|
||||
}
|
||||
return append(bitsetEncodeBytes(nonZeroBitset), nonZeroBytes...)
|
||||
}
|
||||
|
||||
// DecompressBytes decompresses data with a known target size. If the input data
|
||||
// matches the size of the target, it means no compression was done in the first
|
||||
// place.
|
||||
func DecompressBytes(data []byte, target int) ([]byte, error) {
|
||||
if len(data) > target {
|
||||
return nil, errExceededTarget
|
||||
}
|
||||
if len(data) == target {
|
||||
cpy := make([]byte, len(data))
|
||||
copy(cpy, data)
|
||||
return cpy, nil
|
||||
}
|
||||
return bitsetDecodeBytes(data, target)
|
||||
}
|
||||
|
||||
// bitsetDecodeBytes decompresses data with a known target size.
|
||||
func bitsetDecodeBytes(data []byte, target int) ([]byte, error) {
|
||||
out, size, err := bitsetDecodePartialBytes(data, target)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
if size != len(data) {
|
||||
return nil, errUnreferencedData
|
||||
}
|
||||
return out, nil
|
||||
}
|
||||
|
||||
// bitsetDecodePartialBytes decompresses data with a known target size, but does
|
||||
// not enforce consuming all the input bytes. In addition to the decompressed
|
||||
// output, the function returns the length of compressed input data corresponding
|
||||
// to the output as the input slice may be longer.
|
||||
func bitsetDecodePartialBytes(data []byte, target int) ([]byte, int, error) {
|
||||
// Sanity check 0 targets to avoid infinite recursion
|
||||
if target == 0 {
|
||||
return nil, 0, nil
|
||||
}
|
||||
// Handle the zero and single byte corner cases
|
||||
decomp := make([]byte, target)
|
||||
if len(data) == 0 {
|
||||
return decomp, 0, nil
|
||||
}
|
||||
if target == 1 {
|
||||
decomp[0] = data[0] // copy to avoid referencing the input slice
|
||||
if data[0] != 0 {
|
||||
return decomp, 1, nil
|
||||
}
|
||||
return decomp, 0, nil
|
||||
}
|
||||
// Decompress the bitset of set bytes and distribute the non zero bytes
|
||||
nonZeroBitset, ptr, err := bitsetDecodePartialBytes(data, (target+7)/8)
|
||||
if err != nil {
|
||||
return nil, ptr, err
|
||||
}
|
||||
for i := 0; i < 8*len(nonZeroBitset); i++ {
|
||||
if nonZeroBitset[i/8]&(1<<byte(7-i%8)) != 0 {
|
||||
// Make sure we have enough data to push into the correct slot
|
||||
if ptr >= len(data) {
|
||||
return nil, 0, errMissingData
|
||||
}
|
||||
if i >= len(decomp) {
|
||||
return nil, 0, errExceededTarget
|
||||
}
|
||||
// Make sure the data is valid and push into the slot
|
||||
if data[ptr] == 0 {
|
||||
return nil, 0, errZeroContent
|
||||
}
|
||||
decomp[i] = data[ptr]
|
||||
ptr++
|
||||
}
|
||||
}
|
||||
return decomp, ptr, nil
|
||||
}
|
||||
56
common/bitutil/compress_fuzz.go
Normal file
56
common/bitutil/compress_fuzz.go
Normal file
@ -0,0 +1,56 @@
|
||||
// Copyright 2017 The go-ethereum Authors
|
||||
// This file is part of the go-ethereum library.
|
||||
//
|
||||
// The go-ethereum library is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Lesser General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
//
|
||||
// The go-ethereum library is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Lesser General Public License for more details.
|
||||
//
|
||||
// You should have received a copy of the GNU Lesser General Public License
|
||||
// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
// +build gofuzz
|
||||
|
||||
package bitutil
|
||||
|
||||
import "bytes"
|
||||
|
||||
// Fuzz implements a go-fuzz fuzzer method to test various encoding method
|
||||
// invocations.
|
||||
func Fuzz(data []byte) int {
|
||||
if len(data) == 0 {
|
||||
return -1
|
||||
}
|
||||
if data[0]%2 == 0 {
|
||||
return fuzzEncode(data[1:])
|
||||
}
|
||||
return fuzzDecode(data[1:])
|
||||
}
|
||||
|
||||
// fuzzEncode implements a go-fuzz fuzzer method to test the bitset encoding and
|
||||
// decoding algorithm.
|
||||
func fuzzEncode(data []byte) int {
|
||||
proc, _ := bitsetDecodeBytes(bitsetEncodeBytes(data), len(data))
|
||||
if !bytes.Equal(data, proc) {
|
||||
panic("content mismatch")
|
||||
}
|
||||
return 0
|
||||
}
|
||||
|
||||
// fuzzDecode implements a go-fuzz fuzzer method to test the bit decoding and
|
||||
// reencoding algorithm.
|
||||
func fuzzDecode(data []byte) int {
|
||||
blob, err := bitsetDecodeBytes(data, 1024)
|
||||
if err != nil {
|
||||
return 0
|
||||
}
|
||||
if comp := bitsetEncodeBytes(blob); !bytes.Equal(comp, data) {
|
||||
panic("content mismatch")
|
||||
}
|
||||
return 0
|
||||
}
|
||||
181
common/bitutil/compress_test.go
Normal file
181
common/bitutil/compress_test.go
Normal file
@ -0,0 +1,181 @@
|
||||
// Copyright 2017 The go-ethereum Authors
|
||||
// This file is part of the go-ethereum library.
|
||||
//
|
||||
// The go-ethereum library is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Lesser General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
//
|
||||
// The go-ethereum library is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Lesser General Public License for more details.
|
||||
//
|
||||
// You should have received a copy of the GNU Lesser General Public License
|
||||
// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package bitutil
|
||||
|
||||
import (
|
||||
"bytes"
|
||||
"math/rand"
|
||||
"testing"
|
||||
|
||||
"github.com/ethereum/go-ethereum/common/hexutil"
|
||||
)
|
||||
|
||||
// Tests that data bitset encoding and decoding works and is bijective.
|
||||
func TestEncodingCycle(t *testing.T) {
|
||||
tests := []string{
|
||||
// Tests generated by go-fuzz to maximize code coverage
|
||||
"0x000000000000000000",
|
||||
"0xef0400",
|
||||
"0xdf7070533534333636313639343638373532313536346c1bc33339343837313070706336343035336336346c65fefb3930393233383838ac2f65fefb",
|
||||
"0x7b64000000",
|
||||
"0x000034000000000000",
|
||||
"0x0000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000f0000000000000000000",
|
||||
"0x4912385c0e7b64000000",
|
||||
"0x000034000000000000000000000000000000",
|
||||
"0x00",
|
||||
"0x000003e834ff7f0000",
|
||||
"0x0000",
|
||||
"0x0000000000000000000000000000000000000000000000000000000000ff00",
|
||||
"0x895f0c6a020f850c6a020f85f88df88d",
|
||||
"0xdf7070533534333636313639343638373432313536346c1bc3315aac2f65fefb",
|
||||
"0x0000000000",
|
||||
"0xdf70706336346c65fefb",
|
||||
"0x00006d643634000000",
|
||||
"0xdf7070533534333636313639343638373532313536346c1bc333393438373130707063363430353639343638373532313536346c1bc333393438336336346c65fe",
|
||||
}
|
||||
for i, tt := range tests {
|
||||
data := hexutil.MustDecode(tt)
|
||||
|
||||
proc, err := bitsetDecodeBytes(bitsetEncodeBytes(data), len(data))
|
||||
if err != nil {
|
||||
t.Errorf("test %d: failed to decompress compressed data: %v", i, err)
|
||||
continue
|
||||
}
|
||||
if !bytes.Equal(data, proc) {
|
||||
t.Errorf("test %d: compress/decompress mismatch: have %x, want %x", i, proc, data)
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// Tests that data bitset decoding and rencoding works and is bijective.
|
||||
func TestDecodingCycle(t *testing.T) {
|
||||
tests := []struct {
|
||||
size int
|
||||
input string
|
||||
fail error
|
||||
}{
|
||||
{size: 0, input: "0x"},
|
||||
|
||||
// Crashers generated by go-fuzz
|
||||
{size: 0, input: "0x0020", fail: errUnreferencedData},
|
||||
{size: 0, input: "0x30", fail: errUnreferencedData},
|
||||
{size: 1, input: "0x00", fail: errUnreferencedData},
|
||||
{size: 2, input: "0x07", fail: errMissingData},
|
||||
{size: 1024, input: "0x8000", fail: errZeroContent},
|
||||
|
||||
// Tests generated by go-fuzz to maximize code coverage
|
||||
{size: 29490, input: "0x343137343733323134333839373334323073333930783e3078333930783e70706336346c65303e", fail: errMissingData},
|
||||
{size: 59395, input: "0x00", fail: errUnreferencedData},
|
||||
{size: 52574, input: "0x70706336346c65c0de", fail: errExceededTarget},
|
||||
{size: 42264, input: "0x07", fail: errMissingData},
|
||||
{size: 52, input: "0xa5045bad48f4", fail: errExceededTarget},
|
||||
{size: 52574, input: "0xc0de", fail: errMissingData},
|
||||
{size: 52574, input: "0x"},
|
||||
{size: 29490, input: "0x34313734373332313433383937333432307333393078073034333839373334323073333930783e3078333937333432307333393078073061333930783e70706336346c65303e", fail: errMissingData},
|
||||
{size: 29491, input: "0x3973333930783e30783e", fail: errMissingData},
|
||||
|
||||
{size: 1024, input: "0x808080608080"},
|
||||
{size: 1024, input: "0x808470705e3632383337363033313434303137393130306c6580ef46806380635a80"},
|
||||
{size: 1024, input: "0x8080808070"},
|
||||
{size: 1024, input: "0x808070705e36346c6580ef46806380635a80"},
|
||||
{size: 1024, input: "0x80808046802680"},
|
||||
{size: 1024, input: "0x4040404035"},
|
||||
{size: 1024, input: "0x4040bf3ba2b3f684402d353234373438373934409fe5b1e7ada94ebfd7d0505e27be4035"},
|
||||
{size: 1024, input: "0x404040bf3ba2b3f6844035"},
|
||||
{size: 1024, input: "0x40402d35323437343837393440bfd7d0505e27be4035"},
|
||||
}
|
||||
for i, tt := range tests {
|
||||
data := hexutil.MustDecode(tt.input)
|
||||
|
||||
orig, err := bitsetDecodeBytes(data, tt.size)
|
||||
if err != tt.fail {
|
||||
t.Errorf("test %d: failure mismatch: have %v, want %v", i, err, tt.fail)
|
||||
}
|
||||
if err != nil {
|
||||
continue
|
||||
}
|
||||
if comp := bitsetEncodeBytes(orig); !bytes.Equal(comp, data) {
|
||||
t.Errorf("test %d: decompress/compress mismatch: have %x, want %x", i, comp, data)
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// TestCompression tests that compression works by returning either the bitset
|
||||
// encoded input, or the actual input if the bitset version is longer.
|
||||
func TestCompression(t *testing.T) {
|
||||
// Check the the compression returns the bitset encoding is shorter
|
||||
in := hexutil.MustDecode("0x4912385c0e7b64000000")
|
||||
out := hexutil.MustDecode("0x80fe4912385c0e7b64")
|
||||
|
||||
if data := CompressBytes(in); bytes.Compare(data, out) != 0 {
|
||||
t.Errorf("encoding mismatch for sparse data: have %x, want %x", data, out)
|
||||
}
|
||||
if data, err := DecompressBytes(out, len(in)); err != nil || bytes.Compare(data, in) != 0 {
|
||||
t.Errorf("decoding mismatch for sparse data: have %x, want %x, error %v", data, in, err)
|
||||
}
|
||||
// Check the the compression returns the input if the bitset encoding is longer
|
||||
in = hexutil.MustDecode("0xdf7070533534333636313639343638373532313536346c1bc33339343837313070706336343035336336346c65fefb3930393233383838ac2f65fefb")
|
||||
out = hexutil.MustDecode("0xdf7070533534333636313639343638373532313536346c1bc33339343837313070706336343035336336346c65fefb3930393233383838ac2f65fefb")
|
||||
|
||||
if data := CompressBytes(in); bytes.Compare(data, out) != 0 {
|
||||
t.Errorf("encoding mismatch for dense data: have %x, want %x", data, out)
|
||||
}
|
||||
if data, err := DecompressBytes(out, len(in)); err != nil || bytes.Compare(data, in) != 0 {
|
||||
t.Errorf("decoding mismatch for dense data: have %x, want %x, error %v", data, in, err)
|
||||
}
|
||||
// Check that decompressing a longer input than the target fails
|
||||
if _, err := DecompressBytes([]byte{0xc0, 0x01, 0x01}, 2); err != errExceededTarget {
|
||||
t.Errorf("decoding error mismatch for long data: have %v, want %v", err, errExceededTarget)
|
||||
}
|
||||
}
|
||||
|
||||
// Crude benchmark for compressing random slices of bytes.
|
||||
func BenchmarkEncoding1KBVerySparse(b *testing.B) { benchmarkEncoding(b, 1024, 0.0001) }
|
||||
func BenchmarkEncoding2KBVerySparse(b *testing.B) { benchmarkEncoding(b, 2048, 0.0001) }
|
||||
func BenchmarkEncoding4KBVerySparse(b *testing.B) { benchmarkEncoding(b, 4096, 0.0001) }
|
||||
|
||||
func BenchmarkEncoding1KBSparse(b *testing.B) { benchmarkEncoding(b, 1024, 0.001) }
|
||||
func BenchmarkEncoding2KBSparse(b *testing.B) { benchmarkEncoding(b, 2048, 0.001) }
|
||||
func BenchmarkEncoding4KBSparse(b *testing.B) { benchmarkEncoding(b, 4096, 0.001) }
|
||||
|
||||
func BenchmarkEncoding1KBDense(b *testing.B) { benchmarkEncoding(b, 1024, 0.1) }
|
||||
func BenchmarkEncoding2KBDense(b *testing.B) { benchmarkEncoding(b, 2048, 0.1) }
|
||||
func BenchmarkEncoding4KBDense(b *testing.B) { benchmarkEncoding(b, 4096, 0.1) }
|
||||
|
||||
func BenchmarkEncoding1KBSaturated(b *testing.B) { benchmarkEncoding(b, 1024, 0.5) }
|
||||
func BenchmarkEncoding2KBSaturated(b *testing.B) { benchmarkEncoding(b, 2048, 0.5) }
|
||||
func BenchmarkEncoding4KBSaturated(b *testing.B) { benchmarkEncoding(b, 4096, 0.5) }
|
||||
|
||||
func benchmarkEncoding(b *testing.B, bytes int, fill float64) {
|
||||
// Generate a random slice of bytes to compress
|
||||
random := rand.NewSource(0) // reproducible and comparable
|
||||
|
||||
data := make([]byte, bytes)
|
||||
bits := int(float64(bytes) * 8 * fill)
|
||||
|
||||
for i := 0; i < bits; i++ {
|
||||
idx := random.Int63() % int64(len(data))
|
||||
bit := uint(random.Int63() % 8)
|
||||
data[idx] |= 1 << bit
|
||||
}
|
||||
// Reset the benchmark and measure encoding/decoding
|
||||
b.ResetTimer()
|
||||
b.ReportAllocs()
|
||||
for i := 0; i < b.N; i++ {
|
||||
bitsetDecodeBytes(bitsetEncodeBytes(data), len(data))
|
||||
}
|
||||
}
|
||||
Loading…
Reference in New Issue
Block a user