From fd5d51c9ae3256a1f24cf974dcd02433a259677e Mon Sep 17 00:00:00 2001 From: Zsolt Felfoldi Date: Fri, 5 May 2017 20:11:55 +0200 Subject: common/bitutil: added data compression algorithm --- common/bitutil/compress.go | 93 ++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 93 insertions(+) create mode 100644 common/bitutil/compress.go (limited to 'common/bitutil/compress.go') diff --git a/common/bitutil/compress.go b/common/bitutil/compress.go new file mode 100644 index 000000000..c6c139ab9 --- /dev/null +++ b/common/bitutil/compress.go @@ -0,0 +1,93 @@ +// 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 . + +package bitutil + +/* +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(nonZeroBits(data)), nonZeroBytes(data)...) +where + nonZeroBits(data) is a bit vector with len(data) bits (MSB first): + nonZeroBits(data)[i/8] && (1 << (7-i%8)) != 0 if data[i] != 0 + len(nonZeroBits(data)) == (len(data)+7)/8 + nonZeroBytes(data) contains the non-zero bytes of data in the same order +*/ + +// CompressBytes compresses the input byte slice +func CompressBytes(data []byte) []byte { + if len(data) == 0 { + return nil + } + if len(data) == 1 { + if data[0] == 0 { + return nil + } else { + return data + } + } + + bitsLen := (len(data) + 7) / 8 + nonZeroBits := make([]byte, bitsLen) + nonZeroBytes := make([]byte, 0, len(data)) + for i, b := range data { + if b != 0 { + nonZeroBytes = append(nonZeroBytes, b) + nonZeroBits[i/8] |= 1 << byte(7-i%8) + } + } + if len(nonZeroBytes) == 0 { + return nil + } + return append(CompressBytes(nonZeroBits), nonZeroBytes...) +} + +// DecompressBytes decompresses data with a known target size. +// In addition to the decompressed output, the function returns the length of +// compressed input data corresponding to the output. The input slice may be longer. +// If the input slice is too short, (nil, -1) is returned. +func DecompressBytes(data []byte, targetLen int) ([]byte, int) { + decomp := make([]byte, targetLen) + if len(data) == 0 { + return decomp, 0 + } + if targetLen == 1 { + return data[0:1], 1 + } + + bitsLen := (targetLen + 7) / 8 + nonZeroBits, ptr := DecompressBytes(data, bitsLen) + if ptr < 0 { + return nil, -1 + } + for i, _ := range decomp { + if nonZeroBits[i/8]&(1< Date: Sat, 6 May 2017 18:35:59 +0300 Subject: common/bitutil: fix decompression corner cases; fuzz, test & bench --- common/bitutil/compress.go | 137 +++++++++++++++++++++++++++++++-------------- 1 file changed, 94 insertions(+), 43 deletions(-) (limited to 'common/bitutil/compress.go') diff --git a/common/bitutil/compress.go b/common/bitutil/compress.go index c6c139ab9..a806c0e8b 100644 --- a/common/bitutil/compress.go +++ b/common/bitutil/compress.go @@ -16,78 +16,129 @@ package bitutil -/* -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: +import "errors" -if data only contains zeroes, - CompressBytes(data) == nil -otherwise if len(data) <= 1, - CompressBytes(data) == data -otherwise: - CompressBytes(data) == append(CompressBytes(nonZeroBits(data)), nonZeroBytes(data)...) -where - nonZeroBits(data) is a bit vector with len(data) bits (MSB first): - nonZeroBits(data)[i/8] && (1 << (7-i%8)) != 0 if data[i] != 0 - len(nonZeroBits(data)) == (len(data)+7)/8 - nonZeroBytes(data) contains the non-zero bytes of data in the same order -*/ +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") -// CompressBytes compresses the input byte slice + // 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. func CompressBytes(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 - } else { - return data } + return data } - - bitsLen := (len(data) + 7) / 8 - nonZeroBits := make([]byte, bitsLen) + // 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) - nonZeroBits[i/8] |= 1 << byte(7-i%8) + nonZeroBitset[i/8] |= 1 << byte(7-i%8) } } if len(nonZeroBytes) == 0 { return nil } - return append(CompressBytes(nonZeroBits), nonZeroBytes...) + return append(CompressBytes(nonZeroBitset), nonZeroBytes...) } -// DecompressBytes decompresses data with a known target size. -// In addition to the decompressed output, the function returns the length of -// compressed input data corresponding to the output. The input slice may be longer. -// If the input slice is too short, (nil, -1) is returned. -func DecompressBytes(data []byte, targetLen int) ([]byte, int) { - decomp := make([]byte, targetLen) - if len(data) == 0 { - return decomp, 0 +// DecompressBytes decompresses data with a known target size. 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 DecompressBytes(data []byte, target int) ([]byte, error) { + out, size, err := decompressBytes(data, target) + if err != nil { + return nil, err } - if targetLen == 1 { - return data[0:1], 1 + if size != len(data) { + return nil, ErrUnreferencedData } + return out, nil +} - bitsLen := (targetLen + 7) / 8 - nonZeroBits, ptr := DecompressBytes(data, bitsLen) - if ptr < 0 { - return nil, -1 +// decompressBytes decompresses data with a known target size. 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 decompressBytes(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 := decompressBytes(data, (target+7)/8) + if err != nil { + return nil, ptr, err } - for i, _ := range decomp { - if nonZeroBits[i/8]&(1<= 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 + return decomp, ptr, nil } -- cgit v1.2.3 From 82defe5c5663ca0c28563f8a111d327c87726267 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?P=C3=A9ter=20Szil=C3=A1gyi?= Date: Mon, 8 May 2017 10:40:48 +0300 Subject: common/compress: internalize encoders, add length wrappers --- common/bitutil/compress.go | 72 +++++++++++++++++++++++++++++++--------------- 1 file changed, 49 insertions(+), 23 deletions(-) (limited to 'common/bitutil/compress.go') diff --git a/common/bitutil/compress.go b/common/bitutil/compress.go index a806c0e8b..c057cee4a 100644 --- a/common/bitutil/compress.go +++ b/common/bitutil/compress.go @@ -19,21 +19,21 @@ package bitutil import "errors" var ( - // ErrMissingData is returned from decompression if the byte referenced by + // errMissingData is returned from decompression if the byte referenced by // the bitset header overflows the input data. - ErrMissingData = errors.New("missing bytes on input") + errMissingData = errors.New("missing bytes on input") - // ErrUnreferencedData is returned from decompression if not all bytes were used + // 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") + errUnreferencedData = errors.New("extra bytes on input") - // ErrExceededTarget is returned from decompression if the bitset header has + // 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") + errExceededTarget = errors.New("target data size exceeded") - // ErrZeroContent is returned from decompression if a data byte referenced in + // 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") + errZeroContent = errors.New("zero byte in input content") ) // The compression algorithm implemented by CompressBytes and DecompressBytes is @@ -55,8 +55,20 @@ var ( // 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. +// 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 @@ -81,27 +93,41 @@ func CompressBytes(data []byte) []byte { if len(nonZeroBytes) == 0 { return nil } - return append(CompressBytes(nonZeroBitset), nonZeroBytes...) + return append(bitsetEncodeBytes(nonZeroBitset), nonZeroBytes...) } -// DecompressBytes decompresses data with a known target size. 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. +// 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) { - out, size, err := decompressBytes(data, target) + 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 nil, errUnreferencedData } return out, nil } -// decompressBytes decompresses data with a known target size. 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 decompressBytes(data []byte, target int) ([]byte, int, error) { +// 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 @@ -119,7 +145,7 @@ func decompressBytes(data []byte, target int) ([]byte, int, error) { return decomp, 0, nil } // Decompress the bitset of set bytes and distribute the non zero bytes - nonZeroBitset, ptr, err := decompressBytes(data, (target+7)/8) + nonZeroBitset, ptr, err := bitsetDecodePartialBytes(data, (target+7)/8) if err != nil { return nil, ptr, err } @@ -127,14 +153,14 @@ func decompressBytes(data []byte, target int) ([]byte, int, error) { if nonZeroBitset[i/8]&(1<= len(data) { - return nil, 0, ErrMissingData + return nil, 0, errMissingData } if i >= len(decomp) { - return nil, 0, ErrExceededTarget + return nil, 0, errExceededTarget } // Make sure the data is valid and push into the slot if data[ptr] == 0 { - return nil, 0, ErrZeroContent + return nil, 0, errZeroContent } decomp[i] = data[ptr] ptr++ -- cgit v1.2.3