0
0
Fork 0
mirror of https://github.com/go-gitea/gitea synced 2024-12-24 21:35:05 +01:00
gitea/vendor/github.com/pingcap/tidb/util/codec/bytes.go
Thomas Boerger b6a95a8cb3 Integrate public as bindata optionally (#293)
* Dropped unused codekit config

* Integrated dynamic and static bindata for public

* Ignore public bindata

* Add a general generate make task

* Integrated flexible public assets into web command

* Updated vendoring, added all missiong govendor deps

* Made the linter happy with the bindata and dynamic code

* Moved public bindata definition to modules directory

* Ignoring the new bindata path now

* Updated to the new public modules import path

* Updated public bindata command and drop the new prefix
2016-11-30 00:26:36 +08:00

197 lines
5.3 KiB
Go

// Copyright 2015 PingCAP, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// See the License for the specific language governing permissions and
// limitations under the License.
package codec
import (
"bytes"
"encoding/binary"
"runtime"
"unsafe"
"github.com/juju/errors"
)
const (
encGroupSize = 8
encMarker = byte(0xFF)
encPad = byte(0x0)
)
var (
pads = make([]byte, encGroupSize)
encPads = []byte{encPad}
)
// EncodeBytes guarantees the encoded value is in ascending order for comparison,
// encoding with the following rule:
// [group1][marker1]...[groupN][markerN]
// group is 8 bytes slice which is padding with 0.
// marker is `0xFF - padding 0 count`
// For example:
// [] -> [0, 0, 0, 0, 0, 0, 0, 0, 247]
// [1, 2, 3] -> [1, 2, 3, 0, 0, 0, 0, 0, 250]
// [1, 2, 3, 0] -> [1, 2, 3, 0, 0, 0, 0, 0, 251]
// [1, 2, 3, 4, 5, 6, 7, 8] -> [1, 2, 3, 4, 5, 6, 7, 8, 255, 0, 0, 0, 0, 0, 0, 0, 0, 247]
// Refer: https://github.com/facebook/mysql-5.6/wiki/MyRocks-record-format#memcomparable-format
func EncodeBytes(b []byte, data []byte) []byte {
// Allocate more space to avoid unnecessary slice growing.
// Assume that the byte slice size is about `(len(data) / encGroupSize + 1) * (encGroupSize + 1)` bytes,
// that is `(len(data) / 8 + 1) * 9` in our implement.
dLen := len(data)
reallocSize := (dLen/encGroupSize + 1) * (encGroupSize + 1)
result := reallocBytes(b, reallocSize)
for idx := 0; idx <= dLen; idx += encGroupSize {
remain := dLen - idx
padCount := 0
if remain >= encGroupSize {
result = append(result, data[idx:idx+encGroupSize]...)
} else {
padCount = encGroupSize - remain
result = append(result, data[idx:]...)
result = append(result, pads[:padCount]...)
}
marker := encMarker - byte(padCount)
result = append(result, marker)
}
return result
}
func decodeBytes(b []byte, reverse bool) ([]byte, []byte, error) {
data := make([]byte, 0, len(b))
for {
if len(b) < encGroupSize+1 {
return nil, nil, errors.New("insufficient bytes to decode value")
}
groupBytes := b[:encGroupSize+1]
if reverse {
reverseBytes(groupBytes)
}
group := groupBytes[:encGroupSize]
marker := groupBytes[encGroupSize]
// Check validity of marker.
padCount := encMarker - marker
realGroupSize := encGroupSize - padCount
if padCount > encGroupSize {
return nil, nil, errors.Errorf("invalid marker byte, group bytes %q", groupBytes)
}
data = append(data, group[:realGroupSize]...)
b = b[encGroupSize+1:]
if marker != encMarker {
// Check validity of padding bytes.
if bytes.Count(group[realGroupSize:], encPads) != int(padCount) {
return nil, nil, errors.Errorf("invalid padding byte, group bytes %q", groupBytes)
}
break
}
}
return b, data, nil
}
// DecodeBytes decodes bytes which is encoded by EncodeBytes before,
// returns the leftover bytes and decoded value if no error.
func DecodeBytes(b []byte) ([]byte, []byte, error) {
return decodeBytes(b, false)
}
// EncodeBytesDesc first encodes bytes using EncodeBytes, then bitwise reverses
// encoded value to guarantee the encoded value is in descending order for comparison.
func EncodeBytesDesc(b []byte, data []byte) []byte {
n := len(b)
b = EncodeBytes(b, data)
reverseBytes(b[n:])
return b
}
// DecodeBytesDesc decodes bytes which is encoded by EncodeBytesDesc before,
// returns the leftover bytes and decoded value if no error.
func DecodeBytesDesc(b []byte) ([]byte, []byte, error) {
return decodeBytes(b, true)
}
// EncodeCompactBytes joins bytes with its length into a byte slice. It is more
// efficient in both space and time compare to EncodeBytes. Note that the encoded
// result is not memcomparable.
func EncodeCompactBytes(b []byte, data []byte) []byte {
b = reallocBytes(b, binary.MaxVarintLen64+len(data))
b = EncodeVarint(b, int64(len(data)))
return append(b, data...)
}
// DecodeCompactBytes decodes bytes which is encoded by EncodeCompactBytes before.
func DecodeCompactBytes(b []byte) ([]byte, []byte, error) {
b, n, err := DecodeVarint(b)
if err != nil {
return nil, nil, errors.Trace(err)
}
if int64(len(b)) < n {
return nil, nil, errors.Errorf("insufficient bytes to decode value, expected length: %v", n)
}
return b[n:], b[:n], nil
}
// See https://golang.org/src/crypto/cipher/xor.go
const wordSize = int(unsafe.Sizeof(uintptr(0)))
const supportsUnaligned = runtime.GOARCH == "386" || runtime.GOARCH == "amd64"
func fastReverseBytes(b []byte) {
n := len(b)
w := n / wordSize
if w > 0 {
bw := *(*[]uintptr)(unsafe.Pointer(&b))
for i := 0; i < w; i++ {
bw[i] = ^bw[i]
}
}
for i := w * wordSize; i < n; i++ {
b[i] = ^b[i]
}
}
func safeReverseBytes(b []byte) {
for i := range b {
b[i] = ^b[i]
}
}
func reverseBytes(b []byte) {
if supportsUnaligned {
fastReverseBytes(b)
return
}
safeReverseBytes(b)
}
// like realloc.
func reallocBytes(b []byte, n int) []byte {
newSize := len(b) + n
if cap(b) < newSize {
bs := make([]byte, len(b), newSize)
copy(bs, b)
return bs
}
// slice b has capability to store n bytes
return b
}