英文:
How to deserialize a non-standard size field?
问题
我必须对来自另一个应用程序的一些二进制消息进行反序列化。我很想使用restruct.io,但是消息结构中的一些字段使用了“非标准”的位数(5位、3位、10位等)。
有没有办法处理这种类型的结构?我已经搜索了一段时间,没有任何成功的结果,所以任何帮助都将非常受欢迎。
提前感谢。
我将尝试给出一个示例来澄清我的问题。给定以下代码:
package main
import (
"encoding/binary"
"fmt"
restruct "gopkg.in/restruct.v1"
)
type MessageType uint8
const (
MessageTypeOne MessageType = iota + 1
MessageTypeTwo
MessageTypeThree
)
// Message is the data to deserialize from the binary stream
type Message struct {
Length uint32 `struct:"uint32"` // message size in bytes (including length)
Type MessageType `struct:"uint8"`
Version uint8 `struct:"uint8:4"` // Just need 4 bits
Subversion uint8 `struct:"uint8:2"` // just need 2 bits
Optional uint8 `struct:"uint8:1"` // just one bit --> '1' means next field is NOT present
NodeName string ``
ANumber uint16 `struct:"uint16:10"` // just need 10 bits
}
// (length(4)+type(1)+(version(4bits)+Subversion(2bits)+Optional(1bit))) = 6 bytes
// need 32bit alignment
func main() {
var inStream = []byte{0x08, // just 8 bytes needed
0x01, // messge type = MessageTypeOne
0x4a, // Version=0100 Subversion=10 Optional=1 ANumber = 0 (MSB bit)
0x00, 0x60, // ANumber(000 0000 011) Padding = 0 0000 for 32 bits alignment
}
var msg Message
err := restruct.Unpack(inStream, binary.BigEndian, &msg)
if err != nil {
panic(err)
}
fmt.Println(msg)
// Expected:
// msg.Length = 8
// msg.Type = 1
// msg.Version = 4
// msg.Subversion = 2
// msg.Optional = 1
// msg.NodeName = ""
// msg.ANumber = 3
}
我将从TCP连接接收inStream,并希望对二进制数据进行反序列化,并获得一个具有预期值的Message结构体。
希望这能澄清我的问题。
再次感谢;)
英文:
I have to deserialize some binary messages coming from another application. I would love to use restruct.io but some fields in the message structure use a "non-standard" number of bits ( 5 bits, 3 bits, ... 10 bits ... ).
Is there any way to handle this type of structs? I have been searching for some time without any success so any help will be very welcomed.
thanks in advance
I wil try to give an example to clarify my question. Given the code:
package main
import (
"encoding/binary"
"fmt"
restruct "gopkg.in/restruct.v1"
)
type MessageType uint8
const (
MessageTypeOne MessageType = iota + 1
MessageTypeTwo
MessageTypeThree
)
// Message is the data to deserialize from the binary stream
type Message struct {
Length uint32 `struct:"uint32"` // message size in bytes (including length)
Type MessageType `struct:"uint8"`
Version uint8 `struct:"uint8:4"` // Just need 4 bits
Subversion uint8 `struct:"uint8:2"` // just need 2 bits
Optional uint8 `struct:"uint8:1"` // just one bit --> '1' means next field is NOT present
NodeName string ``
ANumber uint16 `struct:"uint16:10"` // just need 10 bits
}
// (length(4)+type(1)+(version(4bits)+Subversion(2bits)+Optional(1bit))) = 6 bytes
// need 32bit alignment
func main() {
var inStream = []byte{0x08, // just 8 bytes needed
0x01, // messge type = MessageTypeOne
0x4a, // Version=0100 Subversion=10 Optional=1 ANumber = 0 (MSB bit)
0x00, 0x60, // ANumber(000 0000 011) Padding = 0 0000 for 32 bits alignment
}
var msg Message
err := restruct.Unpack(inStream, binary.BigEndian, &msg)
if err != nil {
panic(err)
}
fmt.Println(msg)
// Expected:
// msg.Length = 8
// msg.Type = 1
// msg.Version = 4
// msg.Subversion = 2
// msg.Optional = 1
// msg.NodeName = ""
// msg.ANumber = 3
}
I will receive inStream from a TCP connection and will want to deserialize the binary data and get a Message struct with the expected values ...
Hope this will clarify my question.
Thanks again
答案1
得分: 1
虽然可能没有通用的包来实现这种自定义结构的打包,但你可以很容易地创建自己的方法,仅提取每个字段所需的位。
func (m *Message) UnmarshalBinary(data []byte) error {
m.Length = binary.BigEndian.Uint32(data[:4])
if int(m.Length) > len(data) {
return fmt.Errorf("not enough bytes")
}
m.Type = MessageType(data[4])
m.Version = data[5] >> 4
m.Subversion = data[5] >> 2 & 0x03
m.Optional = data[5] >> 1 & 0x01
// 如果有可选字符串,将 ANumber 的索引移回
idx := 6
if m.Optional == 0 {
// 移除 ANumber 的最后两个字节
end := int(m.Length) - 2
m.NodeName = string(data[6:end])
idx = end
}
m.ANumber = uint16(data[idx]&0xc0)<<2 | uint16(data[idx]&0x3f<<2|data[idx+1]>>6)
return nil
}
当然,你可以添加更多的边界检查,以返回错误,而不是在索引越界时引发 panic。
我稍微修改了你的 inStream
切片以匹配你的定义,你可以在这里看到示例输出:https://play.golang.org/p/FoNoazluOF
英文:
While there's probably no generic package to implement this custom struct packing, you can easily create your own method extracting just the bits required for each field.
func (m *Message) UnmarshalBinary(data []byte) error {
m.Length = binary.BigEndian.Uint32(data[:4])
if int(m.Length) > len(data) {
return fmt.Errorf("not enough bytes")
}
m.Type = MessageType(data[4])
m.Version = data[5] >> 4
m.Subversion = data[5] >> 2 & 0x03
m.Optional = data[5] >> 1 & 0x01
// move the index for ANumber back if there's an optional string
idx := 6
if m.Optional == 0 {
// remove the last two bytes for ANumber
end := int(m.Length) - 2
m.NodeName = string(data[6:end])
idx = end
}
m.ANumber = uint16(data[idx]&0xc0)<<2 | uint16(data[idx]&0x3f<<2|data[idx+1]>>6)
return nil
}
You can of course add more bound checks to return errors rather than letting this panic when indexing out of bounds.
I modified your inStream
slice slightly to match your definition, and you can see the example output here: https://play.golang.org/p/FoNoazluOF
答案2
得分: 0
我一直在为restruct.io编写一些补丁,以便能够处理位字段...尽管还没有完全测试,但似乎可以工作...
一旦测试完成,我将尝试发送一个拉取请求...
func (e *encoder) writeBits(f field, inBuf []byte) {
var inputLength uint8 = uint8(len(inBuf))
if f.BitSize == 0 {
// Having problems with complex64 type ... so we asume we want to read all
//f.BitSize = uint8(f.Type.Bits())
f.BitSize = 8 * inputLength
}
// destPos: Destination position ( in the result ) of the first bit in the first byte
var destPos uint8 = 8 - e.bitCounter
// originPos: Original position of the first bit in the first byte
var originPos uint8 = f.BitSize % 8
if originPos == 0 {
originPos = 8
}
// numBytes: number of complete bytes to hold the result
var numBytes uint8 = f.BitSize / 8
// numBits: number of remaining bits in the first non-complete byte of the result
var numBits uint8 = f.BitSize % 8
// number of positions we have to shift the bytes to get the result
var shift uint8
if originPos > destPos {
shift = originPos - destPos
} else {
shift = destPos - originPos
}
shift = shift % 8
var inputInitialIdx uint8 = inputLength - numBytes
if numBits > 0 {
inputInitialIdx = inputInitialIdx - 1
}
if originPos < destPos {
// shift left
carry := func(idx uint8) uint8 {
if (idx + 1) < inputLength {
return (inBuf[idx+1] >> (8 - shift))
}
return 0x00
}
mask := func(idx uint8) uint8 {
if idx == 0 {
return (0x01 << destPos) - 1
}
return 0xFF
}
var idx uint8 = 0
for inIdx := inputInitialIdx; inIdx < inputLength; inIdx++ {
e.buf[idx] |= ((inBuf[inIdx] << shift) | carry(inIdx)) & mask(idx)
idx++
}
} else {
// originPos >= destPos => shift right
var idx uint8 = 0
// carry : is a little bit tricky in this case because of the first case
// when idx == 0 and there is no carry at all
carry := func(idx uint8) uint8 {
if idx == 0 {
return 0x00
}
return (inBuf[idx-1] << (8 - shift))
}
mask := func(idx uint8) uint8 {
if idx == 0 {
return (0x01 << destPos) - 1
}
return 0xFF
}
inIdx := inputInitialIdx
for ; inIdx < inputLength; inIdx++ {
//note: Should the mask be done BEFORE the OR with carry?
e.buf[idx] |= ((inBuf[inIdx] >> shift) | carry(inIdx)) & mask(idx)
idx++
}
if ((e.bitCounter + f.BitSize) % 8) > 0 {
e.buf[idx] |= carry(inIdx)
}
}
//now we should update buffer and bitCounter
e.bitCounter = (e.bitCounter + f.BitSize) % 8
// move the head to the next non-complete byte used
headerUpdate := func() uint8 {
if (e.bitCounter == 0) && ((f.BitSize % 8) != 0) {
return (numBytes + 1)
}
return numBytes
}
e.buf = e.buf[headerUpdate():]
return
}
英文:
I have been working on some patch for restruct.io to be able to work with bitfields .... Still not fully tested but seems to work ...
Will try to send a pull request once tested ...
func (e *encoder) writeBits(f field, inBuf []byte) {
var inputLength uint8 = uint8(len(inBuf))
if f.BitSize == 0 {
// Having problems with complex64 type ... so we asume we want to read all
//f.BitSize = uint8(f.Type.Bits())
f.BitSize = 8 * inputLength
}
// destPos: Destination position ( in the result ) of the first bit in the first byte
var destPos uint8 = 8 - e.bitCounter
// originPos: Original position of the first bit in the first byte
var originPos uint8 = f.BitSize % 8
if originPos == 0 {
originPos = 8
}
// numBytes: number of complete bytes to hold the result
var numBytes uint8 = f.BitSize / 8
// numBits: number of remaining bits in the first non-complete byte of the result
var numBits uint8 = f.BitSize % 8
// number of positions we have to shift the bytes to get the result
var shift uint8
if originPos > destPos {
shift = originPos - destPos
} else {
shift = destPos - originPos
}
shift = shift % 8
var inputInitialIdx uint8 = inputLength - numBytes
if numBits > 0 {
inputInitialIdx = inputInitialIdx - 1
}
if originPos < destPos {
// shift left
carry := func(idx uint8) uint8 {
if (idx + 1) < inputLength {
return (inBuf[idx+1] >> (8 - shift))
}
return 0x00
}
mask := func(idx uint8) uint8 {
if idx == 0 {
return (0x01 << destPos) - 1
}
return 0xFF
}
var idx uint8 = 0
for inIdx := inputInitialIdx; inIdx < inputLength; inIdx++ {
e.buf[idx] |= ((inBuf[inIdx] << shift) | carry(inIdx)) & mask(idx)
idx++
}
} else {
// originPos >= destPos => shift right
var idx uint8 = 0
// carry : is a little bit tricky in this case because of the first case
// when idx == 0 and there is no carry at all
carry := func(idx uint8) uint8 {
if idx == 0 {
return 0x00
}
return (inBuf[idx-1] << (8 - shift))
}
mask := func(idx uint8) uint8 {
if idx == 0 {
return (0x01 << destPos) - 1
}
return 0xFF
}
inIdx := inputInitialIdx
for ; inIdx < inputLength; inIdx++ {
//note: Should the mask be done BEFORE the OR with carry?
e.buf[idx] |= ((inBuf[inIdx] >> shift) | carry(inIdx)) & mask(idx)
idx++
}
if ((e.bitCounter + f.BitSize) % 8) > 0 {
e.buf[idx] |= carry(inIdx)
}
}
//now we should update buffer and bitCounter
e.bitCounter = (e.bitCounter + f.BitSize) % 8
// move the head to the next non-complete byte used
headerUpdate := func() uint8 {
if (e.bitCounter == 0) && ((f.BitSize % 8) != 0) {
return (numBytes + 1)
}
return numBytes
}
e.buf = e.buf[headerUpdate():]
return
}
通过集体智慧和协作来改善编程学习和解决问题的方式。致力于成为全球开发者共同参与的知识库,让每个人都能够通过互相帮助和分享经验来进步。
评论