可变长度二进制补码int64

Question

I'm attempting to write a Go program to parse ans.1 BER two's complement integer encoding. However, the integer can have either 1, 2, 3 or 4 byte length encoding (depending on its size).

According to the specification (http://www.itu.int/ITU-T/studygroups/com17/languages/X.690-0207.pdf) the leftmost bit is always the complement.

What's a clean way to do this?

func ParseInt(b []byte) (int64, error) {
    switch len(b) {
    case 1:
        // this works
        return int64(b[0]&0x7f) - int64(b[0]&0x80), nil
    case 2:
        // left most byte of b[0] is -32768
    case 3:
        // left most byte of b[0] is -8388608
    case 4:
        // left most byte of b[0] is -2147483648 (and so on for 5, 6, 7, 8)
    case 5:
    case 6:
    case 7:
    case 8:
    default:
        return 0, errors.New("value does not fit in a int64")
    }
}


ParseInt([]byte{0xfe})       // should return (-2, nil)
ParseInt([]byte{0xfe, 0xff}) // should return (-257, nil)
ParseInt([]byte{0x01, 0x00}) // should return (256, nil)

Answer 1

Easier to understand if you read the bytes from the end:

• You don't have to shift the last byte
• Left-shift the last byte by 8 (8 bits in a byte)
• Left-shift the 2nd last byte by 16
• ...
• And from the first byte only use 7 bits, the leftmost bit is special.

The leftmost bit of the first byte b[0]&080 tells if you have to add an offset to the result. The offset to be optionally added is -1 multiplied by the number your input would mean by having this one bit set and all others being 0, that is -1 * (1 << (len(b)*8 - 1)) = 0x80 << (len(b)*8 - 8).

Examples. If input is...

• 1 byte:
  int64(b[0]&0x7f) - int64(b[0]&0x80)
• 2 bytes:
  int64(b[0]&0x7f)<<8 + int64(b[1]) - int64(b[0]&0x80)<<8
• 3 bytes:
  int64(b[0]&0x7f)<<16 + int64(b[1])<<8 + int64(b[2]) - int64(b[0]&0x80)<<16

All these cases can be covered with a nice loop.

Here's a compact implementation (try it on the <kbd>Go Playground</kbd>):

func ParseInt(b []byte) (int64, error) {
    if len(b) > 8 {
        return 0, errors.New("value does not fit in a int64")
    }

    var n int64
    for i, v := range b {
        shift := uint((len(b) - i - 1) * 8)
        if i == 0 && v&0x80 != 0 {
            n -= 0x80 << shift
            v &= 0x7f
        }
        n += int64(v) << shift
    }
    return n, nil
}