将byte []转换为数组问题

Question

Please check this code

float f = BitConverter.ToSingle(new byte[] { 0xBF, 0x04, 0x8E, 0xFF }, 0);
byte[] b = BitConverter.GetBytes(f);

this yeilds a strange result. b will be { 0xBF, 0x04, 0xCE, 0xFF }

I guess it is because the value of f is NaN. the reason I'm asking this question is because I use Marsal to convert a stream of bytes into a struct that contains a float, the I swap endianity

The thing is when I get to the field is already messed up (like the above example)

any ideas?

Thanks!

Answer 1

Since your problem appears to an endian one, I'm wondering if checking for NaN and reversing the order and setting a boolean to reverse it back if necessary, would help:

byte[] input = { 0xBF, 0x04, 0x8E, 0xFF };
bool reversed = false;
float f = BitConverter.ToSingle(input, 0);

if (float.IsNaN(f))
{
    reversed = true;
    f = BitConverter.ToSingle(input.Reverse().ToArray(), 0);
}

byte[] b = BitConverter.GetBytes(f);

if (reversed)
    b = b.Reverse().ToArray();

Answer 2

This works:

double f = BitConverter.ToDouble(new byte[] { 0, 0, 0, 0, 0xBF, 0x04, 0x8E, 0xFF }, 0);
byte[] b = BitConverter.GetBytes(f);

This one also does:

int f = BitConverter.ToInt32(new byte[] { 0xBF, 0x04, 0x8E, 0xFF }, 0);
byte[] b = BitConverter.GetBytes(f);

I believe the reason is in the single floating-point storage. The third byte ( { 0xBF, 0x04, [0x8E to 0xBE] , 0xFF } is being +4 'd for some reason while converted.

The good containter for 4 bytes is Int32 and I advise you to use it.