How do I convert an array of floats to a byte[] and back?

Question

I have an array of Floats that need to be converted to a byte array and back to a float[]... can anyone help me do this correctly?

I'm working with the bitConverter class and found myself stuck trying to append the results.

The reason I'm doing this is so I can save runtime values into a IO Stream. The target storage is Azure Page blobs in case that matters. I don't care about what endian this is stored in, as long as it input matches the output.

static  byte[] ConvertFloatToByteArray(float[] floats)
        {
            byte[] ret = new byte[floats.Length * 4];// a single float is 4 bytes/32 bits

            for (int i = 0; i < floats.Length; i++)
            {
               // todo: stuck...I need to append the results to an offset of ret
                ret = BitConverter.GetBytes(floats[i]);

            }
            return ret;
        }


 static  float[] ConvertByteArrayToFloat(byte[] bytes)
{ //to do }

Solution 1

If you're looking for performance then you could use Buffer.BlockCopy. Nice and simple, and probably about as fast as you'll get in managed code.

var floatArray1 = new float[] { 123.45f, 123f, 45f, 1.2f, 34.5f };

// create a byte array and copy the floats into it...
var byteArray = new byte[floatArray1.Length * 4];
Buffer.BlockCopy(floatArray1, 0, byteArray, 0, byteArray.Length);

// create a second float array and copy the bytes into it...
var floatArray2 = new float[byteArray.Length / 4];
Buffer.BlockCopy(byteArray, 0, floatArray2, 0, byteArray.Length);

// do we have the same sequence of floats that we started with?
Console.WriteLine(floatArray1.SequenceEqual(floatArray2));    // True

Solution 2

There's the BitConverter.ToSingle(byte[] value, int startIndex) method that should help out here.

Returns a single-precision floating point number converted from four bytes at a specified position in a byte array.

Your probably want something like (untested):

static float[] ConvertByteArrayToFloat(byte[] bytes)
{
    if(bytes == null)
        throw new ArgumentNullException("bytes");

   if(bytes.Length % 4 != 0)
        throw new ArgumentException
              ("bytes does not represent a sequence of floats");

    return Enumerable.Range(0, bytes.Length / 4)
                     .Select(i => BitConverter.ToSingle(bytes, i * 4))
                     .ToArray();
}

EDIT: Non-LINQ:

float[] floats = new float[bytes.Length / 4];

for (int i = 0; i < bytes.Length / 4; i++)
    floats[i] = BitConverter.ToSingle(bytes, i * 4);

return floats;

Solution 3

You are not moving the position when you copy the float[i] into the byte array, you should write something like

Array.Copy(BitConverter.GetBytes(float[i]),0,res,i*4);

instead of just:

ret = BitConverter.GetBytes(floats[i]);

the inverse function follow the same strategy.

Solution 4

static float[] ConvertByteArrayToFloat(byte[] bytes)
{
    if(bytes.Length % 4 != 0) throw new ArgumentException();

    float[] floats = new float[bytes.Length/4];
    for(int i = 0; i < floats.Length; i++)
    {
        floats[i] = BitConverter.ToSingle(bytes, i*4);
    }

    return floats;
}

View more solutions

Comments

[makerofthings7]

I love linq as much as I don't understand it! Which is a lot! Since this will be a critical performance function, I'd like to use a system level object if possible. Also the extension .ToArray() doesn't seem to exist in my .NET 4 command line project. Do you know the namespace?

[Ani]

@makerofthings7: It's in System.Linq.Enumerable.

[R. Martinho Fernandes]

@maker: System.Linq, as always.

[makerofthings7]

@marthinho Yes, have System.Linq but ToArray() didn't appear. Odd

[Ani]

@makerofthings7: I was missing a parenthesis (fixed). That might have been the problem. Try now? Also, you might want to use the non-LINQ version if this is performance-critical. Profile both. :)

[makerofthings7]

last parameter is 4 for Array.Copy. +1 I would mark this as the answer to the question (since it addresses the question in the title), but Ani gave so much time into his answer. Only a 5 point difference anyway...

[Nicholas Terry]

Awesome solution!!!!! I was going to multiply the float by like 10000000 and shift it, thus losing some precision. This is even better!

[heltonbiker]

Instead of a literal 4 you can use sizeof(float).

[El Marce]

@NicholasTerry if you ever find yourself in the position of having to take the low-level road, instead of multiplying by a million, you can also store the float in its bit representation: en.wikipedia.org/wiki/Single-precision_floating-point_format