4 bytes array into float

sandycoolxyz

Hello,
I am using qt 4.8.5.
I have an array of 4 elements of _u8 type.
I would like to convert it into float. But I end up in wrong value.

= {1, 0, 0, 0} (hex values are 01 00 00 00)

qint32 value;
float value2;
for(i = 0; i < 4; i++)
      {
        value = (value << 8) + (data[i] & 0xff);                // if MSB is bit number 1
      }
value2 = (float)value;

If I out the value as int, the value I get is 16777216.
If I out the value2 as float, I get 1.67772e07. But I am expecting a float value 2.35099e-38(https://gregstoll.dyndns.org/~gregstoll/floattohex/).
How can I convert the 4 byte array into float?

jsulm

@sandycoolxyz Shouldn't it be like:

qint32 value = 0;
for(i = 0; i < 4; i++)
{
     value |= data[i]<<(8*i);
    // or (depending on byte order in the array
    value |= data[i]<<((4-i)*8);
}

Paul Colby

One option, assuming your array is always big-endian, and your code has to support both big an little-endian platforms:

    quint8 a[] = {1, 0, 0, 0};
    quint32 localEndian = qFromBigEndian<quint32>(a);
    qDebug() << *reinterpret_cast<float *>(&localEndian);

Output:

2.35099e-38

Cheers.

sandycoolxyz

Thanks Paul Colby and jsulm for the reply.
@jsulm I believe Its the same code as mine, just using addition instead of OR.

@Paul-Colby Thanks for the solution.

Cheers.

mvuori

Problems like that are exactly why people like to avoid bitwise operation when possible... For example, in this case many would prefer using an union of float & array for the conversion as it is more readable and maintainable years later...

jsulm

@mvuori But the union would not solve the problem with endianness.