reading single float binary files

[John van Groningen]

>Does anybody have a Clean function that can read a binary file (Unix
>or Windows) of single floats as Clean Reals (double floats)? Thanks,

You could read the float as an Int (using freadi) and convert this 32 bit integer to two 32 bit integers that represent a 64 bit Real, and then convert these to a Real using the 'ints_to_real' function below.
The 'real_to_ints' and 'ints_to_real' functions should work on all platforms, but the order of the Ints depends on the byte order used by the processor.
The 'int_float_to_real' function does not handle all cases and has not been tested very well.

Or you could write such a function in C and use htoclean to call this C function.

Regards,

John van Groningen


real_to_ints :: !Real -> (!Int,!Int);
real_to_ints r = code {
        pop_b 0
        };

ints_to_real :: !(!Int,!Int) -> Real;
ints_to_real r = code {
        pop_b 0
        };

int_float_to_real :: !Int -> Real;
int_float_to_real f
        # e=f bitand 0x7f800000;
        # v=f bitand 0x7fffffff;
        | e==0
                | v==0
//                      = ints_to_real (0,f); // little endian
                        = ints_to_real (f,0); // big endian
                        = abort
                "int_float_to_real for denormalized number not implemented";
        | e<>0x7f800000
                # i2 = f << 29;
                # i1 = (f bitxor v) bitor (v>>3) + ((1023-127)<<20);
//              = ints_to_real (i2,i1); // little endian
                = ints_to_real (i1,i2); // big endian
                = abort
                "int_float_to_real for NAN or infinity not implemented";

Start
//      = ints_to_real (real_to_ints 1.1);
//      = ints_to_real (1023<<20,0);
        = int_float_to_real 0x3f800000;