Question

I want to read in some data from a file. Say an integer:

fread(&var1, 4, 1, f);

Where var1 would be an integer. But then I got to thinking that this is not safe as there isn’t any guarantee that an integer is 4 bytes long. (I’m ignoring other issues like feof and ferror for the sake of this question).

I also soon realised that there were even more issues than just int size, such as the endianness of the system, and probably others which I haven’t even thought of.

So, what is the best way to ensure that you data being read in is interpreted properly? So far, the only thing I can think of is to just store the data as text rather than as binary data, read in the text, and convert it at run time. I would guess that no matter the solution though, if you wanted to ensure that it is portable, it would always involve some form of conversion anyway.

Thank you.

Answer 1

To avoid the size problem, you should be doing:

fread(&var1, sizeof(var1), 1, f);

If you’re worried that the size of int might vary between the platform that writes the data and the platform that reads the data, then you have a more fundamental problem. In this scenario, you should avoid using int, short, etc., and use the types defined in <stdint.h>, such as int16_t, uint32_t.

To deal with endianness issues, you should consider writing helper functions that explicitly write/read the individual bytes in a known order, such as:

void write_uint32_t(uint8_t *buf, uint32_t x)
{
    buf[0] = (uint8_t)(x >> 0);
    buf[1] = (uint8_t)(x >> 8);
    buf[2] = (uint8_t)(x >> 16);
    buf[3] = (uint8_t)(x >> 24);
}

All of the above applies only to integer types. For floating-point types, there is no perfect universal solution.