> > I can't find such functions, but it is the only option left. Therefore, > > such a function has to be written, may as well be called qsprintf() or > > something. It doesn't need to be complicated, only handle "%d", "%s" > > and "%f", maybe with a few standard flags. There is a number of such > > functions in GNU Go, by the way. > > All I have found in GnuGO were wrappers for standard library functions. > Were you accidentally referring to those, or did I miss something?
I meant those, just didn't check how exactly they were written. > Anyway, I have rewritten utils_format_double() to a "direct" version > that only uses some math-library routines (most importantly log10), but > does the formatting itself. I'd use a few constants instead, since the number of digits after point is limited to 6 anyway. But the main thing for it is to work ;), so do it as you see fit. > Next problem now that the komi is stored > correctly, reading it back truncates the decimals. Without actually > verifying, I'm pretty sure that's because atof() expects a ',', but > finds a '.'. So if you don't object, I'll write another function > utils_parse_double() to be the inverse of utils_format_double(). Fine. > I'm not sure whether we need our own formatting for %d (or even %s), so > for the time being, it's probably easiest to just use > sprintf(str, "...%s...", ..., utils_format_double(f), ...); > instead of > sprintf(str, "...%f...", ..., f, ...); One important consideration is that utils_format_double() works with a static buffer. It's very inconvenient when you need to write several numbers in one string because you need to copy, free allocated memory etc. I think I'll write such a function when I finally get to working with Quarry again. > I also > want to specify utils_format_double() a little more rigorously. Up to > now, it used as many decimals as sprintf spits out, truncating zeros. My > preliminary version always uses exactly one decimal. Probably it's > better to use up to n decimals if needed, where n is either a constant > (5?) or an additional parameter. > E.g. 1 -> 1.0 > 1.5 -> 1.5 > 1.12345 -> 1.12345 > 1.1234567 -> 1.12346 > Thoughts? sprintf() formats with six digits after point by default. I think 6 is a good limit, so what you described above is fine except that 1.1234567 should be formatted as "1.123457". > > Please provide ChangeLog entries > > along with patches (if you use Emacs, `C-x 4 a' does everything but > > writing the actual text for you). > > I'm using vim - that's the masochist in me :) Well, this means more work for you, sorry. Paul
