Re: [glob2-devel] Nicowar desync

[Nuage]

> Look, I don't have the time nor the patience to learn fixed point
> decimal algorithms, of which I would need many to be able to work. Its
> quite a waist to make my own integer class especially when their are
> many already made, and many of those could be easily imported into the
> source. Especially the rational class from boost, which would single
> handedly solve my problems. Many boost things would solve my problems,
> I certainly don't see why not using boost is such a big problem, when
> the usage of boost has such substantial benifits. Boost can almost be
> garunteed to be on every linux system, and windows is definitly not a
> problem, boost has portability in mind, and for the one (there would
> be only one) person making the compilation on windows, installing
> boost for the compile would be a snap.
> 
> I've hinted at it many times already, Boost is a good thing. Can we
> please drop the dependancy issue against boost, and have it before
> 1.0? It would make life, my life in particular, allot easier, and the
> code would be plenty more beautifull.

Fixed points is nothing to learn about it's just a way to use integers in some
cases. A kind of trick.

Example. If you want to make it simpler, let's use power of 10, and a 2 decimal
precision. In this case an integer with value "100" is interpreted by humans as
"1". Here you can typecast anything to fixed by multiplying by 100.

int intToFixed(int i) {return i * 100;}
int fixedToInt(int f) {return f / 100;}

And instead of "100" power, using a power of 2 can slightly improve speed in
some cases.


Now about boost dependency. I don't really know about the extra work and
complexity it adds, but I have nothing against it. Note that I make no work at
all to maintain the packages and other things like this related to the
dependencies. Maybe someone else will disagree. What about the boost macosx
support by the way ? So just tell us about boost so we can agree or not.