On 11/14/06, Rhythmic Fistman <[EMAIL PROTECTED]> wrote:
> On 11/14/06, skaller <[EMAIL PROTECTED]> wrote:
>
> > > I'm with 3. floats are just tricky.
> >
> > Clearly distrib and many other related axioms don't hold,
> > but simply giving up isn't an option either.
>
> Ok, recognising that the axioms won't compose more than
> a few times "close" works fine. I used a similar thing
> for a mathlib I wrote a while back, occasionally adjusting
> the definition of close for various operations and sometimes
> even for different machines (ppc->386 lost some precision
> for some calculations).
>
> > > val result, error = witherror (x * (y * z));
> >
> > I think you would use a record:
> >
> > struct {
> > approx: float;
> > lobound: float;
> > hibound: float;
> > ...
> > }
> >
> > and just define + - * / etc etc on it using a typeclass,
> > then you could write algorithms which work for real numbers,
> > including this version of them .. I guess this is monadic
> > programming .. Felix does have typeclasses now :)
>
> That would work? Cool bananas. That reminds me, I have a
> type class question...
>
Here's something pretty neat/cute, thanks to Jimmy, for testing when
there aren't any further floats available thanks to ieee float
lexographical ordering.
http://www.cygnus-software.com/papers/comparingfloats/comparingfloats.htm
-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Felix-language mailing list
Felix-language@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/felix-language
