[boost] Re: C++ Standard Library proposal - Math functions forStatistics

From: Hubert Holin (Hubert.Holin_at_[hidden])
Date: 2003-04-16 09:29:49

somewhere in the E.U., le 16/04/2003

   Bonjour

In article <m3r883uztg.fsf_at_[hidden]>,
 Gabriel Dos Reis <gdr_at_[hidden]> wrote:

> Hubert Holin <Hubert.Holin_at_[hidden]> writes:
> 
> 
> [...]
> 
> |       So, again I must ask, is someone interested in what I could put 
> | forward,
> 
> I believe that would be most welcome.  Please continue!
> 
> -- Gaby

      OK, so I will code some functions and propose them for addition to 
what we have here.

      I'll start with the very simplest ones (log1p, expm1, cosm1) as I 
have a fairly clear idea (I hope ;-)  ) of what to expect from them and 
how to code them.

      Then perhaps cbrt and pow (though I would like to know the precise 
C99 wording to see what people expect from them, say as to the range of 
the integer exponent for pow or the behaviour of cbrt on negative reals 
(that ought to be obvious, but then again...)).

      Things will become really interesting with beta, erf and erfc.

      At that stage, I guess time will be ripe to discuss in more 
details what we want effectively out of the list of proposed functions 
and proceed, with these examples in hand.

      I will attempt to code for accuracy first, where its importance is 
expected (say around 0 for expm1), with a fall-back on standard library 
functions elsewhere.

   A bientot

         Hubert Holin