(no object) aka gammaln

[frerejacque <daniel frerejacque free fr>]

Jody Goldberg a écrit :
> On Wed, Nov 12, 2003 at 11:32:27AM +0800, chengj wrote:
> > Hi !
> > who know the method of implemention of Excel's Function gammaln()?
>
> I doubt anyone knows the exact implementation of MS Excel's gammaln
> routine, but we do have a version in gnumeric can appears to produce
> comperable results.
>
>     gnumeric/plugins/fn-eng/functions.c:gnumeric_gammaln
> _______________________________________________
> gnumeric-list mailing list
> gnumeric-list gnome org
> http://mail.gnome.org/mailman/listinfo/gnumeric-list
------------------------------------------------------------
Beeing not involved in the maintenance nor development,
I feel I have few rights on gnumeric,
but the problem rised by the email from chengj seems fundamental:

If Excel is wrong on something,
there is no point to mimic that mistake.
                  ----------------------
If gnumeric implement "gammaln" (or some other function) it
should be done to give the best results from the point
of view of numerical computing.
Algorithms for special functions should be found easily.
"Numerical recipes" by F.Press and others,
is a good starting point,
but the implementer has to be careful to the multiple
traps with the implementation of special functions
(troncation, error propagation etc.).

Daniel Frerejacque
------------------