Re: Ada (GNAT) and GNU Scientific Library

["Dr. Adrian Wrigley" <amtw@linuxchip.demon.co.uk.uk.uk>]

On Fri, 17 Nov 2006 13:28:23 -0800, george wrote:

> brian.b.mcguinness@lmco.com wrote:
>> I would think that it would be more useful to translate the library to
>> Ada
>> as a generic package, so any floating type could be used, but of
>> course it would then be necessary to maintain synchronization
>> with the original version.
> Which may not be such a trivial thing to do:
> 
> aldar portage # cat sci-libs/gsl/ChangeLog|grep \*
> *gsl-1.8 (11 Apr 2006)
> *gsl-1.7 (14 Oct 2005)
> *gsl-1.6 (16 Jan 2005)
> *gsl-1.4 (28 Dec 2004)
> *gsl-1.4 (18 Nov 2003)
> *gsl-1.1.1 (12 Jun 2003)
> *gsl-1.3 (14 Feb 2003)
> *gsl-1.2 (06 Aug 2002)
> *gsl-1.0 (1 Feb 2002)
> 
> Expect two new releases per year on average. If anybody is ready to
> undertake this, sure, but I think the more realistic aim would be to
> have a binding..

Yes.  Brian's comment suggests translating the package in order to
make the floating type generic(?).  Surely this isn't necessary?
The most you would want is a thick binding which chose the GSL code
for single or double precision according to the precision of the
generic parameter.  Gnat *may* be able to optimise the instance
to eliminate any dead code.

And since my name was mentioned higher up the thread, I'm not planning
on digging out the few bindings that I made for public use - they were
coded without much thought, just to try out GSL in my application. I
didn't use them in the end because the routines didn't meet my needs at
the time. Perhaps in the future, I will try again, because my needs and
the GSL have changed and expanded significantly.

A decent binding would probably best work out if someone in the Ada
community was actually using GSL in Ada regularly, since they
would have working, maintained code as a base.  If that's where
my coding is likely to go, I'll seriously consider putting it
on a public license from the start.
--
Adrian