This is the mail archive of the
gsl-discuss@sources.redhat.com
mailing list for the GSL project.
Re: C++ wrapper
- To: gsl-discuss at sources dot redhat dot com
- Subject: Re: C++ wrapper
- From: "E. Robert Tisdale" <edwin at netwood dot net>
- Date: Sun, 26 Nov 2000 20:34:41 +0000
Ivo Kwee wrote:
> GSL is a great library
> but math-intensive code in C++ is rather unreadable
> using the standard C-namings.
>
> I am thinking to start coding up a simple (?) C++ wrapper
> providing operator overloading etc.
> Why has no one started this yet?
The GSL library developers have not yet finalized the API.
If you implement a C++ wrapper on top of the GSL API,
you should expect to revise you wrappers to accommodate
GSL revisions at least until version 1.0 is released.
> Anyway, I would be OK to code something up
> but I need some ready established interface specification.
>
> Earlier, SVMT has been mentioned.
> Is this the "best" one to go for?
Yes.
Mostly because it is the ONLY proposal for a standard API
besides the one proposed by the BLAS Technical Forum
http://netlib.uow.edu.au/utk/papers/blast-forum.old/blast-forum.html
> Anyone knows "better" or more "accepted" standards?
> I was thinking myself
> if MatLab-like conventions would be better
> but they are not really OO style. Suggestions?
There are serious problems with the existing Matlab API.
Matlab supports just one type: a double precision matrix.
This simplifies the interface and is convenient
for developing short application programs
but it creates problems when you try to implement
larger more complicated application programs.
The MathWorks may eventually extent the Matlab language
to provide support single and extended precision floating-point
as well as fixed-point (integral) types
and vector and third order and higher order tensor objects
as well as matrix objects so that numerical programmers
can write fast, efficient reliable applications in Matlab.