dolfin team mailing list archive

[Robert C.Kirby <kirby@xxxxxxxxxxxxxxx>]

I refuse to add C++ to the output list since I am not a swig developer 
and that is currently outside of swig's capabilities.  In gnu 
compilers, at least, the mapping from C bindings to C++ bindings is:

extern "C" {
#include "foo.h"
}

SWIG and other tools are incapable of inferring what the right OO 
hierarchy is unless the underlying code already is architected in such 
a way.  You can't automate such architecting.  On the other hand, one 
could do the OO design in some abstract environment (SIDL? UML?)  which 
could be translated into Python, C++, Java, but one *still* must tell 
what class system there is.

What SWIG would let you do is this:
1.) First, you wrap up all your C code into C++
2.) Then, you use swig to bind your C++ into any other language you 
choose.

On Oct 26, 2004, at 1:51 PM, Anders Logg wrote:

> I think you should add C++ to the output list. There are many people
> who prefer to live inside C++. It is also the only option if not all
> the tools are available (and I don't think you can always assume they
> are).
>
> If a tool is missing, you could of course create the tool in C/C++
> with an appropriate wrapper, but that's an extra step.
>
> /Anders
>
> On Tue, Oct 26, 2004 at 01:27:28PM -0500, Robert C.Kirby wrote:
>
> > My thought is that:
> >
> > input = { C, C++ }
> > output = { Python, perl, mzscheme, ruby, ocaml, pike, java }, but not
> > C++.
> >
> > Has Beazley updated it?
> >
> > On Oct 26, 2004, at 12:26 PM, Matthew Knepley wrote:
> >
> > > ridg@xxxxxxxxxxxxxxx (Ridgway Scott) writes:
> > > > If the issue is just a C++ interface for PETsc, then can't you
> > > > just use SWIG to do this? PETsc is a library, SWIG provides
> > > > wrappers, and off you go?
> > >     Yes, this is correct, although I thought SWIG was strictly 
> > > Python.
> > > We can find some way to do it easily.
> > >
> > >          Matt
> > > > 				Ridg