01 Nov 2000 22:22:04 -0600
"Darren/Torin/Who Ever..." <email@example.com> writes:
> I suspect that having easy bindings for Perl, Python, Tcl/Tk, and
> Ruby pretty much says that we'll use SWIG.
If SWIG does a good job, then yes, but if it's not going to provide a
good solution, then I'd rather just see each interested group create
and maintain bindings for that language until/if/unless there's a
unified solution that actually DTRT. I don't know enoug about SWIG
these days to know the answer to that.
> P.S. SWIG also supports Guile but I doubt we care about that.
We originally used SWIG (this is becoming a FAQ :>), but at the time
it had a very poor guile implementation, and they seemed quite a bit
less than receptive to help on that front, so I went with g-wrap.
(Christopher Lee was great :>.). That's how we ended up here. Where
we go later is a different question...
Actually, using guile's kinda nice. We have a lot of control over
what happens, and how it happens. For example, I'm about to see about
adding both dynamic linking (a la "dlopen") so that you can just say
(use-modules (gwp gnucash))
(use-modules (gwp open-gl))
and have it suck in the system and g-wrap libs you need automagically
at runtime, and a fancier type system so that we can handle fancier
types (like glib containers) with more transparency. The latter is
tougher, and I need to do some head-scratching before I can tell what
the right answer is...
Rob Browning <firstname.lastname@example.org> PGP=E80E0D04F521A094 532B97F5D64E3930