opencog-dev team mailing list archive

["Moshe Looks" <holism@xxxxxxxxx>]

---------- Forwarded message ----------
From: Moshe Looks <madscience@xxxxxxxxxx>
Date: Sun, May 25, 2008 at 9:20 AM
Subject: Re: [Opencog-dev] OpenCog Shell
To: linasvepstas@xxxxxxxxx
Cc: opencog@xxxxxxxxxxxxxxxx, opencog-dev@xxxxxxxxxxxxxxxxxxx


Linas,

> I've used swig in the past, but have been generally
> unsatisfied with results. It sounds good in theory, and
> works great for hello-world tests, but fails to  bind
> correctly to language features -- its not transparent,
> its very much in-your-face.  I remember spending
> many, many days writing swig binding code trying to
> get a C-language linked list converted to an ordinary
> perl array, and back. I eventually gave up when it became
> clear that swig simply didn't have enough of its guts
> implemented to make stuff like this transparent.
>
Sorry to hear that swig didn't live up to expectations. How long ago
was this? Its possible that things have improved since, as development
seems fairly active...

> Myself, I'm partial to scheme+guile.  Quite unlike swig,
> the interface is very, very transparent.  ...although you do
> have to write glue code for every function you want to
> export, it beats the pants off of trying to diddle swig-internal
> crap to accomplish the same thing --  swig is not transparent.
>
Well, my understanding is that Guile is C-only whereas swig at least
attempts to support some C++ constructs (obviously much more
difficult) is it possible that a C-to-swig interface would be easier
to manage than C++-to-swig?

- Moshe