--- Ovid <[EMAIL PROTECTED]> wrote:
> --- Richard Jelinek <[EMAIL PROTECTED]> wrote:
> > Our NLP/NLU application does run on both environments, as long as
> we
> > do not use the Yaswi module (newest version fails with the 4.0.9
> on
> > production as well as with 5.5.x if we want to put that on
> > development
> > system).
>
> From what I understand about Yaswi and SWI-Prolog, I know it runs
> runs
> on 5.1.x and 5.2.x and the 5.1 series is the minimum. 5.5.x
> definitely fails.
Only versions 5.x.x are supported. It just need a minor change to
work on 5.4.x. Versions on the 4.x.x series are too incompatible, I
don see the point on supporting them.
About SWI-Prolog 5.5.x, the foreign interface has not been extended
yet to support unicode strings/atoms so I have to wait until it is
done.
> > You know - I'm almost ashamed to say this, because I haven't
> > contributed a single line to either project, but the
> > consolidation-driven manager in me feels a certain bitterness
> about
> > seeing double effort and two halfway-done projects.
>
> Believe me, I certainly understand that. I certainly had no
> intention
> of releasing my own, but I couldn't find anything else :(
>
> > Ovid: Is it really less work to push AI::Prolog to where you
> probably
> > want it to be (engine ported to C), than harness development (and
> > thus power) of the SWI engine and team up with Salvador?
>
> That depends, I suppose, on whether or not Salvador has the time.
> I
> realize that there's a lot of work involved. I would like to work
> on
> providing a cleaner interface, docs, tests, etc. to Yaswi, but I
> confess that my C is rusty -- anyone reading through
> AI::NeuralNet::Simple can tell you that :) In order for me to feel
> really confident, I'd need a bit of a roadmap on the C portion to
> know what's really involved in developing on that side.
I don't have too much time, but I promisse to fix any bug you report
to me, specially on the C side, and apply any patch improving the
docs, tests, etc. :-)
> Personally, I think Salvador's module is a much better starting
> point.
> Mine, however, is being actively developed and has no external
> dependencies. It's a tough call.
>
> Salvador, are you listening? :)
>
> > 3) interface. We had good experience with Inline::C, I just don't
> get
> > it, why it shouldn't be possible to embed prolog programs in
> Perl
> > exactly that way. (the HEREDOC of AI::Prolog comes nearest to
> this
> > requirement). So why no Inline::Prolog?
>
> I have not personally looked into doing Inline::Prolog. Others I
> have
> spoken with tell me that doing this would be a bit of a nightmare
> though targeting a C interface to Prolog might not be so bad. This
> sounds like the route that Salvador is taking. I suppose I should
> look
> into this again. Maybe it's easier than I thought?
I don't believe that an Inline::Prolog module would be comparable to
Inline::C
Mapping C to Perl is semantically obvious, there are several input
parameters and one output parameter and the functions are always
deterministic so it is the same in Perl and Prolog. All that is
required is converting the data formats.
On the other side semantically Prolog is completely different to
Perl, any predicate argument can be used as input, output or both,
and predicates can be undeterministic. How would you map member/2 to
Perl for example? how would you call it from Perl?
For a high level interface to Prolog I would better use something
similar to Perl DBI, or even the real DBI.
Cheers,
- Salvador
__________________________________
Do you Yahoo!?
Take Yahoo! Mail with you! Get it on your mobile phone.
http://mobile.yahoo.com/maildemo
