Re: unification

[Andy Wingo]

Hi Stefan,

On Tue 13 Apr 2010 13:55, address@hidden (Ludovic Courtès) writes:

> stefan <address@hidden> writes:
>
>> I did a small try to extend guile to handle unification. The result is in
>>
>> http:///c-lambda.se/gp.tar.gz
>>
>> In there is an example of a unification solution of the Einstein riddle.
>> The solution takes 150ms on my PC. Gnu prolog execute it in about 16ms.
>> I'm using a c-extension linked in which is a fun play with the tagging 
>> system 
>> in guile (ducks!)
>
> I haven’t looked at the code but that sounds like an interesting and
> useful project!

Agreed! I looked briefly at the code, but this is one of my many
ignorant areas.

I would also suggest implementing this in Scheme, as a macro. That way
you program as in prog.scm, but when it's expanded, it expands out to
the exact thing you need to do the job. A macro extends the compiler
with support for a mini-language -- something like prolog in your case.

>> By the way. I can help out as well. My main interest now is in type theory 
>> but 
>> I start to dig into guile now and I would probably be able to help out if 
>> you 
>> like in a short time, as my knowledge matures.
>>
>>  Are there any suitable tickets to close?

Besides the points Ludovic mentioned, if you are interested you could
try writing a type inferencer for tree-il. Do it as a weak map from
tree-il to some type representation, then propagate down to the leaves
and back. At least, that will work for many cases -- you might know a
better implementation strategy.

Have fun with Guile,

Andy
-- 
http://wingolog.org/