Re: [PRD] Rule instances, refraction and Modify [Was: Re: Fwd: Clips behavior]

Gary,

Gary wrote on 02/02/2010 17:27:26:
> 
> > The transformation that you propose does not work, since ?y will 
> > always be bound to 1 and ?z to B, whether the instance matched foo(1) 
> > or bar(B) (and, thus, the two different instances will always have the 

> > same bindings for all their universal variables).
>
> I don't understand. Surely if the only fact is foo(1), then ?y is bound 
> to 1 and ?z is unbound. Similarly, if the only fact is bar(B), then ?y 
> is unbound and ?z is bound to B...

Unfortunately, that rule

> > Rule: FORALL ?x, ?y, ?z, IF test(?x) AND ( ?y=1 AND foo(?y) OR ?
> > z=B AND bar(?z) ) THEN...

is unsafe (since ?y and ?z are not bound in every disjunct).

The rule:

FORALL ?x, ?y, ?z, IF test(?x) AND ?y=1 AND ?z=B AND (foo(?y) OR bar(?z))

is safe, but the binding of the universal variables does not discriminate 
the two instances :-(

And I see no easy way to explain that ?y is not bound if the instance 
matches bar(?z); I mean, not easier than introducing the notion of matched 
facts, of something similar.

If we do not want to add the dummy (disjunct counter) variable, we can 
explicitely "annotate" each rule instance with the disjunct it matches in 
the disjunctive normal form (as defined for the purpose of checking 
safeness); if the rule is not disjunctive, the disjunct if the condition 
itself.

And two instances are equal only if they bind the universal variables to 
the same values, and if they matched the same disjunct.

That would mimick the way it actually works in most engines (if not all): 
that is, disjunctive rules are really handled as one rule per disjunct, 
with all the same action part...

Actually, I would prefer that: it requires a change in the definition of 
rule instance (esp. rule instances equality), but it is simpler (more 
intuitive) than adding the notion of matched facts, and it is less 
artificial than adding a fake universal variable just for differentiating 
the disjuncts...

What do you think?

Cheers,

Christian

IBM
9 rue de Verdun
94253 - Gentilly cedex - FRANCE
Tel. +33 1 49 08 35 00
Fax +33 1 49 08 35 10


Sauf indication contraire ci-dessus:/ Unless stated otherwise above:
Compagnie IBM France
Siege Social : 17 avenue de l'Europe, 92275 Bois-Colombes Cedex
RCS Nanterre 552 118 465
Forme Sociale : S.A.S.
Capital Social : 611.451.766,20 ?
SIREN/SIRET : 552 118 465 03644