🚀 go-pugleaf

In article <TORKEL.92Dec21175440@bast.sics.se> torkel@sics.se (Torkel Franzen) writes:
>(Mark Johnson) writes:
>
>   >This is a nice way of viewing things, but I don't know any Prolog that
>   >gives you all of the constraints C.  Consider the following program:
>   >p(a) :- q(X,X).
>   >q(Y, f(Y)).
>    ....
>   >That is, sicstus only returns constraints related to variables appearing
>   >the goal.  But the constraints that only have cyclic solutions may be
>   >associated with other variables.
...
>  However, your example doesn't have anything to do with circular terms
>in particular. You might as well define q by the clause q(b,b).

Only that the constraints that q introduces, viz. X=f(X),
are not satisfiable wrt the standard equality theory.
The answer Sicstus prolog returns gives the user no indication
of this.  BTW, I don't think this is a problem with Sicstus, only
that it doesn't solve all the problems involved with cyclic solutions.

Mark

P.S.  I'm interested in this approach.  I know of work by 
Hoefeld and Smolka ``Definite Relations over Constraint Languages'',
and Chen and D.S. Warren ``C-Logic of Complex Objects'' and
``Abductive Logic Programming'' that describe the decomposition
into a relational and constraint components. Any other pointers
to relevant literature?


Mark Johnson
Cognitive Science, Box 1978
Brown University