Re: [Xsb-development] XSB crashing

[David W. <wa...@cs...>]

Terrance Swift writes:
 > We do have an auto_table declaration, which for Prolog files decides=
 which
 > predicates need to be tabled in order to remove loops in the static =
call
 > graph.  It might help, but you should be aware of two issues:
 >=20
 > 1) I don't know how well it works with HiLog (it may table all the H=
iLog
 > predicates).

I tried it with the progran Sudhakar sent and it didn't table
anything, and so there was still the loop.  I was somewhat surprised.
And, as Terry mentions, it's not very precise with HiLog. =20

The program is a little odd:

Z( A, B) :-
=A0=A0=A0=A0=A0=A0=A0X( B ).
Z(A, C) :-
=A0 =A0 =A0 =A0Y(A, B),
=A0=A0=A0=A0=A0=A0=A0Z(B, C).
/* Depends is reflexive */
Y(A,B) :-
=A0 =A0 =A0 =A0(
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0( atomic(A), A =3D B) ;
=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0( atomic(B),=A0B =3D A)
=A0 =A0 =A0 =A0 =A0 =A0).
X( _) :- fail.

The clause for Y/2 says that Y is essentially equality (on atoms).  I
would have thought what you would want was something like:

Y(A,A) :- (Y(A,B) ; Y(B,A)).

This clause would give the reflexive closure of other ways of defining=20=

Y.

Maybe what you want are predicates like:
transitive(parent).
..
reflexive(sibling).
..

which define properties of relationship names.  And then you could
have rules like:

Z(A,B) :- transitive(Z), Z(A,C), Z(C,A).

Y(A,A) :- reflexive(Y), (Y(A,B) ; Y(B,A)).

 > 2) As it is invoked by our file-based compiler, I don't think it han=
dles
 > cases in which a predicate in file A (say p/2) and a predicate in fi=
le B
 > (say q/2) call each other.

-David