Re: Strange clause/2 predicate behaviour in SWI-Prolog

On 2006-03-24, Jose' de Siqueira <jose@xxxxxxxxxxx> wrote:
The following behaviour of clause/2 in SWI-Prolog is very strange, to
say the least:

Welcome to SWI-Prolog (Multi-threaded, Version 5.6.7)
Copyright (c) 1990-2006 University of Amsterdam.
SWI-Prolog comes with ABSOLUTELY NO WARRANTY. This is free software,
and you are welcome to redistribute it under certain conditions.
Please visit http://www.swi-prolog.org for details.

For help, use ?- help(Topic). or ?- apropos(Word).

?- clause((A,B),C).

A = _G187
B = _G188
C = _G187, _G188

There isn't much rule or consensus on the clauses you can access using
clause/2. The only rule is that clause/2 works on dynamic predicates.
Beyond that, all bets are off. This isn't a real problem.

Besides the usual exercise in Prolog meta-programming generally
resulting in very much broken meta-interpreters there isn't much point
in using clause/2 for anything but environment support predicates such
as listing (basically clause/2 followed by portray_clause/1).

SWI-Prolog is very generous, in the sense that clause/2 succeeds on
anything but predicates defined in C. Its up to the user to decide
how to use it. If you have a meta-interpreter handling ,/2 you
shouldn't be calling clause/2 as well and hope it says 'no'. So,
the code becomes

prove((A,B)) :- !, % things we meta-interpret
prove(A),
prove(B).
.....
prove(Head) :- % things with inspectable clauses
clause(Head, Body), !,
prove(Body).
prove(Head) :- % all the rest
Head.

Richard O'Keefe has provided a decent meta-interpreter implementation
capable of handling the cut. Have a look at that if you are serious on
meta-interpretation.

Cheers --- Jan

P.s. Current SWI-Prolog still has a rule (A,B) :- A, B. This is why
clause/2 succeeds. Actually this rule serves no purpose anymore
and could be deleted. I think I'll leave it in, just make students
aware of the issues :-)
.