Successful Predicate not filling Variable that must be unified.

[Stephen Coda]

Hi everyone, thanks in advance to anyone who can help me here.

I'm new to Prolog and couldn't find an equivalent question to the one I'm posting so, given this knowledgebase:

t(pb(4)).

t(pb(4),L):-

member(pb(11),L).

tt(P,L):-

findall(T,t(P),R),

findall(T,t(P,L),R).

pv(A):-

tt(pb(4),[pb(A)]).

When I give the queries:

 ?- pv1(11).

true.

 ?- pv1(12).

false.

These are as I'd expect but, when I give:

 ?- pv1(A).

true.

This too is right except it doesn't bind anything to A for me. Can anyone explain what's going on here and how i can sensibly get the binding?

With trace I can clearly see the correct unification taking place, highlighted purple:

   Call: (6) pv1(_G19347) ? creep

   Call: (7) tt(pb(4), [pb(_G19347)]) ? creep

^  Call: (8) findall(_G19430, t(pb(4)), _G19432) ? creep

   Call: (13) t(pb(4)) ? creep

   Exit: (13) t(pb(4)) ? creep

^  Exit: (8) findall(_G19426, user:t(pb(4)), [_G19440]) ? creep

^  Call: (8) findall(_G19426, t(pb(4), [pb(_G19347)]), [_G19440]) ? creep

   Call: (13) t(pb(4), [pb(_G19347)]) ? creep

   Call: (14) lists:member(pb(11), [pb(_G19347)]) ? creep

   Exit: (14) lists:member(pb(11), [pb(11)]) ? creep

   Exit: (13) t(pb(4), [pb(11)]) ? creep

^  Exit: (8) findall(_G19426, user:t(pb(4), [pb(_G19347)]), [_G19440]) ? creep

   Exit: (7) tt(pb(4), [pb(_G19347)]) ? creep

   Exit: (6) pv1(_G19347) ? creep

So I just don't understand why there's no binding for A. This is all a simple instance of the code as I can get, the real version is validating a bunch of rules somewhat similar to 't' in the example against a list of things.

Thanks once again for any advice.

Steve.

[Kilian Evang]

On 04/21/2015 01:00 PM, Stephen Coda wrote:
> This too is right except it doesn't bind anything to A for me. Can
> anyone explain what's going on here

findall(Template, Goal, Bag) finds all solutions for Goal and collects
the corresponding instantiations of Template in a list (Bag). The way it
finds multiple solutions is that after each solution, it *backtracks*
and tries to find another. Backtracking undoes variable bindings, so
after findall/3 exits, Template and Goal are in the same state as before
it was called, only Bag has been instantiated.

> and how i can sensibly get the binding?

How about not using findall/3 and calling t/[1,2] directly?

[Stephen Coda]

Thanks, 

I can call the t[1/2] directly. Originally I was trying to avoid this as I wouldn't necessarily since I didn't know how many t[1/2]s there would be.

Is there any way to enumerate a complete set of solutions to a given goal? Can't think of an obvious way myself.

[Kilian Evang]

On 04/21/2015 03:16 PM, Stephen Coda wrote:
> Is there any way to enumerate a complete set of solutions to a given
> goal? Can't think of an obvious way myself.

Interactively, you can enumerate all solutions to a query by hitting the
; key afer each solution, see
http://www.swi-prolog.org/man/quickstart.html#sec:2.1.2 .

To make a list of all solutions, use findall/3, e.g.:

?- findall(A, t(pb(4), [pb(A)]), As).
As = [11].

[Stephen Coda]

Thanks Kilian,

I've changed tack to enumerate my rule tests explicitly, since there's no strong reason not to, it just seemed a little inelegant.

Steve