[SOLVED] Surprising behavior when writing some debugging code with prolog_trace_interception

Hi! In the code below you can find some simple predicates I created, that exhibit a surprising behavior.

The main predicate, explainSuccess is a simple debugging tool I’m writing to check out swipl introspection capabilities. If you pass explainSuccess a goal that succeeds, it will show you the instantiations of the variables.

I have several question on how I could improve this, but the first one is: at the bottom of the file I included two queries: one calls goal, and succeeds, the other calls (goal, true) and fails :exploding_head:

could someone explain how that’s possible? Is that because I’m using trace in the definition of the explainSuccess predicate?

:- use_module(library(clpfd)).
:- use_module(library(dcg/basics)).
:- use_module('reif.pl').

%% A simple explanation mechanism:

prolog_trace_interception(exit, Frame, _PC, continue) :-
    prolog_frame_attribute(Frame, level, Level),
           (Level = 11)
        -> (prolog_frame_attribute(Frame, goal, Goal),
        ;  (Level = 12)
        -> (prolog_frame_attribute(Frame, goal, Goal),
        ;  true
prolog_trace_interception(_Port, _Frame, _PC, continue).

reconstruct(Out) :-
    findall(Subclause, debugging_info(subclause(Subclause)), Subclauses),
    Out = (Clause :- Subclauses).

explainSuccess(G, Out) :-

%% A simple predicate to test:

zip([], [], []).
zip(As, [], []) :- dif(As, []).
zip([], Bs, []) :- dif(Bs, []).
zip([A|As], [B|Bs], [A-B|ABs]) :-

enumerate(As, Enumerated) :-
        length(As, L),
        zip(As, Ns, Enumerated).

%% The problem I described in the topic

%?- explainSuccess(enumerate([a,b,c,d], X), Out).
%@ X = [a-1, b-2, c-3, d-4],
%@ Out =  (enumerate([a, b, c, d], [a-1, b-2, c-3, d-4]):-[system:length([a, b, c, d], 4), lists:numlist(1, 4, [1, 2, 3, 4]), zip([a, b, c, d], [1, 2, 3, 4], [a-1, b-2, c-3, ... - ...])]) ;
%@ false.

%?- explainSuccess(enumerate([a,b,c,d], X), Out), true.
%@ false.

It seems to me it is not succeding, it says %@ false; am I missing something?

Thanks for looking into this @swi! Eventually it fails, but I have one solution there, right? Before the false there’s an assignment to X and Out. In fact I obtain false when I backtrack, searching for the second solution. But I don’t see why that first solution shouldn’t be reported in the goal, true query.

Ok, I found the problem, at last:
It had to do with the fact that I was hard-coding the level at which the calls I was interested happened!(initially I noted that every top-level goal seem to start at level 11, and so I was interested only in levels 11 and 12. But when I changed the implementation of the predicates, I added another layer, and that meant that my original code didn’t work)

At this point I’d just like to hear your general suggestions on how I could better implement the above code, if you have any.
Thanks! :slight_smile: