Tracking variable information?

[jjtolton]

Update: I found a way to do this with clpz, so no need to roll my own CLP. But I am still interested!

I find myself in the odd position of possibly needing to track information about variables, such as the domain of the variable, constraints placed on the variable, and so on. It's a bit trickier than I thought so I was wondering if there is any prior art on this. I notice in particular that if I unify a variable, it seems as though my normal methods of tracking the data associated with it have become orphaned.

Has anyone had to deal with this before?

:- use_module(library(assoc)).
:- use_module(library(lists)).
:- use_module(library(lambda)).

data(A) :-
   empty_assoc(A0),
   put_assoc(_X0, A0, [domain([0,1,2,3])], A1),
   put_assoc(_X1, A1, [domain([a,b,c,d])], A).

get(A,K,V) :-
        get_assoc(K,A,V).

?- Vs+\(data(A),
        assoc_to_keys(A, [XA,XB]),
        maplist(get(A), [XA,XB], Vs)
       ).
   Vs = [[domain([0,1,2,3])],[domain("abcd")]].
   
?- Vs+\(data(A),
        assoc_to_keys(A, [XA,XB]),
        XA=5,
        maplist(get(A), [XA,XB], Vs)
       ).
   false.

?- Vars0^Vars+\(data(A),
                assoc_to_keys(A,[XA,XB]),
                term_variables(XA,Vars0),
                XA=5,
                term_variables(XA,Vars)
               ).
   Vars0 = [5], Vars = [].

[UWN]

Can you express the problems you encountered in the form of expected and unexpected queries instead?

[jjtolton]

Not only can I not, I'm not even sure I should! But I will try to describe what it would look like if I could. However, I am not certain if what I am asking is an anti-pattern and CSP implementations should be handled differently.

Suppose a series of predicates exist: var_data/2, var_data_next/3, and var_unified_t/2. These predicates are special because they are interrogating the variable itself. I say they are special because I suspect they are not "logical" -- meaning, when you reason about if A then B, A is true therefore B is true, I'm not sure if there is a logical formalism for if the variable A has been discussed or here are the various things that have been said about the variable A.

Here are some cases that would be expected and unexpected:

?- var_data_next(X, [], [data]),
   var_unified_t(X, false),
   var_data(X, Data),
   X=5,
   var_unified_t(X, true),
   var_data(X, Data1).
   Data=[data], X=5, Data1=[data]. % expected


?- var_data_next(X, [], [data]),
   var_unified_t(X, false),
   var_data(X, Data),
   X=5,
   var_unified_t(X, true),
   var_data(X, Data1).
   false. % unexpected

So if I could formulate my question in terms of expected and unexpected queries, that is what it would look like.