Unexpected evaluation_error for CLP(ℤ) goal expansion

[triska]

With clpz3.pl from https://www.metalevel.at/clpz/scryer/clpz3.pl, I get:

?- clpz:goal_expansion(cis(A,B), Y).
caught: error(evaluation_error((clpz:goal_expansion)/2),goal_expansion/2)

Expected: Success, with

A = B,
Y = true.

[triska]

I now get:

?- clpz:goal_expansion(cis(A,B), Y).
Y = true .

However, I expect the binding A = B in addition to that.

(Please also see #237 for the unexpected blank symbol that occurs before the ..)

[mthom]

Yes, I figured. I'm not sure how to implement that in pure Prolog without scouring the variable list, and checking that the attached labels aren't equal. Otherwise, the toplevel will post many X = X goals, which I want to avoid.

[triska]

Ideally, the toplevel should use the variable_names/1 option of read_term/3 (or one of its bootstrapped built-in predicates) to read the query together with the variable names that are used:

http://www.complang.tuwien.ac.at/ulrich/iso-prolog/WDCor3#7.10.3

These variable names should then be used when the residual goals are displayed. A binding like X = X need not be emitted. Only created bindings and residual goals need to be displayed, and the variable names only need to be plugged in when printing the answer.

[triska]

Maybe an internal Prolog predicate that determines whether its argument is a free variable could help with this? It should succeed for WAM cells that point to themselves (if indeed variables are represented in this way in Scryer!). For such variables, it is not necessary to emit a binding.

[triska]

The binding works nicely now, so I'm closing this issue, thank you a lot!

Please see #253 for the crash that can now be observed with this query!