What do you predict this query will produce?

[jasonhemann]

I hope this is interesting---I was somewhat surprised. What do you expect this query will produce as output?

$ scryer-prolog -f
?- use_module(library(pairs)).
   true.
?- use_module(library(dif)).
   true.
?- dif(X,[]), pairs_keys_values(X,X,X).

As in, what will be the printed representation of the output? Obviously, the query would not succeed if we had the occurs check flag on.

[UWN]

?- X = [K-V], pairs_keys_values(X,X,X).
   X = [K-K], K = K-K, V = K-K. % nice!
?- X = [K-_], pairs_keys_values(X,X,X).
   X = [K-K], K = K-K. % nice!
?- X = [_-_], pairs_keys_values(X,X,X).
   X = ... . % ugly gigantic blurb

[jasonhemann]

I expected a rational term, I just didn't expect it to be that gigantic. Is there a reason that a user would want it to unfold that far?

It's not so bad for pairs_keys_values(X,_,X), but wouldn't the user prefer the more compact representation?

[jasonhemann]

Also that is an obviously better way to represent the constraint I had in mind.

[UWN]

In order to write the answer compactly you need a new auxiliary variable.

What the Scryer top level loop currently does, is to never introduce such auxiliary variables and to use a depth limit in stead. If you have a blam-ish term, you get a very large ugly but abbreviated term. Here is the minimal version:

?- X=s(X), T = t(X).
   X = s(X), T = t(s(X)).
?- T = t(X), X = s(X).
   T = t(s(X)), X = s(X).
?- T +\ ( X=s(X), T = t(X) ).
   T = t(s(s(s(s(s(s(s(s(s(s(s(s(s(s(s(s(s(s(s(...)))))))))))))))))))).
?- T +\ ( T = t(X), X = s(X) ).
   T = t(s(s(s(s(s(s(s(s(s(s(s(s(s(s(s(s(s(s(s(...)))))))))))))))))))).

[jasonhemann]

TY.