max_depth/1 now incorrect

[UWN]

?- A = B*A, write_term(A,[max_depth(5)]).
   outputs("_601031* ..."), unexpected,  A = B*A.

[mthom]

This happens because of the cyclicity of A, not the max_depth setting.

[UWN]

But in the context of max_depth being positive, there are simply no cycles. So what is expected is what SICStus or GNU produce:

% SICStus:
| ?- A = B*A, write_term(A,[max_depth(5)]), fail.
_1185*(_1185*(_1185*(_1185*(_1185* ...))))
% GNU:
| ?- A = B*A, write_term(A,[max_depth(5)]), fail.
_280*(_280*(_280*(_280*(_280* ...))))

[UWN]

One consequence of this is that the internal representation is leaked to the outside. Consider the equivalent¹ query:

?- A=B*A1,A1=B*A, write_term(A,[max_depth(5)]).
_43*(_43* ...)   A = B*(B*A), A1 = B*(B*A1). % unexpected.

In both SICStus and GNU the very same text is written (in SICStus, the variable numbering is now different, so to get exactly the same result:

| ?- A=B*A1,A1=B*A, write_term(A,[variable_names(['B'=B]),max_depth(5)]).
B*(B*(B*(B*(B* ...))))

---

1 Strictly speaking, this equivalence only holds for **rational tree unification**, but that is the unification default here