findnsols/4 what is it supposed to do?

[Jean-Luc-Picard-2021]

The Ciao Playground gives me:

?- findnsols(3, X, between(1,6,X), L).
L = [1,2,3] ? ;
no

On the other hand SWI-Prolog gives me:

?- findnsols(3, X, between(1,6,X), L).
L = [1, 2, 3] ;
L = [4, 5, 6].

Bug or feature?

[jfmc]

I'd say that this is a feature and not a bug. According to SWI's documentation https://www.swi-prolog.org/pldoc/man?predicate=findnsols/4 the SWI version is based on Ciao but added the non-deterministic behavior to implement the chunk option in library(pengines).

Ciao's behavior sounds more natural to me for findnsols/3 but SWI's extension is definitively very useful, but adopting the full SWI semantics (e.g., allow nb_setarg/3 to change the chunk size) sounds a bit scary... I'd prefer a more controlled interface, like passing some reference to the query with some interface to modify the aggregation or search parameters?

[Jean-Luc-Picard-2021]

An interesting improvement over findall/3 in the wild,
would work for both SWI-Prolog and Ciao Prolog?

has_unique_answer(G):-findnsols(2,G,G,Sols),!,Sols=[G].

https://arxiv.org/pdf/1709.04302

But SWI-Prolog does copy the goal G before it runs
findnsols/4. Is this necessary? I don't find that the goal
G is copied in the ordinary findall/3. Only the template.

I think the above example gives a rational why copy is necessary.
Without copy the G might show a solution, depending on
implementation of findnsols/4 maybe the 2nd or 3rd solution,

this might prevent binding G with the 1st solution.
Does Ciao Prolog copy the 3rd argument of findnsols/4,
before it runs the 3rd argument? Or does it have other

means to prevent leaking some binding?

BTW: Such abominations are not really necessary if the Prolog
system has a library(sequence). One can combine limit/2
with findall/3. There is also no goal copy issue:

has_unique_answer(G):-findall(G,limit(2,G),Sols),!,Sols=[G].

In as far the Ciao Prolog behaviour of findnsols/4 becomes
also redundant if limit/2 and findall/3 are availablle. Does
Ciao Prolog have limit/2 ?