Cut, If-Then in CLP(Z)

[notoria]

The construct (->)/2 is used in fold_statement/3:

scryer-prolog/src/lib/clpz.pl

Line 3832 in 7e22c12

( List = [] -> Statement = 1

The construct !/0 is used intervals_to_domain/2:

scryer-prolog/src/lib/clpz.pl

Line 1715 in 7e22c12

intervals_to_domain([], empty) :- !.

Is it safe to remove rewrite it?

[notoria]

Could the monotonic mode remove this:

scryer-prolog/src/lib/clpz.pl

Line 3728 in 7e22c12

reify_(E, B) --> { var(E), !, E = B }.

Something similar to this:

scryer-prolog/src/lib/clpz.pl

Line 2984 in 7e22c12

{ var(E), !, \+ monotonic }.

[notoria]

Is !/0 missing here:

scryer-prolog/src/lib/clpz.pl

Line 3707 in 7e22c12

reifiable(E) :- var(E), non_monotonic(E).

Should the unification in the head be delayed:

scryer-prolog/src/lib/clpz.pl

Line 2316 in 7e22c12

single_value(V, V) :- var(V), !, non_monotonic(V).

[triska]

Thank you a lot for looking at this! It would definitely be nice to use better constructs in the implementation.

Regarding these concrete cases: Even though (->)/2 or !/0 play a role in all these cases, the situations are all quite distinct:

• l. 3832: this was added before Scryer benefited from first instantiated argument indexing which you generously contributed nearly 4 years ago in 657e4f1, and it can therefore now be written efficiently without using impure constructs
• l. 1715: removing !/0 would leave an unintended alternative
• l. 3728: yes
• l. 3707: no, due to the calling context in which the predicate is always used
• l. 2316: it need not, for the same reason as above.

[notoria]

Thank you a lot for looking at this! It would definitely be nice to use better constructs in the implementation.

It's no trouble.

For intervals_to_domain/2, this could be done:

intervals_to_domain([], empty).
intervals_to_domain([I|Is], D) :-
    intervals_to_domain_(I, Is, D).

intervals_to_domain_(I0, [], from_to(M,N)) :-
    I0 = M-N.
intervals_to_domain_(I0, [I1|Is0], D) :-
        Is = [I0|[I1|Is0]],
        length(Is, L),
        FL is L // 2,
        length(Front, FL),
        append(Front, Tail, Is),
        Tail = [n(Start)-_|_],
        Hole is Start - 1,
        intervals_to_domain(Front, Left),
        intervals_to_domain(Tail, Right),
        D = split(Hole, Left, Right). % can be moved to the head.

But depending on the context there might be some choice points.

I was hoping to change reify_(E, B) --> { var(E), !, E = B }. into reify_(E, B) --> { var(E), !, non_monotonic(E), E = B }. but reify_(E, B) --> { var(E), !, \+ monotonic, E = B }. could work too.

Could the monotonic mode be used here:

scryer-prolog/src/lib/clpz.pl

Line 1188 in 7e22c12

cis_goals(V, V) --> { var(V) }, !.

Is it possible for the second factor to be the integer and not the first:

scryer-prolog/src/lib/clpz.pl

Line 3171 in 7e22c12

linsum(A*N, S0, S) --> { integer(N) }, !, mulsum(A, N, S0, S).