syntax for recursive predicate abstractions

[mcsoto]

What is the current syntax for recursive predicate abstractions?

test(X) :- P=( _(X):- X = 1 ),P(Y),print(Y).

This works fine.

test(X) :- P=( _(X):- X = 1 ; P(X) ),P(X),print(X).
test(X) :- P = {Member -> ''(X) :- X=1; Member(X)},P(X).

Even a syntax closer to the one in the paper doesn't seem to, I've tried it mostly in the webassembly app and it hangs (should this be moved to issues)?

[jfmc]

Thank you for reporting. Could you tell us the paper you are referring to?

Ciao predicate abstractions (like { ... }) are just terms. When they pass though some meta_predicate, they are reinterpreted as closures. Recursive definitions like those in your example create cyclic terms, thus that it hangs. Here is a simple workaround:

:- use_package(hiord).

:- meta_predicate p1(pred(1),?).
p1(X,X).

test(X) :-
    p1({ P -> ''(X) :- X = [] ; X = [a|Xs], P(Xs) }, P),
    P(X).

The code above works as follows:

• the p1/2 predicate reinterprets the first arg as an anonymous predicate with one argument, then unifies to the second arg
• the pred abstraction passed to p1/2 does not contain any cyclic term

Actually this is a limitation of the pred abstractions that we'd like to fix (e.g., by turning {...} into an special evaluable functor using functional notation). Also, predicate abstractions are currently represented as terms and dynamically interpreted. This makes them slower than named predicates.