support functional predicate definition syntax

[mndrix]

Ciao (with :=) and GRIPS (with <-) and Mercury (with =) provide special syntax for defining predicates that behave like functions.

In both GRIPS and Mercury, function definition implies determinism. I think Ciao makes a mistake here by allowing nondeterministic predicates to be defined with functional notation.

Using functional definitions should imply determinism. Mercury's = should be enough syntax. I've never seen =/2 as a top level functor. So, assuming --/1 performs decrement, our definition should be

fact(0) = 1
fact(N) = N * fact(--N)

which desugars into

fact(0, F) :- 
    !,
    F = 1.
fact(N, F) :- 
    succ(Nminus, N),
    fact(Nminus, F0),
    F is F0*N.

There's no protection against fact(-1) looping forever, but most functional languages don't protect against that either.