Reputation: 43
Is there any way to apply simplifications to uninterpreted functions defined in z3, rather than the goals and subgoals ?
I have the following z3 code :
(declare-fun f (Bool Bool) Bool)
(assert (forall ((b1 Bool) (b2 Bool))
(implies b2 (f b1 b2))))
(assert (exists ((b1 Bool) (b2 Bool))
(not (f b1 b2))))
(check-sat)
(get-model)
And I get the following output:
sat
(model
(define-fun b1!1 () Bool
false)
(define-fun b2!0 () Bool
false)
(define-fun k!7 ((x!1 Bool)) Bool
false)
(define-fun f!8 ((x!1 Bool) (x!2 Bool)) Bool
(ite (and (= x!1 false) (= x!2 true)) true
false))
(define-fun k!6 ((x!1 Bool)) Bool
(ite (= x!1 false) false
true))
(define-fun f ((x!1 Bool) (x!2 Bool)) Bool
(f!8 (k!7 x!1) (k!6 x!2)))
)
It turns out that by applying rewrite rules to the definition of f, we can get that f is equal to the second argument (x!2) by the following derivation:
(f!8 (k!7 x!1) (k!6 x!2))
= (f!8 false (k!6 x!2))
= (f!8 false x!2)
=(x!2)
Is there any way to get z3 to produce the following definition automatically ?
(define-fun f ((x!1 Bool) (x!2 Bool)) Bool
(x!2))
Thanks for your help. Regards, Oswaldo.
Upvotes: 4
Views: 533
Reputation: 21475
One option is to ask Z3 to evaluate the expression (f x y)
where x
and y
are fresh Boolean constants. The eval
command will evaluated (f x y)
in the current model, and will produce y
in your example. Here is the complete example (also available online here):
(declare-fun f (Bool Bool) Bool)
; x and y are free Boolean constants that will be used to create the expression (f x y)
(declare-const x Bool)
(declare-const y Bool)
(assert (forall ((b1 Bool) (b2 Bool))
(implies b2 (f b1 b2))))
(assert (exists ((b1 Bool) (b2 Bool))
(not (f b1 b2))))
(check-sat)
(eval (f x y))
Upvotes: 4