I'm trying to encode an arithmetic for positive reals with a constant infinity in Z3. I successfully obtained the result in SMT2 with the following pair encoding
(declare-datatypes (T1 T2) ((Pair (mk-pair (first T1) (second T2)))))
(declare-const infty (Pair Bool Real))
(assert (= infty (mk-pair true 0.)))
(define-fun inf-sum ((p1 (Pair Bool Real)) (p2 (Pair Bool Real))) (Pair Bool Real)
( ite
(first p1)
p1
(ite
(first p2)
p2
(mk-pair false (+ (second p1) (second p2)))
)
)
)
where a pair (true, _) encodes infinity while (false, 5.0) encodes the real 5. This works and I can solve constraints over it very fast.
I tried a similar approach with Z3py using z3 axioms over the following datatype:
MyR = Datatype('MyR')
MyR.declare('inf');
MyR.declare('num',('r',RealSort()))
MyR = MyR.create()
inf = MyR.inf
num = MyR.num
r = MyR.r
r1,r2,r3,r4,r5 = Consts('r1 r2 r3 r4 r5', MyR)
n1,n2,n3 = Reals('n1 n2 n3')
msum = Function('msum', MyR, MyR, MyR)
s = Solver()
s.add(ForAll(r1, msum(MyR.inf,r1)== MyR.inf))
s.add(ForAll(r1, msum(r1,MyR.inf)== MyR.inf))
s.add(ForAll([n1,n2,n3], Implies(n1+n2==n3,
msum(MyR.num(n1),MyR.num(n2))== MyR.num(n3))))
s.add(msum(r2,r4)==MyR.num(Q(1,2)))
print s.sexpr()
print s.check()
I can't get it to work (it times out). I guess the problem is in trying to prove the consistency axioms. However I couldn't find another way to encode my arithmetic in Z3py.
Is anyone aware of what is the equivalent of the Z3 SMT2 approach of above in z3py?
Thank you