Reputation: 6589
Consider the following goals:
Goal forall (x y: I = I), x = y.
Proof.
Abort.
Goal forall (x y: tt = tt), x = y.
Proof.
Abort.
Both I
and tt
are members of singleton types. The former lives in Prop
, the latter in Set
. These are very uncomplicated types, so I wouldn't expect them to support rich equality types in any setting. Are these goals provable in Coq without resorting to the Uniqueness of Identity Proofs axiom?
Upvotes: 1
Views: 72
Reputation: 33519
A common sufficient condition is for the type to have decidable equality; see also this FAQ: https://github.com/coq/coq/wiki/The-Logic-of-Coq#i-have-two-proofs-of-an-equality-statement-can-i-prove-they-are-equal
Upvotes: 5