This deduction is not true in the presence of `coerce`. Specifically, with
newtype Age = MkAge Int x :: IORef Int
we can say
y :: IORef Age y = coerce x
But, I do think Ashley's idea might work for TestCoercion, instead of TestEquality. I'm not 100% sure without More Thought, but it seems reasonable.
Richard
On Jun 1, 2014, at 2:13 AM, Ashley Yakeley
The base package includes the class Data.Type.Equality.TestEquality, for types from which type equality can be deduced from values.
Given values ra rb of types "IORef a" and "IORef b", if they are the same value (i.e. ra == unsafeCoerce rb), then one can deduce "a ~ b". Therefore, IORef can be made an instance of TestEquality. The same is true for various other reference types.
I propose these instances in base:
instance TestEquality IORef instance TestEquality (STRef s)
-- Ashley Yakeley
_______________________________________________ Libraries mailing list Libraries@haskell.org http://www.haskell.org/mailman/listinfo/libraries