Should MVar be an instance of Data.RefMonad

[ivanperez-keera]

MVars with an initial value could be an instance of Data.RefMonad. However, the current functional dependency would not make this possible without "newtyping" the IO monad.

Would having this instance make sense?

[conal]

Perhaps, though I suspect it would be inconvenient to have to work with an alternative, newtype-wrapped IO.

I'd forgotten all about Data.RefMonad, and I haven't used the TypeCompose package for years.

[ivanperez-keera]

An alternative would be to have the instances in separate modules.

The bad side of this is that it would mean having orphan instances. It would be great to have named instances in Haskell. At the FPLab at Nottingham we had a number of discussions about this, although I don't know if anyone is actively working on it.