6 Aug 20:46 2013

## ScopedTypeVariables

Evan Laforge <qdunkan <at> gmail.com>

2013-08-06 18:46:26 GMT

2013-08-06 18:46:26 GMT

Occasionally I have to explicitly add a type annotation, either for clarity or to help choose a typeclass instance. Usually top-level type annotations take care of this, but sometimes it's convenient to only annotate a certain value, e.g. one argument of a lambda. I've noticed that while vanilla haskell is happy to allow me to put type annotations on variables where they are used (e.g. '\x -> f (x :: T)'), if I put it on the variable where it is bound (e.g. '\(x :: T) -> f x'), it wants me to turn on ScopedTypeVariables. I think ScopedTypeVariables is a nice extension and I'm sure it comes from a perfectly respectable family and all, but it feels like annotations on arguments comes in as a side-effect. Would it make sense to split it into a separate extension like TypesOnArguments so I can more accurately express my deviation from haskell2010 orthodoxy? Or is there some deeper tie between scoped type variables and annotations on arguments? Now that I think of it, it seems inconsistent that 'x :: A -> B; x a = ...' is valid, but 'x = \(a :: A) -> (...) :: B' is not. Doesn't the former desugar to the latter? And what about getting ScopedTypeVariables into haskell prime? As far as I know everyone loves it, or at least no one actually hates it :)