28 Dec
2011
28 Dec
'11
5:33 p.m.
Quoting Simon Peyton-Jones
for example. Singleton unboxed tuples are a perfectly valid data type; it's just that we don't (now) have a name for their constructor.
Well, Haskell *does* have a mechanism for giving two different implementations to a particular name... class UnboxedUnit a where (# #) :: a instance UnboxedUnit (##) instance UnboxedUnit (a -> (# a #)) That only leaves solving the equivalent problem for the type-level name (# #). ~d