On Jul30, Claus Reinke wrote:
one could turn that promise into a type-checked guarantee by using explicit sum types (and thus structural rather than name-based typing), but that gets awkward in plain haskell.
I don't think the choice of whether you label your variants with names or with numbers (in1, in2, in3...) has anything to do with the choice of whether you require your cases to be exhaustive or not.
i was talking about name-based (as in: this is the sum type named List) vs structural (as in: this is the sum type build from Cons and Nil) typing. the former hides (in-)exhaustiveness from the type system, the latter exposes it.
I don't think the juxtaposition you're setting up here, between "name-based" and "structural", is correct. You could just as well choose the elimination form for "name-based" sum types (i.e., datatypes) to be an exhaustive case analysis. It just happens that Haskell chooses to permit non-exhaustive case expressions. -Dan