On Wed, Apr 1, 2015 at 3:23 AM, Keean Schupke <[email protected]> wrote:
> Hmm but that variable does not follow normal unification rules. Its not a
> type variable, but looks like a kind of expansion variable?

Right, variables like ?ar are not type variables. We haven't finished
deciding exactly what they are. (We have multiple working options.)
The unification rules might not be too bad, because you never end up
with anything but a plain variable in the surface syntax. But they're
not like unification for HM arrows.

What is an expansion variable? I found a couple of papers by Sebastien
Carlier, and from one of the abstracts, it doesn't sound like what
we're doing.

Conceptually ?ar is some type-level information about the concrete
arity an afn has. The only reason an afn is truly arity-abstract is
because its arity parameter is a variable.

> We agree that arity-aware types break paretricity right?

I don't see why.
_______________________________________________
bitc-dev mailing list
[email protected]
http://www.coyotos.org/mailman/listinfo/bitc-dev

Reply via email to