> > which may in turn be combined further. So in this view we are starting
> with
> > all arities equal to 1 (which is the curried view) and coercing the
> arities
> > *upwards* within the type rule for /app/.
>

This is precisely what the logic code I posted for type inference did,
although now I am thinking about the compiler passes, I think I do want an
'arity-abstract' type to indicate that the specialisation is possible.

I am also thinking about the opposite, collecting all the types used in the
different applications together at the definition, and specialising there.

Keean.
_______________________________________________
bitc-dev mailing list
[email protected]
http://www.coyotos.org/mailman/listinfo/bitc-dev

Reply via email to