On Tue, Feb 24, 2015 at 2:19 AM, Keean Schupke <[email protected]> wrote: >> Shap's proposal solves the problem. It allows arity-abstract functions >> as arguments, and it never introduces allocations. It doesn't use >> subtyping. > > This sounds impossible to me.
Care to try again to find a problem? > If my solution needs to allocate so does shap's. That's a strong claim. > We both have the same inlining and application operations on the > AST. Shap's solution doesn't rely on inlining. What's an application operation on the AST? > Remember subtyping is just a relation on types, that only admits values the > compiler produces valid results for. You can't just say "it's just a type checking thing" and expect an implementation to appear. You need to say how to implement it. Your subtyping allows strictly more programs than instantiation of arity-abstract types does. You need to make those work. _______________________________________________ bitc-dev mailing list [email protected] http://www.coyotos.org/mailman/listinfo/bitc-dev
