On Tue, Dec 23, 2014 at 6:12 PM, William ML Leslie < [email protected]> wrote:
> On 24 December 2014 at 12:10, Jonathan S. Shapiro <[email protected]> > wrote: > > Oh. Somebody (William?) asked earlier whether readonly and mutable > modifiers > > made sense. The answer is yes, but uncomfortably. > > Oh, that's not what I was trying to ask. I meant, "are they an > interesting problem?" > I'm not an ivory tower guy. Problems are interesting only to the extent that the results provide value to humans. > Type checking in a system with subtyping involves more than > unification Yes. The thing that ultimately made me abandon the v0 code base was the conclusion that we could no longer avoid subtyping, and that rebuilding the type inference engine to deal with subtyping was a rewrite in any case. > These > mutability annotations - which I think I conceptually want to > associate with the static region variable, not the type - are > constraints, and their variance is important. > I'm not convinced that they are constraints, but it's certainly possible that we modeled them badly. Can you expand on why you believe they should associate with the region variable? > ... I think for informal discussion it makes more sense to talk about types > /matching/ - if I require a given type, does the expression provide > it? If not, complain. > Tempting, but that intuition breaks down quickly. A nearby intuition is "does the expression provide a *compatible* type (up to mutability)". That's exactly what copy compatibility (which we modeled as a constraint) dealt with. shap
_______________________________________________ bitc-dev mailing list [email protected] http://www.coyotos.org/mailman/listinfo/bitc-dev
