I now think I can more or less see how to get single inheritance into BitC.
We won't infer it, because that wouldn't give a most general typing, but I
think we can handle it when it is user-specified. If we introduce
inheritance into the language, I believe that we should continue to separate
the definition of type from the definition of implementation. In this view,
a class definition is mainly seen as a type class definition.

Questions for the group:

1. How important is inheritance in the eyes of people here?
2. Can anybody identify an "acceptable cost" strategy for transcoding C++ to
BitC without it?
3. Can anybody identify a sensible way to interoperate effectively with the
Common Type System without single inheritance?
4. Given that this can be deferred, how much impact does its presence or
absence likely to have on the early evolution of the BitC standard library?
Is that good or bad?

To those of you coming at this from the Haskell and O'Caml (i.e. purely
functional) worlds, I'm particularly interested in your experience and
input, but a gentle reminder that BitC needs to be able to interoperate....

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

Reply via email to