On Sun, Jun 7, 2015 at 4:26 AM, Keean Schupke <[email protected]> wrote: > I would argue that requirements belong with algorithms, and properties > belong in the type. > > Now where you draw the line is tricky, and you need to carefully consider > the whole problem domain.
I can go along with that terminology. > I general I think everything should have an interface defined by a type > class, so a range is any type that implements 'begin' and 'end' that return > the beginning and end of the range. As such range is a type constraint, that > can be inferred from a function type: > > f :: Range r => r -> r > > r can be any type, and the constraint propagates as required. This doesn't do anything like what I think Shap is talking about. For example, to a function that doubles an integer, you should be able to give the type: range[1,5)->range[2,10) Saying that if you give it an int (1 <= z < 5), you'll get an int (2 <= z' < 10). Come to think of it, because ints are discrete, it could be range[1,5)->range[2,9) in other words range[1,4]->range[2,8] _______________________________________________ bitc-dev mailing list [email protected] http://www.coyotos.org/mailman/listinfo/bitc-dev
