On Thu, Apr 2, 2015 at 7:15 AM, Matt Oliveri <[email protected]> wrote:

> *Sigh*
> I wonder how Shap will feel about this tangent. This really has
> nothing to do with BitC anymore.
>
> Really, "Why is it not a type?" is an odd question. When you learned
> to count, did you think numbers were types? Why should the natural
> numbers be types?
>

Well, I'd sort of like this to move to a different subject heading damned
fast so that we can find the discussion later, but the discussion itself is
no problem.

As I've already mentioned, BitC is going to adopt a Nat kind. One reason we
want natural numbers to be able to be interpreted as types is that it gives
us more sensible constructions for array and vector types that allow us to
check and elide bounds checks for many implementations of library-level
collection types.

And anybody who has looked at the performance of Java or C# collections
will understand that eliminating those bounds checks actually matters! To
real systems programmers, even.

I can't resist noting, however, that William can't count. In his count of
messages he lost track of a couple of orphaned subject lines that should
have been included. :-)


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

Reply via email to