On Thu, Nov 4, 2010 at 5:19 PM, Ben Kloosterman <[email protected]> wrote:

>  Aren’t regions good for  heap but problematic for static data ( and maybe
> stack) ? tagging all  static data would be expensive ( though static
> immutable data is often const or readonly so maybe there is some way)
>

I don't see why - the tagging in question happens entirely at compile time;
it's merely a labelling discipline. Yes, there are languages in which there
is a first-class notion of a region at run time (typically a heap region).
That's a fine idea, but it's also a great example of how the term "region"
has come to be overloaded in the literature.


>  > Other collections should be indexable, and it should be possible to
> construct nat-indexable collections that do not use arrays as their
> implementation. But "vector as reference to array whose size is not part of
> its type" is still a very useful concept.
>
>
>
> I suppose it depends on how dependable types work out eg the dynamic array
> discussed in the ring buffer . If it is handled well by dependable types I
> see no purpose for vectors in the language ( obviously it will be in the run
> time)
>

Sorry, but do you mean *dependent* types?


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

Reply via email to