Re: [Sbcl-devel] *backend-info-environment* unused?

[Brian S. <br...@de...>]

Raymond Toy wrote:

>>>>>>"Brian" == Brian Spilsbury <br...@de...> writes:
>>>>>>
>
>    Brian> In good news, when I get the array type stuff sorted out (to allow
>    Brian> specialisation on non-primitive types), we'll be able to add a
>    Brian> universally specialisable array type, implemented like (array T), but
>    Brian> with an extra slot for a type-predicate, which could be checked at
>    Brian> high safety, and used for type-inferrence otherwise. So, a
>    Brian> (simple-array cons (*)) might be actually meaningful, and help the
>    Brian> type-inferrence.
>
>Doesn't it already support this for type inference?
>
>(defun foo (x n)
>  (declare (type (simple-array cons (*)) x)
>	   (fixnum n))
>  (values (car (aref x n)) (cdr (aref x n)) (aref x n)))
>
>(compile 'foo)
>
>(describe 'foo)
>=>
>Its result type is:
>  (VALUES T T CONS)
>
It does within the scope of the declaration, but if I were to use the 
array elsewhere it would turn into a (simple-array T (*)) [unless there 
were additional declarations...]

I'm not sure that it's a significant win to allow fully specialised 
array types, but it shouldn't be hard to do.

Brian