Re: Access type to member procedure of instance (Object Oriented programming in Ada)

["Randy Brukardt" <randy@rrsoftware.com>]

"Adam Beneschan" <adam@irvine.com> wrote in message 
news:7f09e5c5-f118-4431-a3c2-4abd2145dc14@googlegroups.com...
On Wednesday, November 21, 2012 2:12:23 PM UTC-8, Randy Brukardt wrote:
...
>It occurs to me that if you're right, we didn't need "access protected 
>procedure/function" either,
>since that could also be implemented with wrappers, I think (there's a 
>difference in that a
>dereference that calls a protected subprogram is a protected action, but I 
>think the logic
>needed to make it a protected action could probably be stuck inside the 
>wrapper.)

Right, there is no implementation need for access-to-protected. But there is 
a difference here: the semantics of a protected call is a bit different than 
the semantics of an unprotected call, and there are cases where you need to 
make that visible. So there is a semantic reason beyond the implementation 
reason for having them separate.

In any case, it's certainly true that Ada 83 and Ada 95 tried to avoid 
requiring wrappers in any circumstances (and that may be why we have 
access-to-protected -- I don't remember any discussions on that), but Ada 
2005 and especially Ada 2012 have given up on that and definitely require 
them in a lot of cases. So much of the reason for convention Intrinsic and 
the like is really OBE.

But do note that there are some cases where even a wrapper can't bail you 
out. The reason you can't in general take 'Access of a subprogram in a 
generic body is because a shared implementation can't make that work, even 
if wrappers are used. (To take one example.) So we can't just get rid of 
most of those rules without some thought. And then it's questionable whether 
it would be worth the effort.

                                Randy.