Re: Subprogram Renaming

[Jonas Nygren <ehsjony@ehs.ericsson.se>]

Robert Dewar wrote:
> 
> Jonas asked:
> 
> "But it could be put in the private part, without 'exposing the renaming'.
> Robert mentioned before that this had a performance drawback but I can't see
> that this would be the case if you have a generic package."
> 
> No, the performance penalty of creating a body occurs only if the renaming
> is in the body. Actually, as Bob Duff pointed out, not all compilers
> will take advantage of the possible performance gain (equiavlent to
> automatic inlining) that completion-with-renaming can provide (GNAT does
> not, since it requires renaming capability at the linker level, a feature
> that not all linkers provide). However, this performance penalty can in
> any case be completely eliminated by use of pragma Inline so in the case
> where the renaming-as-body appears in the body it is a shorthand and
> nothing more.
> 
> In the private part it is a different feature, but I don't understand
> Jonas' comment above. Sure in Ada 83 you could put a renaming in the
> private part, and it wouldn't expose the renaming -- it also would not
> expose ANYTHING and could not be used by a client of the package and
> therefore would be completely useless!

Well, I don't know anything about Ada 83 and just a little bit more
about Ada 95 :-)

> 
> The ability to use renaming-as-body in the private part is a new Ada 95
> feature which is quite useful, sometimes it can completely eliminate a
> package body that would otherwise be required in Ada 83.

Sorry, I perhaps misunderstood youre earlier posting. I read this as you
were advocating that all renaming should be in the body and then I tried
to point out that there could be reasons for doing the renaming in the
private part of the spec. My motivation for this was perhaps not the best.



-------------------------------------------------------
--    Jonas Nygren
--    ehsjony@ehs.ericsson.se
-------------------------------------------------------