Actual subprogram and generic instantiation in same unit

Actual subprogram and generic instantiation in same unit

[Oliver Kellogg]

To my amazement the following compiles:

-- file: inst.ads
with Generic_Package;  -- takes a parameterless generic formal procedure P

package Lump is

   procedure Manipulate;

   package Instantiation is new Generic_Package (P => Manipulate);

end Lump;


Poor compiler has to instantiate a package using a procedure
whose body it has not yet seen...

Is that considered good Ada style?
Are there cases where this lumping together could cause problems?

Thanks,

Oliver M. Kellogg

Re: Actual subprogram and generic instantiation in same unit

[Duncan Sands]

On Monday 27 October 2003 13:33, Oliver Kellogg wrote:
> To my amazement the following compiles:
>
> -- file: inst.ads
> with Generic_Package;  -- takes a parameterless generic formal procedure P
>
> package Lump is
>
>    procedure Manipulate;
>
>    package Instantiation is new Generic_Package (P => Manipulate);
>
> end Lump;
>
>
> Poor compiler has to instantiate a package using a procedure
> whose body it has not yet seen...

Why should that be a problem?  After all, it has all the information it needs
to generate calls to the procedure.

Duncan.

Re: Actual subprogram and generic instantiation in same unit

[Oliver Kellogg]

Duncan Sands <baldrick@free.fr> wrote in message news:<mailman.227.1067260813.25614.comp.lang.ada@ada-france.org>...
> >
> > Poor compiler has to instantiate a package using a procedure
> > whose body it has not yet seen...
> 
> Why should that be a problem?  After all, it has all the information it needs
> to generate calls to the procedure.
> 

As a layman, I would assume this is the hard part of implementing
generics in Ada compilers - it's what sets them apart from simple
text substitution robots.

From my personal experience I would still advise against such a
usage; it's a test for the compiler when the scenario gets more
complicated, in particular when the compiler implements generics
with the text expansion approach.

--Oliver