Re: Ada 2005 Language Designer

[Adam Beneschan <adambeneschan@gmail.com>]

On Wednesday, October 8, 2014 1:35:37 PM UTC-7, Jeffrey Carter wrote:
> On 10/08/2014 11:30 AM, Adam Beneschan wrote:

> > A package is not an object, neither in the sense defined by the Ada RM nor in
> > any other relevant sense.  You can't create multiple instances of it.  And
> > being able to create a variable object (or constant object, or expression
> > that produces an object) and use the dot-notation to call an operation of
> > that object is one of the things programmers came to expect of OO languages
> > that Ada 95 didn't give them; Package.Operation notation is not close to the
> > same thing.
> 
> Pkg as object was a standard technique for implementing OO designs in Ada 83, in
> which a pkg represented an object in the design. Multiple instances were
> sometimes implemented as instantiations of a parameterless generic pkg. That
> seems like a relevant sense of "object" to me.
> 
> Of course, you couldn't have components of such things, so they clearly weren't
> objects in the Ada sense.

And you couldn't pass it as a parameter to any other subprogram, nor return one from a function.  And you couldn't declare a variable of a class type, and then assign one or the other generic instantiation to that variable.  Those inabilities would make a generic instantiation of a package a very poor substitute for an OO-type object. 

But I suppose the original point, that Ada always had Object.Operation notation because you could use a package as an object, may make sense.  In fact, we could make yet another case: Ada has always had Object.Operation notation, as long as Operation was a record component name.  And then we could try to argue that a record component is a kind of operation.  That should be equally convincing.

                              -- Adam