Re: Integer'Class

["Dmitry A. Kazakov" <mailbox@dmitry-kazakov.de>]

On Tue, 5 Oct 2004 02:10:55 +0200, Rick Santa-Cruz wrote:

> sorry for so many questions.... maybe that's easy to answer, but I don't 
> find any senseful answer to that:
> Why can't I use: Integer'Class, but I can derive a new class from Integer in 
> writing:
> type New_Int is new Integer range 1..100;
> 
> Isn't this unlogical? Or have I missunderstand something completly?

Not at all. It is quite logical, all types should have T'Class, (probably
T'Class ones too (:-)). But when Ada 83 was extended to Ada 95, a decision
was taken not to go that far. I suppose, because the amount of changes
needed was considered too large. So only tagged types may be grouped in
classes, presently. Constructors/destructors were implemented as a hack
etc. Alas.

As for New_Int, it is not just deriving from Integer. It is something like:

-- This is not Ada!
type <Anonymous_Subtype> is -- Deriving a subtype
   new Integer
      with null record              -- No extension
      with constraint range 1..100; -- Constraint on the values
type New_Int is new <Anonymous_Subtype>; -- Cloning the subtype

The result is a clone of a subtype of Integer. It is an unrelated to
Integer type. Therefore it cannot be put into Integer'Class, it induces a
class of its own. A related construct is:

subtype Sub_Int is Integer range 1..100;

This only puts a constraint, but produces no true new type. So you cannot
override Integer operations in Sub_Int, only overload (with nasty
consequences). This is a heavy heritage of Ada 83, which considered all
subtypes as constrained bases.

I do hope that Ada 2010 will have T'Class for all specific types, let's
see.

-- 
Regards,
Dmitry A. Kazakov
http://www.dmitry-kazakov.de