We should introduce aliased types

We should introduce aliased types

[Victor Porton]

I think we should introduce aliased types to allow self-referencing objects 
and objects which we can access for object's internal purposes:

type T is aliased
   record
      Ptr: access all T;
   end;

Every value of an aliased type should be aliased (whether we specify that 
the particular value if aliased or don't specify).

Then we could create a self-referencing object by a code (say in Initialize) 
like this:

declare
   X: T;
begin
   X.Ptr = X.Unchecked_Access;
end;

-- 
Victor Porton - http://portonvictor.org

Re: We should introduce aliased types

[Victor Porton]

Victor Porton wrote:

> I think we should introduce aliased types to allow self-referencing
> objects and objects which we can access for object's internal purposes:
> 
> type T is aliased
>    record
>       Ptr: access all T;
>    end;
> 
> Every value of an aliased type should be aliased (whether we specify that
> the particular value if aliased or don't specify).
> 
> Then we could create a self-referencing object by a code (say in
> Initialize) like this:
> 
> declare
>    X: T;
> begin
>    X.Ptr = X.Unchecked_Access;
> end;

It sometimes make sense to make only the private view of a type aliased.

-- 
Victor Porton - http://portonvictor.org

Re: We should introduce aliased types

[Victor Porton]

Victor Porton wrote:

> I think we should introduce aliased types to allow self-referencing
> objects and objects which we can access for object's internal purposes:
> 
> type T is aliased
>    record
>       Ptr: access all T;
>    end;
> 
> Every value of an aliased type should be aliased (whether we specify that
> the particular value if aliased or don't specify).
> 
> Then we could create a self-referencing object by a code (say in
> Initialize) like this:
> 
> declare
>    X: T;
> begin
>    X.Ptr = X.Unchecked_Access;
> end;

One situation in which we need aliased types (not a single aliased variable) 
is Initialize() procedure for controlled types.

-- 
Victor Porton - http://portonvictor.org

Re: We should introduce aliased types

[Randy Brukardt]

"Victor Porton" <porton@narod.ru> wrote in message 
news:lrgp2h$h3n$1@speranza.aioe.org...
...
> One situation in which we need aliased types (not a single aliased 
> variable)
> is Initialize() procedure for controlled types.

This is already true. Controlled types are by-reference, so it's legal to 
take 'Access of a subprogram parameter of the type. Specifically:

    procedure Initialize (Obj : T) is
    begin
         Obj.Ptr := T'Unchecked_Access;
    end Initialize;

is legal.

Did you actually try this before writing??

                                       Randy.

Re: We should introduce aliased types

[Victor Porton]

Randy Brukardt wrote:

> "Victor Porton" <porton@narod.ru> wrote in message
> news:lrgp2h$h3n$1@speranza.aioe.org...
> ...
>> One situation in which we need aliased types (not a single aliased
>> variable)
>> is Initialize() procedure for controlled types.
> 
> This is already true. Controlled types are by-reference, so it's legal to
> take 'Access of a subprogram parameter of the type. Specifically:
> 
>     procedure Initialize (Obj : T) is
>     begin
>          Obj.Ptr := T'Unchecked_Access;
>     end Initialize;
> 
> is legal.
> 
> Did you actually try this before writing??

Sorry, Randy. I've tried but not with a tagged type. Now I see it does 
already work with tagged types.

-- 
Victor Porton - http://portonvictor.org

Re: We should introduce aliased types

[Randy Brukardt]

"Victor Porton" <porton@narod.ru> wrote in message 
news:lrgnoc$di6$1@speranza.aioe.org...
>I think we should introduce aliased types to allow self-referencing objects
> and objects which we can access for object's internal purposes:
>
> type T is aliased
>   record
>      Ptr: access all T;
>   end;
>
> Every value of an aliased type should be aliased (whether we specify that
> the particular value if aliased or don't specify).

You do know that this is already true of almost all limited types (without 
any declaration)?

I.e.

    type T;
    type Ptr_Holder (D : access T) is limited null record;
    type T is limited record
       Ptr : Ptr_Holder (T'Access);
    end record;

is legal Ada.

It's not allowed for non-limited types because assigning the object would 
break the self-referential link (it would point to the wrong object 
afterwards).

And you can't directly declare a component as
    Ptr : access T := T'Access;
because of accessibility; the problem isn't whether T is aliased or not. 
"access T" is a library-level type, while T'Access might refer to a local 
object, so the access type could outlive the object (in particular, this 
component could be copied into an object of another library-level type), and 
that's not allowed.

Your idea would have to have the same restriction, so it wouldn't help any.

                                  Randy.

Re: We should introduce aliased types

[Randy Brukardt]

"Randy Brukardt" <randy@rrsoftware.com> wrote in message 
news:lrgqnq$4i3$1@loke.gir.dk...
> "Victor Porton" <porton@narod.ru> wrote in message
...
> And you can't directly declare a component as
>    Ptr : access T := T'Access;
> because of accessibility; the problem isn't whether T is aliased or not. 
> "access T" is a library-level type, while T'Access might refer to a local 
> object, so the access type could outlive the object (in particular, this 
> component could be copied into an object of another library-level type), 
> and that's not allowed.

Actually, you can do that so long as you use Unchecked_Access:

      type T is limited record
          Ptr : access T := T'Unchecked_Access;
      end record;

                                 Randy.

Re: We should introduce aliased types

[sbelmont700]

On Friday, August 1, 2014 2:54:06 PM UTC-4, Victor Porton wrote:
> I think we should introduce aliased types to allow self-referencing objects 
> 
> and objects which we can access for object's internal purposes:
> 

I would rather see "not aliased" subprogram parameters (and I guess also return values) to help do fun things with constructor functions.  At the very least it would make smart pointers safer, but I have lots of other dreams about limited types (though it would probably be too complex to really work).

-sb