!standard 3.9.5                                          06-12-11  SI99-0003-1/03
!standard 15.49
!class binding interpretation 06-04-20
!status ARG Approved  11-0-0  07-06-01
!status work item 07-01-05
!status ARG Approved  10-0-0  06-11-17
!status work item 06-04-20
!status received 06-02-13
!priority High
!difficulty Easy
!qualifier Omission
!subject Support overriding indicators

!summary

Support queries of new overriding indicators for subprograms.

!question

The Amendment (Ada 2005) defines overriding indicators for subprograms; should these be added
to Asis.Declaration to allow access to this designation? (yes)

!recommendation

(See summary.)

!wording


Add a new clause after 3.9.4 (type Declaration_Kinds):

3.9.x type Overriding_Indicator_Kinds

Type Overriding_Indicator_Kinds classifies declarations and specifications having
an overrriding indicator.

type Overriding_Indicator_Kinds is (
  Not_An_Overriding_Indicator,

  No_Overriding_Indicator,          -- 8.3.1 (2)
  An_Indicator_of_Overriding,       -- 8.3.1 (2)
  An_Indicator_of_Not_Overriding);  -- 8.3.1 (2)


The comments list a reference to the definition in ISO/IEC 8652:2007 Ed. 3 for each kind
of declaration.


Add a new clause to the end of section 15:

15.xx function Overriding_Indicator_Kind

   function Overriding_Indicator_Kind 		-- 8.3.1 (2)
     (Declaration : Asis.Declaration)
      return Overriding_Indicator_Kinds;

  Declaration specifies the subprogram declaration to query.

  Returns the kind of Overriding_Indicator for the subprogram declaration.

  Returns Not_An_Overriding_Indicator for any unexpected Element.

  Expected Declaration_Kinds:
     A_Procedure_Declaration
     A_Function_Declaration
     A_Procedure_Body_Declaration
     A_Function_Body_Declaration
     A_Null_Procedure_Declaration
     A_Procedure_Renaming_Declaration
     A_Function_Renaming_Declaration
     An_Entry_Declaration
     A_Procedure_Body_Stub
     A_Function_Body_Stub
     A_Procedure_Instantiation
     A_Function_Instantiation



!discussion



!appendix

Subject: Summary of the proposals for updating the ASIS Standard for Ada 2005
From: Sergey I. Rybin
Date: Mon, 13 Feb 2006 13:55:06 +0300

Formally the ASIS revision process is supposed to be started only after
finishing all the technical work with the new revision of the Ada Standard,
but I think we can start some preliminary technical work with new ASIS
definition already now (all the changes in Ada definition are stable
enough to start the ASIS revision).

At AdaCore we have developed a draft version of the new ASIS specification
and most of the proposed changes are already implemented in the latest
version of GNAT/ASIS technology. The attached file contains the summary
of the changes in the ASIS Standard we have made to adopt the new Ada
features. The interesting thing I've realized when working on
our draft of the new ASIS specification is that if we would like
to keep the upward compatibility as much as possible and if we
would like to keep changes as small as possible,
the new Ada definition gives us very limited freedom in making changes
in the ASIS spec. It does not mean that we think that our proposals
are the only possible solution for ASIS 2005, but it gives us some
hope that we can do ASIS revision for Ada 2005 fast enough.

There is quite a natural mapping of the proposed changes in the ASIS
specification onto the existing set of Ada Issues used for Ada revision,
and the attached file is structured and ordered by AIs. The only
exception is the new values in the ASIS Element Classification needed
to represent the new Ada 2005 predefined attributes and pragmas (of course,
this could be also mapped onto AIs, but I do not think that it would make
sense, because this is the easiest and completely straightforward part
of the ASIS revision).

My main goal of sending this ASIS 2005 Draft summary is to start
the ASIS 2005 discussion. The document itself would hardly be a convenient
background for this discussion - we will have at least split it on
changes related to different language issues, we will also have to
find a way for documenting the revision discussion, probably we
may need a separate mailing list for it. But let's get started!

Sergey Rybin


.....

---------------------------------
-- AI-218 Overriding indicator --
---------------------------------

RM-06  8.3.1

* Add two boolean queries to Asis.Declaration:

--  15.#??? function Is_Overriding_Declaration
------------------------------------------------------------------------------

   function Is_Overriding_Declaration
     (Declaration : Asis.Declaration)
      return        Boolean;

------------------------------------------------------------------------------
--  Declaration - Specifies the subprogram declaration to query
--
--  Returns True if the declaration contains the overriding indicator of the
--  form "overriding"
--
--  Returns False for any unexpected Element.
--
--  Expected Declaration_Kinds:
--       A_Procedure_Declaration
--       A_Function_Declaration
--       A_Procedure_Body_Declaration
--       A_Function_Body_Declaration
--       A_Null_Procedure_Declaration
--       A_Procedure_Renaming_Declaration
--       A_Function_Renaming_Declaration
--       An_Entry_Declaration
--       A_Procedure_Body_Stub
--       A_Function_Body_Stub
--       A_Procedure_Instantiation
--       A_Function_Instantiation
--
--

------------------------------------------------------------------------------
--  15.#??? function Is_Not_Overriding_Declaration
------------------------------------------------------------------------------

   function Is_Not_Overriding_Declaration
     (Declaration : Asis.Declaration)
      return        Boolean;

------------------------------------------------------------------------------
--  Declaration - Specifies the subprogram declaration to query
--
--  Returns True if the declaration contains the overriding indicator of the
--  form "not overriding"
--
--  Returns False for any unexpected Element.
--
--  Expected Declaration_Kinds:
--       A_Procedure_Declaration
--       A_Function_Declaration
--       A_Procedure_Body_Declaration
--       A_Function_Body_Declaration
--       A_Null_Procedure_Declaration
--       A_Procedure_Renaming_Declaration
--       A_Function_Renaming_Declaration
--       An_Entry_Declaration
--       A_Procedure_Body_Stub
--       A_Function_Body_Stub
--       A_Procedure_Instantiation
--       A_Function_Instantiation

****************************************************************

Subject: Re: Summary of the proposals for updating the ASIS Standard for Ada 2005
From: Jean-Pierre Rosen
Date: Sun, 19 Mar 2006 08:15:35 +0100

Sergey I. Rybin a écrit :

> Jean-Pierre Rosen wrote:
>
>> I have reviewed Sergey's proposal, there is just one point where
>> I'd suggest a different solution.
>>
>> Instead of having Is_[Not_]Overriding_Declaration, shouldn't the
>> "[not] overriding" be considered a trait (and adding to Trait_Kind accordingly)?
>
> Actually you would need not two, but at least four values:
>
>    An_Abstract_Overriding_Trait
>    An_Abstract_Not_Overriding_Trait
>    An_Overriding_Trait
>    A_Not_Overriding_Trait
>
> I personally do not like the approach based on Trait_Kinds, so I
> have suggested boolean queries. For me, it looks much more clear.
>
> (I'm not insisting, I'm only proposing :)

I understand your point, and I am unsure...

1) It is true that traits are not very palatable. There should have been sets
of traits, instead of addressing combinations by providing more litterals.
But it's way too late.

2) I don't like adding more functions, especially boolean functions just to
check the presence of something. The "trait" approach has some merit, if you
consider the number of boolean queries that it replaces.

3) Maybe, there could be an orthogonal enumeration "overriding_trait_kind"
   and a single query:
   type Overriding_Trait_Kind is
      (no_overriding_trait,
       an_overriding_trait,
       a_not_overriding_trait);

Not the best of both worlds. Rather a mitigation of both drawbacks.

I'd like to hear the opinion of others. Anybody listening here?

****************************************************************

From: Randy Brukardt
Date: Friday, January 05, 2007   1:35 PM

The ARG meeting agreed on:

type Overriding_Indicator_Kinds is (Not_An_Overriding_Indicator,
   An_Indicator_Of_Overriding, An_Indicator_Of_Not_Overriding);

The intended semantics was that Not_An_Overriding_Indicator is returned
(1) if there is no overriding_indicator or (2) if the query is malformed.

However, this is different than every other ASIS query, where there are
separate literals for (1) and (2). While I'd rather raise an exception
for malformed queries, it's years too late for that. But mixing the
error case and a perfectly normal case is nasty.

I think the type should be:

type Overriding_Indicator_Kinds is (Not_An_Overriding_Indicator,
  No_Overriding_Indicator,          -- 8.3.1 (2)
  An_Indicator_of_Overriding,       -- 8.3.1 (2)
  An_Indicator_of_Not_Overriding);  -- 8.3.1 (2)

Pascal agreed that this is better, and we mutually agreed to return
this issue to the ARG for further consideration.

****************************************************************