Re: Ragged Array Proposal

["Robert I. Eachus" <eachus@mitre.org>]

Nick Roberts wrote:

> I seem to get the impression that the cognoscenti consider this particular
> idea worthless. I am perfectly serious about it, however; I am completely
> convinced that it would be both workable and valuable (extremely so, in
> fact). I beseech anyone who might be interested to take a serious look; I
> would be delighted to have feedback (_any_ kind of feedback), either to me
> by e-mail or to this newsgroup.
 
   Not worthless, just excessive.  There are already two ways to do what
you seem to want: arrays of Ada.Strings.Unbounded.Unbounded_String, or a
generic
package that does what you want using Controlled types.  If you want to
supply a package like that it would probably be considered, if not for
the next standard, at least to be provided by all implementors.  (And if
you do it before the ARG meeting Monday, it might even get considered
there...)

   Note: I have implemented a slightly more general version of that
package, where my major concern was to provide extensible records. 
Hmmm.  What you want seems to be a generic package with a specification
like:

   with Ada.Finalization;
   generic
     type Index is (<>);
   package Ragged is

     type Element is private with null;
     type Element_Reference is access all Element'Class;
     type Ragged_Array is array (Index range <>) of Element_Reference;
    
     procedure Assign_To(R: in out Ragged_Array; I: in Index, E:
Element);
     function Dereference(R: Ragged_Array; I: Index) return
Element'Class;
   
   private
     -- implemenation defined...
   end Ragged;

   Any comments?

-- 

                                        Robert I. Eachus

with Standard_Disclaimer;
use  Standard_Disclaimer;
function Message (Text: in Clever_Ideas) return Better_Ideas is...