Re: Depth First Search of a Char_Matrix?

[Shark8 <onewingedshark@gmail.com>]

On Saturday, April 27, 2013 1:31:49 PM UTC-6, Simon Wright wrote:
> Shark8 writes:
> 
> > On Saturday, April 27, 2013 11:25:57 AM UTC-6, Jeffrey Carter wrote:
> >> 
> >> There is absolutely no reason to use access types (especially not
> >> anonymous access types) or addresses for this problem.
> >
> > Sure, but there's no real reason not to: the illustration was simply a
> > DFS on something that might-not-exist and so null/some-object map to
> > that perfectly fine. (It's an error to ask "what color is the piece at
> > X,Y?" when the element X,Y contains no piece; this situation must be
> > addressed somewhere, and access/null works.)
> 
> As a general rule of thumb it is a bad idea to use access types unless
> you Absolutely Have To.
> 
> To solve this problem, a discriminated record would have done just fine:
> 
>    type Cell (Occupied : Boolean := False) is record
>       case Occupied is
>          when True =>
>             Piece : Piece_Colour;
>             Alive : Boolean;
>          when False =>
>             null;
>       end case;
>    end record;

Doh! So obvious.

> 
> And anonymous access types are worse, see the recent problems with
> anonymous task access types.

I know; I was the one who started that thread -- and that's why I've been spending my recent time thinking about access types (which probably explains why that was what popped into my head first).