Re: C/C++ programmer giving Ada95 a chance -- writing an emulator.

[tmoran@bix.com]

> Here's one way to use dispatching on tagged types ...
>    Op_Table : array (Opcode) of Op_Root
>     := (Op_Root (O0),
>         Op_Root (O1),
>         Op_Root (O2),
>         Op_Root (O3),
>         others => OU
>         );
>    ...
>       Func (Op_Root'Class (Op_Table (IR)));

> The main loop, on a sparc with gnat 3.12p -gnatp -O3, is 9 instructions
  It also works with Gnat 3.12p NT.  But both ObjectAda and Janus
print a series of "undefined opcode".  It seems to me that is
actually the correct behavior, since Op_Table is a table of Op_Root,
not a table of Op_Root'class.  I'd be inclined to report this as
a Gnat error.
  As alternatives, you might make Op_Table a table of
access-to-Op_Root'class or use a "case statement IR is".
The latter, of course, is explicitly what you were trying to avoid.

> I don't think it is any clearer or better than my original.
  Agreed.  It's slower, more complex, and, it seems to me, has
no benefits.