Re: bitwise comparators

[Robert Iredell Eachus <rieachus@earthlink.net>]

Brian Rogoff wrote:

> Hmmm, I'll have to go back and look, or simply ask you to expand on this,
> since from what follows I must have misunderstood your position. The time
> I *really* rely on nested procedures is when I'm instantiating a nested
> procedure as a parameter of some generic subprogram, and simulating
> downward funargs. Its rare that I feel I need them, and probably worse for
> me than for some others since it can be argued that I'm programming FP
> style in Ada.

      Exactly.  There is nothing "wrong" with ordinary nested procedures.  It is just
that
when you need context or locall declared types, the "right" abstraction is so often a
generic instantiation or task entry that a nested procedure has become a red flag to me.

Try this.  Take some random package body that contains nested procedures, and move all
the nested procedures out into the package body.  Does the package compile and run just
fine?

      Incidently, if you look through the gnat source,  I will probably appear to be the
most
egregious violators of this rule.   (My sins appear in the run-time packages for the IS
Annex,
in processing Picture strings.  However, recursive descent parsers are a very special
case...