Re: Assertions

[Ray Blaak <blaak@infomatch.com>]

"Nick Roberts" <nickroberts@callnetuk.com> writes:
>    pragma Assume (Condition);
>    pragma Verify (Condition);

I would also allow optional description strings that would be part of the
raised exception, accessible via Ada.Exceptions. Most importantly, the strings
would be evaluated only if Condition is False:

    pragma Assume (Condition [, description]);
    pragma Verify (Condition [, description]);

This would eliminate once and for all code like:

  if Debug and not Condition then
    Verify (False, "complicated string that should be evaluated only "
                   & "when necessary: " & DescriptionOf (currentObject));
  end if;

> For all four of these pragmas, if the test of the Condition is actually
> done, and it fails (evaluates to False), the predefined exception
> Program_Error is raised.  (If this exception is raised by a pragma which
> takes the place of a declaration, the exception is immediately propagated
> out the innermost enclosing block, and is never handled by an exception
> handler inside this block.)

I would suggest another exception name, so that one can distinguish the fact
that a programmer-defined check failed, as opposed to a mistake in
general. Maybe: Assertion_Error, or both Verification_Error and
Assumption_Error.

> It may well be better to use 'Assert' throughout rather than 'Verify'.
> Comments please.

Well, "Assert" gives rise to discussions as to what it really means,
especially in comp.lang.ada :-). The idea is to side-step the issue by letting
the programmer decide which type of assertion is desired. Whatever names are
chosen should clearly reflect the intention.

-- 
Cheers,                                        The Rhythm is around me,
                                               The Rhythm has control.
Ray Blaak                                      The Rhythm is inside me,
blaak@infomatch.com                            The Rhythm has my soul.