Re: Assertions in Ada

[Ken Garlington <kennieg@flash.net>]

Tucker Taft wrote:
> 
> Note that during the 9X design we did consider adding more complete
> assertion/invariant constructs, but dropped them in the desire to
> reduce the overall scope of the revision.

At least in the meetings I attended, there was also a significant
conflict among the attendees as to the goals, syntax, etc. of
such constructs. I never heard any significant coalition that
seemed to agree in any of these areas.

Even now, it appears that pragma Assert has some variations
among vendors (e.g. the example code in my paper, which works
under GNAT, has to have the pragmas moved to work with the
Aonix products).

> My biggest complaint about Eiffel (and Java, for that matter), however,
> is that they don't physically separate interface from implementation (except
> via the notion of "deferred" classes).  The normal response is that
> you can use a tool to "extract" the interface any time you want,
> but that seems to break the whole notion of "design by contract."

I agree, although maybe we're being too Ada-centric about this. It
does seem to complicate matters to have to deal with an extraction,
rather than "the thing" itself.
 
> --
> -Tucker Taft   stt@inmet.com   http://www.inmet.com/~stt/
> Intermetrics, Inc.  Burlington, MA  USA