Re: 'Cyclone', a safer C--reinventing the wheel

[Brian Rogoff <bpr@bpr.best.vwh.net>]

On Tue, 27 Nov 2001, Mark Lundquist wrote:
> "Dmitry A. Kazakov" <dmitry@elros.cbb-automation.de> wrote in message
> news:3c034c88.494656@News.CIS.DFN.DE...
> > On Mon, 26 Nov 2001 10:42:54 GMT, "Mark Lundquist"
> >
> > >I do think that Ada or an Ada-like language should have something like
> > >template specialization / partial specialization in C++.
> >
> > I think it would be better to review tagged types and provide
> > user-written compile-time subroutines to reduce use of generics.
> 
> Actually, template "metaprogramming" doesn't really excite me

I admit, I find it pretty neat, but that's not what I want when I long for 
automatic instantiation. Once you have a combination of type inference and 
overloading its too easy to get non-termination as a feature of type
checking. If I really want template metaprogramming, better to just have 
Lisp style procedural macros. 

> (and I don't think generics should be macros anyway).

As an implementation strategy, its fine, but I agree that generics are a
different beast. 

> I'm interested in template specialization for the purpose for which it
> was originally designed, and also for trait programming.

Yes, and in an Ada like language generics should still obey the contract
model. 

> When you say "review tagged types", is that related, or an entirely
> different concern?  (I remember you were the one who wanted multiple
> dispatch a while back... :-)

MD is a nice feature, and I like it in Lisp/Dylan, but I think its a bit
much for a language like Ada. 

I'm of the opinion that genericity is the part of Ada that could use some 
enhancement. Once Ada gets some Java-like interface capability in its OO 
part, that should be enough. 

-- Brian