Re: any_function (callback) library

From: williamkempf_at_[hidden]
Date: 2001-04-10 08:37:18

--- In boost_at_y..., Doug Gregor <gregod_at_r...> wrote:
> On Monday 09 April 2001 09:48, you wrote:
> [snip]
> > Looks mostly good to me from this posting, but I need to look at 
the
> > actual code yet.  The only thing that jumps out is the assignment 
to
> > int.  This would allow code such as the following:
> >
> > any_function func(myfunctor);
> > func = 10; // clears out the target, but this line doesn't make 
sense
> >
> > I also worry about possible ambiguities resulting between this
> > assignment and the templated assignment, though I'm not sure what
> > interactions could cause this.  I'd lean towards eliminating this
> > assignment entirely, since clear will accomplish the same thing 
and
> > is more understandable in any event.
> >
> > Bill Kempf
> 
> I've found no problem with ambiguities thus far. So far I'm 
outnumbered, but 
> I like the ability to say:
> 
> any_function<...> func();
> func = 0;
> 
> Because it more closely mirrors a normal function pointer. Would a 
(runtime) 
> assertion change your mind?

The likelyhood of ambiguities is remote at best.  The likelyhood of 
misuse is greater, though still small.  However, any chance for 
accidental misuse should be very carefully considered and eliminated 
if possible.  A runtime assertion would help, but I'd still feel 
better if this were just eliminated.  The call to clear() just reads 
better, has no chance for accidental misuse and is intuitive, while 
assignment to 0 may not be.

Bill Kempf