From: Reid Sweatman (reids_at_[hidden])
Date: 1999-08-16 12:46:43
> I think that this would be the pattern for all methods that
> return iterators:
> call the method on the underlying data structure, and delete
> and redispatch
> if non-existing. At least for containers with stable iterators.
>
> Sure would be nice if there were some way to say this once
> and for all, as opposed
> to having to replicate the code for it everywhere.
>
>
> Hmmm.... that tickled a neuron. Sure enough, Stroustrup's
> D&EofC++ says that
> his original C with Classes had call and return methods where
> such hooks could be
> applied, and on page 268 he talks about how such method
> combination can be
> acheived manually - unfortunately, requiring code replication.
>
> So, since the feature is *not* in C++, maybe the right thing
> to do is to have an external
> program write the code for me.
Well, PERL makes a hell of a C++ preprocessor <g>. However, you could also
get away without having to hook all access methods by treating them the way,
say, the vector template treats the [ ] operator; if there's a valid object
there, it accesses it, else, it inserts one. Except in your case, it would
*delete* the nulled one it found. I suspect, though that this wouldn't
fulfill the things you really want to do with this class.