From: jsiek_at_[hidden]
Date: 2000-06-13 16:34:29
Dave Abrahams writes:
> Your version has the advantage of doing some concept checking on the
> wrapped iterators to make sure that they really support (some of) the
> operations required. It also has the advantage of being able to wrap
> a RandomAccessIterator so it looks like a BidirectionalIterator, with
> a truly restricted interface (and why would we want to do that?) My
> version has the advantage of being simpler, and working on VC6. If
> you can handle the VC6 and custom tags issues, I don't see a reason
> not to go with your implementation. We would need much more thorough
> tests than the ones I've written, though.
I just finished the VC6 port and Intel C++ using both the VC++ headers
and STLport. I've updated the vault.
The custom tag issue will be more difficult... as of right now I don't
see a way to handle that. The problem is that class template
specialization requires exact matches. Therefore if you have a
sub-class of std::random_access_iterator_tag, you'll get the default
unspecialized iterator_adaptor instead of the specialization for
std::random_access_iterator_tag :( I'll need to put my thinking cap
on...
As for testing... one thing I've been thinking that boost needs is a
good set of *generic* tests that cover all the STL concepts. I think
several people have peices of this coded already and it might not be
too hard to pull it together.
Cheers,
Jeremy