Roland Richter <roland@flll.jku.at> writes:
Neal Becker wrote:
I like your suggestion, Roland. Here is my adaptation (I will have to add credits and copyright later). It counts wraps as you suggest. I found it easier to implement and understand by basing it on an integral offset from a base. It all becomes pretty simple by calculating a "realposition" which is wraps*size+position. This is like a virtual index into an infinitely long sequence.
Definitely, this solution is elegant and better readable. Plus, it passes the test I use for cyclic_iterator (in the sandbox) after a minor modification:
For unknown reason, MSVC chokes on
typename std::iterator_traits<BaseIterator>::value_type
twice ('iterator_category' : is not a member of 'operator``global amespace'''), but accepts
typename boost::iterator_value<BaseIterator>::type
Right. iterator_value uses boost::iterator_traits, which contains bajillions of workarounds for vc6/7 inadequacies.
I suggest to add your implementation to Boost.Iterator! Does this require a review, or mini review, if there is something like that?
A review would help, but isn't strictly neccessary. It requires at least some extra confidence on my part. If cycle_iterator is thoroughly tested, including concept checks, and the documentation is readable and complete, that will go a long way. I don't know the current status of these things... -- Dave Abrahams Boost Consulting http://www.boost-consulting.com