On Feb 20, 2013, at 1:05 PM, Neil Groves <neil@grovescomputing.com> wrote:
On Wed, Feb 20, 2013 at 6:55 PM, Marshall Clow <mclow.lists@gmail.com>wrote:
I'm proposing to change the interface to:
template<typename InputIterator, typename OutputIterator, typename Predicate> std::pair<InputIterator, OutputIterator> copy_while ( InputIterator first, InputIterator last, OutputIterator result, Predicate p );
i.e, changing the return type to return both iterators.
I think that certainly a change needs to be made to accomodate the extra information. I wonder how wise it is to return a pair that when the InputIterator type is the same as the OutputIterator type that it could be mistaken for, and accepted by, range algorithms as input? (That's my fault - doh!)
Whoa. That's something I didn't know about. Boost.Range will take std::pair<Iter, Iter> as a range? Off to do more reading. ;-)
I had similar agonising moments with the range return types of many algorithms and eventually decided that I would provide a template parameter that the user could select to choose the return type. The boost::unique algorithm is one example of such an affected function. See http://www.boost.org/doc/libs/1_53_0/libs/range/doc/html/range/reference/alg...
Of course, this might be over-the-top and perhaps it is better to provide no such option. This judgement I leave in your capable hands. It is another option that could potentially provide backward compatibility.
What this means: * If you're not calling copy_while (or copy_until), then this change won't affect you. * If you're not using the return value, then this change won't affect you. * If you are using the return value, then you will have to change your code thus: Old foo = copy_while ( first, last, out, p ); New: foo = copy_while ( first, last, out, p ).second; * if you were not using these calls because they didn't return the input iterator, now you can.
Questions? Comments? (Except for the "How could you miss that when you wrote those routines?" - I've already asked myself that) Improvements?
My comments are above, but ultimately I think you are on the right track and that the change will not be a painful one.
Thanks for the feedback. I've checked in the proposed change (and tests!) as revision 83064. More comments welcomed. -- Marshall Marshall Clow Idio Software <mailto:mclow.lists@gmail.com> A.D. 1517: Martin Luther nails his 95 Theses to the church door and is promptly moderated down to (-1, Flamebait). -- Yu Suzuki