Arno Schödl wrote:
So here is the implementation, as it is running in our production code. As far as I can see, there is no boost::partition_point, so I rolled my own. The code is only half-boostified. Ignore the make_XXX_range functions. BOOST_CUSTOM_PARTITION is currently a macro, because IMO that is the only way to be able to add it to existing iterators. If Boost.Iterator and Boost.MultiIndex are both willing to include it into their iterators, I am happy to make it a base class a la iterator_facade and integrate it into the existing code base.
Hi Arno, I've had a look at your code and am a bit puzzled about some details: Your BOOST_CUSTOM_PARTITION macro overloads lower_bound, etc for the iterator type of interest: as far as I can see, this is not necessary since boost::lower_bound could use partition_point through ADL, which is the only customization point we need. To put it more succintly, why can't you just have namespace boost{ template<typename _It,typename Val,typenameBinPred> _It lower_bound(_It itBegin,_It itEnd,const Val &x,BinPred pred){ using namespace boost; return partition_point(itBegin,itEnd,!boost::bind(pred,_1,x)); // ADL used } } and let users overload partition_point when needed? template<typename FilterPred,typename It,typename UnaryPred> filter_iterator<FilterPred,It> partition_point( filter_iterator<FilterPred,It> itBegin, filter_iterator<FilterPred,It> itEnd, UnaryPred pred ){ using namespace boost; return make_filter_iterator( itBegin.predicate(), partition_point( itBegin.base(), itEnd.base(), pred ), itBegin.end() ); } This would't have to use fancy overloading macros or base classes, and it covers all the use cases your current scaffolding does. Joaquín M López Muñoz Telefónica, Investigación y Desarrollo -- View this message in context: http://www.nabble.com/-MultiIndex--STL-algorithm-support-for-multi_index_con... Sent from the Boost - Dev mailing list archive at Nabble.com.