OK. Fair enough. But I was thinking if flat_set could be configured so that the underlying container (vector/ptr_vector) could be changed, that would be awesome. Anyway, that was just my thoughts. Many thanks for your patience and very clear answers. ________________________________ From: Andrey Semashev <andrey.semashev@gmail.com> To: boost@lists.boost.org Sent: Wednesday, July 18, 2012 8:12 PM Subject: Re: [boost] Proposal: Linear map On Wednesday 18 July 2012 08:01:09 Amir Ansari wrote:
2. Is there some specialization for flat_set with the same semantics as ptr_vector ?
I'm not sure I understand. Do you intend to store pointers in flat_set? If so, you can store std::unique_ptr's or shared_ptr's and specify your custom ordering predicate to order pointers by their pointees.
The documentation for boost::ptr_vector says that using the smart pointers inside a vector isn't ideal... it's one of the reasons why ptr_vector exists. I was wondering if that logic applies still when flat_set holds pointers and is there a solution available?
If you have unique_ptr, standard containers can hold exclusive pointers, which practically defeats the efficiency penalty of storing shared_ptr's. Pointer containers also provide other features but if all you need is store pointers in containers, unique_ptr and STL containers are enough. The same applies to flat_set. _______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost