Boost C++ Libraries: Ticket #4107: ptr_map::mapped_type incorrect

status changed; resolution set

Thu, 04 Nov 2010 20:44:00 GMT

status new → closed
resolution → invalid

boost::ptr_map is not a drop in replacement for std::map. mapped_type is always a pointer, the container is designed to handle the memory management tasks of the heap allocated object.

anonymous — Thu, 04 Nov 2010 21:10:32 GMT

I know. But isn't ptr_map designed to provide the interface of std::map as close as possible? When 'reading' ptr_map, it looks like it contains (references to) objects. So it'd be great if mapped_type reflected that.

Fri, 05 Nov 2010 13:23:41 GMT

The value_type would have to be translated from std::pair<Key, T*> to std::pair<Key, T> which couldn't be done without copying T. I suppose a reference wrapper could be used but then the value type is std::pair<Key, boost::reference_wrapper<T> > which I don't think is an improvement.

anonymous — Fri, 05 Nov 2010 13:38:05 GMT

I'm not sure what you're trying to say. If you look at the code example, I'm returning a pointer, so no copy is necessary.

Fri, 05 Nov 2010 13:46:06 GMT

yes but you are taking the address of the value here '&i->second' hence the value_type (i) would have to be std::pair<Key, T>.

anonymous — Fri, 05 Nov 2010 14:42:31 GMT

The address operator works fine, that's not the problem.

Fri, 05 Nov 2010 15:04:07 GMT

For 'second' to be T and not T*, which is required in your code fragment, it would have to be a copy of the stored T*.

anonymous — Fri, 05 Nov 2010 15:12:07 GMT

Why can't it be a reference to T?

Fri, 05 Nov 2010 15:34:26 GMT

that would require std::pair<Key, T&>.

Fri, 07 Oct 2011 13:11:16 GMT

Just for the record: "T::value_type::second_type*" works.