Thorsten Ottosen <thorsten.ottosen <at> dezide.com> writes:

Andrew Hain skrev:

...

Two quick queries.

Would it make sense to have the iterator increment and decrement operators

and

...

begin() member function of ptr_vector< nullable<> > skip missing records so that applications have no need to check?

Well, it is always a design trade-off. If we did that, then people who need to know the index of the null value is in trouble.

I do think it would make sense to add some utility iterator/range:

boost::ptr_vector<T> vec; ... BOOST_FOREACH( T& e, boost::filter_nulls(vec) )

or something. I'm open to naming ideas

Something like that could be useful.

...

What is stopping pointer container classes from being used with opaque

types

...

as a compilation firewall?

Can you be more specific? I have some trouble knowing the precise problem.

I have been looking at replacing a homebrewed template class with ptr_vector. The homebrewed template supports opaque classes with a nasty trick that only allows the template to be used with types known to its definition; this is one of the reasons I am trying to replace it. During my test I created a clone template that calls a Clone() member function in each class and I got compiler errors when the class arguments to ptr_vector are undefined. Would opaque classes work with a separate out-of-line definition of each class?

Thanks

-Thorsten _______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost