Re: [boost] associative_vector: request for formal review

From: John E. Potter (jpotter_at_[hidden])
Date: 2000-06-01 15:42:19

On Thu, 1 Jun 2000, Mark Rodgers wrote:

> It suggests that there is probably some rule that could be used to
> switch between the two algorithms.  Either
> 
> 1.  size() < N, or
> 2.  distance(first, last) > M, or
> 3.  distance(first, last) / size() > P.

Since the iterators are input_iterators, there is a bit of a
problem with distance.  Dispatch on iterator_category would
allow it for forward and above.

> I would think that at least vector_multiset's ranged constructor 
> could take advantage of this alternative.

Yes, here there is a clear advantage to fill and stable_sort.

John