Out of curiosity, how are you detecting if a member function swap exists?

it detects at compile-time if the type of the argument(s) has a member swap function

In general, there's no way of knowing if a member swap function is more efficient than a single assignment. It isn't even for all the examples you gave - boost::array<T,N>::swap performs N boost::swaps, which, if T doesn't provide an efficient swap, results in N calls to std::swap, and 3*N assignments. In general I'm wary of using swap in a generic context unless the intent is to swap (i.e. not as a potential optimization of assignment).

The library is non-intrusive and generic, it will perform an efficient move in terms of swap for any type providing a member swap operation, e.g, std::vector, boost::array, user-defined types etc.