on Thu Dec 25 2008, Arno Schödl <aschoedl-AT-think-cell.com> wrote:
Inviting users to overload in your namespace is simply not a viable customization approach in general, so ADL is it.
O.k., agreed.
Why do you think ADL is o.k. for customizing boost::partition_point (soon std::partition_point, N2666), but not for customizing std::lower_bound? AFAIK, the same arguments apply to both.
Because nobody wants to customize lower_bound, upper_bound, equal_range, and binary_search when they could instead customize partition point alone.
I thought about what you are really after. You are probably envisioning future standardization,
I wasn't thinking about standardization.
where partition_point is a fundamental operation similar to swap,
Yes
and lower_bound/... implementations must invoke ADL for partition_point, just like C++0x requires it for swap.
Yes
As a start, we are now adding such a lower_bound/... into boost. Is that your thinking?
Sorta. -- Dave Abrahams BoostPro Computing http://www.boostpro.com