On Apr 12, 2006, at 5:36 PM, Paul Mensonides wrote:
-----Original Message----- From: boost-bounces@lists.boost.org [mailto:boost-bounces@lists.boost.org] On Behalf Of Doug Gregor
When we initially type-check lower_bound, we need to determine if there is an advance() that we can call. Both #1 and #2 have been seen, so we check them. Can we call #1? Sure, because every ForwardIterator is also an InputIterator. Can we call #2? Nope, our ForwardIterator is not guaranteed to be a BidirectionalIterator.
What happens if you have more than one seed?
The program is ill-formed and the compiler will report an ambiguity when type-checking the template definition.
Is the name resolution related to the seed is implicitly added to the concept requirements that are checked immediately prior to instantiation? Or is the name resolution done during normal instantiation (with a restricted set of possible bindings)?
The latter.
[*] Yes, instantiation can still fail if partial ordering of function templates returns an ambiguity. We can synthesize it all we want, but it has yet to actually bite us in practice.
Well, it seems like an reasonable tradeoff--though I guarantee that the above will bite eventually (which is no worse than what we have now).
Everything will bite you eventually, unless you give up specialization entirely. For more information on the issue, see: Jaakko Järvi, Douglas Gregor, Jeremiah Willcock, Andrew Lumsdaine, and Jeremy Siek. Algorithm specialization in generic programming: Challenges of constrained generics in C++. Note: Accepted for publication in PLDI, June 2006. Doug