Niels Dekker - mail address until 2008-12-31 wrote:
I still wonder if this infinite recursion is a compiler bug as well. But anyway, it's clear to me now that there's a good reason to keep swap_impl out of the boost namespace. :-)
And secondly, swap_impl allows the user to specify the type of both function arguments by one single template argument. While the current boost::swap utility no longer does. For example:
it's still possible to get those use-cases as long as it doesn't involve specializing/overloading swap_impl, e.g. (please include some explanatory comments!:-) // various workarounds to avoid compiler bugs that make the compiler // find boost::swap before std::swap make us // - put this outside the boost namespace, // - use a using-directive rather than a using-declaration, // - and make boost::swap have two template arguments. namespace boost_swap_impl{ template<typename T> ...swap_using_adl... } // we provide swap_using_adl mostly because boost::swap // has two template arguments, so it's harder to specify the type // as an explicit template argument (but it could also potentially // be useful to avoid a bit of runtime slowness in a few situations). namespace boost{ using ::boost_swap_impl::swap_using_adl; // or "using namespace ::boost_swap_impl;" if it's necessary for some reason } Although I couldn't say how strong the use-case is; probably it's rarely if ever a good idea to specify *that* template argument explicitly. The function name already confuses me a little though: why should a function named boost::swap do ADL when the nearly equivalent name std::swap doesn't? The name swap_using_adl, besides that not everyone knows the acronym "adl", makes it less obvious whether it can find the generic std::swap implementation (through 'using' not ADL). But maybe the main function has to be named boost::swap to convince clueless people to use it instead of std::swap... -Isaac