On Fri, Mar 5, 2010 at 1:37 PM, Neil Groves wrote:

On Fri, Mar 5, 2010 at 1:29 PM, Robert Jones wrote:

...

I'm sure I'm being a bit dense here, but apart from handling native arrays what does boost::swap give you that std::swap doesn't?

Extensibility. You call the qualified boost::swap but you can have implementations of swap in a namespace along with the swapee (is that a word?!). This facility is hugely useful when creating your own types and libraries for which you desire swap support. Users that call boost::swap will automatically leverage your customized swap behaviour.

Ahh, so it's a universal solution? "swap(a,b)" would use ADL but not find std::swap, and "std::swap(a,b)" would not find via ADL, whereas "boost::swap(a,b)" does both? Ok, cool! Any many boost libs define a swap(), but not a std::swap()! - Rob

On Fri, Mar 5, 2010 at 10:30 AM, Joseph Gauterin < joseph.gauterin@googlemail.com> wrote:

It's really a replacement for the common idiom:

using std::swap; swap(a,b);

IIRC one should write 'using namespace std' to avoid artificially preferring std::swap over swap from other namespaces.

Which enables the use of ADL with a fallback to std::swap. Unfortunately, there are a number of compilers that choke on that idiom. boost::swap abstracts the idiom in a more portable way (and adds support for arrays).

While your point is completely correct, I think that it is an important point that boost::swap allows the use of qualified calls to swap while maintaining extensibility. This allows one to avoid using XXX::swap in header files etc. This is extremely useful when writing header only libraries for example. Neil Groves