On Sun, 5 Jul 2009, joel wrote:

Jeremiah Willcock wrote:

...

Is there a standard way to get a function object that extracts a fixed element from a tuple? There is a boost::tuples::detail::get_class<N> that does what I want, but it does not have result_type and does not appear to be officially supported. isn't boost::fusion::at_c a proper PFO that works on tuple ?

Do you mean http://spirit.sourceforge.net/dl_more/fusion_v2/libs/ fusion/doc/html/fusion/sequence/intrinsic/functions/at_c.html>? That does not appear to be a function object, but a plain function. Also, shouldn't something like this (and getting the nth element type from a tuple) be in the tuple library itself? -- Jeremiah Willcock

On Sun, 5 Jul 2009, joel wrote:

Jeremiah Willcock wrote:

...

That does not appear to be a function object, but a plain function. Also, shouldn't something like this (and getting the nth element type from a tuple) be in the tuple library itself? More like : http://www.boost.org/doc/libs/1_39_0/libs/fusion/doc/html/fusion/sequence/in... and : http://www.boost.org/doc/libs/1_39_0/libs/fusion/doc/html/fusion/sequence/in...

But a quick look at the implementation, it's not a PFO even if it supports the result_of protocol.

I wanted something that would be a particular class, not a function object, for efficiency reasons. I ended up writing my own, but can still replace it if get_class is fixed up. It would be good to have it in 1.40 so I can fix my code in 1.40 that uses it. -- Jeremiah Willcock

...

I wanted something that would be a particular class, not a function object, for efficiency reasons. I don't follow you here. A polymorphic function object is a mere functor supporting result_of protocol. No efficiency penalty arise using them.

That's what I meant -- a function object (polymorphic or not), not a function template. -- Jeremiah Willcock