On Wed, Sep 8, 2010 at 2:48 PM, Steven Watanabe wrote:

AMDG

Robert Jones wrote:

...

I expected this to work

#include #include

typedef boost::fusion::vector V;

int first( const V & v ) { return at_c<0>( v ); }

but I find I need a "using namespace boost::fusion;". Why isn't this covered by ADL?

When calling a function template with explicit template arguments, ADL is only activated if at least one overload is found by normal lookup.

Thanks for that Steven - could you possibly give me a C++ Standard reference for that? Thx, Rob.

On Wed, Sep 8, 2010 at 5:57 PM, Steven Watanabe wrote:

AMDG

Robert Jones wrote:

...

On Wed, Sep 8, 2010 at 2:48 PM, Steven Watanabe

...

wrote:

...

When calling a function template with

explicit template arguments, ADL is only activated if at least one overload is found by normal lookup.

Thanks for that Steven - could you possibly give me a C++ Standard reference for that?

14.8.1/6: "... But when a function template with explicit template arguments is used, the call does not have the correct syntactic form unless there is a function template with that name visible at the point of the call...."

Once again thank you Steven - I've found the reference, in particular the very clear example, but the rationale behind it eludes me! - Rob.

AMDG Robert Jones wrote:

On Wed, Sep 8, 2010 at 5:57 PM, Steven Watanabe wrote:

...

14.8.1/6: "... But when a function template with explicit template arguments is used, the call does not have the correct syntactic form unless there is a function template with that name visible at the point of the call...."

That is just the oddest rule! Going back to my original example, I could make it work just by adding an arbritary declaration of at_c(),

#include #include

typedef boost::fusion::vector V;

template void at_c( const U & );

int first( const V & v ) { return at_c<0>( v ); }

which doesn't even need to be defined anywhere. If such a namespace-enabling template were to be removed, thus breaking the look-up used in my 'first()' function it would be a horrible bug to find!

I think the rationale is that the the compiler needs to know that at_c is a template in order to parse the expression. It's basically the same reason we need to use template and typename to disambiguage. In Christ, Steven Watanabe