Re: [Boost-users] [Spirit] qi::parse at translation unit global scope?

Vitaly Budovski writes:

On 1 February 2012 20:33, Olaf Meeuwissen wrote:

...

I noticed this when I played around with wrapping qi::parse() in a parse() function that does some things before and after qi::parse().

My code had been calling qi::parse(), with the `qi::` namespace prefix, everywhere so I removed from a few calls and compiled before adding a parse() declaration.  Lo and behold, compilation unexpectedly succeeded. I've reduced the behaviour in a minimalistic sample program (attached).

I expect their to be *no* parse() function within the scope of main(). Can anyone clarify why the compiler has no trouble finding one and which one it finds?

Adding a global scope operator to the call, i.e. saying ::parse(), does result in the expected compile error.

The compiler uses Argument Dependent Lookup (ADL) to determine which function to call. One of the parameters is qi::rule, so the compiler looks in that namespace for a matching parse function.

Thanks. For anyone who needs a bit more info on ADL, here's a pointer http://en.wikipedia.org/wiki/Argument_dependent_lookup Based on that, the compiler fails to find a matching class member function (doh, we're in main() ...) for the unqualified call to parse() and then starts looking in qi which is an alias for boost::spirit::qi and finds the parse() free function in that namespace. I see. Hope this helps, -- Olaf Meeuwissen, LPIC-2 FLOSS Engineer -- AVASYS CORPORATION FSF Associate Member #1962 Help support software freedom http://www.fsf.org/jf?referrer=1962