Re: [boost] [range] How to extend Boost.Range?

From: David Abrahams (dave_at_[hidden])
Date: 2005-09-21 07:05:20

Rob Stewart <stewart_at_[hidden]> writes:

> From: David Abrahams <dave_at_[hidden]>
>> Rob Stewart <stewart_at_[hidden]> writes:
>
>> > I've been thinking about this issue and wonder if the
>> > following wouldn't solve the problem neatly for both
>> > conforming and non-conforming compilers:
>> >
>> > // whatever/foo.h
>> > namespace whatever
>> > {
>> >    class foo
>> >    {
>> >       ...
>> >    };
>> >    // for conforming compilers
>> >    unspecified begin(foo const &)
>> >    {
>> >       ...
>> >    };
>> >    // for non-conforming compilers
>> >    namespace boost { namespace range {
>> >       using whatever::begin;
>> >    } }
>> >    // boost/range/begin.hpp
>> >    namespace boost
>> >    {
>> >       inline template <T>
>> >       unspecified begin(T const & range_i)
>> >       {
>> > 	 using boost::range::begin;
>> > 	 return begin(range_i);
>> >       }
>> >    }
>
>> Careful.  Many of the problems you can run into with
>> dispatching onlyshow up in the case where the definition of
>> begin you want to beselected *follows* the template that uses
>> it.
>
> Right.  That was the case you'd been discussing and I missed its
> application here.  However, if the author of whatever::foo
> defines, as a matter of convention or coercion,* begin() in
> namespace boost::range in the same header as foo, and doesn't
> invoke it before it's definition, then the problem order won't
> occur.
>
>> > Did I miss something?
>
>> Two-phase name lookup. 
>>   14.6.4 Dependent name resolution [temp.dep.res]
>
> LOL It seems you like quoting that now that you've found it!  

It's easy to find when it's in a recent Boost posting :)

> I
> do know about two-phase lookup, but I don't have to deal with it
> often so I don't always recognize the trouble it can cause a
> solution like this.
>
> Nevertheless, won't the problem only occur for non-conforming
> compilers and only when the definition of
> boost::range::begin(foot const &) is in a separate header

No, it has nothing to do with *definitions* and everything to do with
declarations.  That function could be defined in a source file.

> (provided the author/maintainer of foo follows the prescribed
> course when it's in the same header)?

No, it will occur anytime boost/range/begin.hpp is included before
whatever/foo.h.

>> Unless you arrange for boost::range to be an associated namespace
>> offoo, you're setting up dangerous #include order dependencies.
>
> Of course.
>
> * It would be documented as the only correct thing to do were
>   this an accepted workaround for non-conforming compilers.

It's not much good if it breaks the conforming compilers.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com