[Bug libstdc++/15910] can't compile self defined void distance(std::vector<T>, std::vector<T>)

["giovannibajo at libero dot it" <gcc-bugzilla at gcc dot gnu dot org>]

------- Additional Comments From giovannibajo at libero dot it  2004-06-11 00:01 -------
Subject: Re:  can't compile self defined void distance(std::vector<T>, std::vector<T>)

bangerth at dealii dot org wrote:

> Alright, so here things go: [...]

Thank you for the perfect analysys!

> I think there is not much we can do about this -- what the user would
> like is of course that std::distance is rejected right away, but we
> could
> only do this by putting more checks into the function signature (via
> SFINAE) that the type that was passed is indeed an iterator.
> Given the wide variability of iterator types, I fear that there is not
> very much we can do about things at this point.

Something like this could work, but it's totally untested:

/****/

// it's an iterator if it has a nested type called iterator_category
template <class T> struct is_iterator
{
    typedef char yes[1];
    typedef char no[2];

    template <class T> yes test(typename T::iterator_category* );
    template <class T> no test(...);

    enum { result = (sizeof(test<T>(0)) == sizeof(yes)) };
};

// or if it's a pointer
template <class T>
struct is_iterator<T*> { enum { result = 1 }; };

/****/

// handy SFINAE to be used in return types
template <bool B, class T> struct enable_if {};
template <class T> struct enable_if<true, T> { typedef T type; }

/****/

namespace std{
    template<typename _InputIterator>
    inline
    typename enable_if<  typename is_iterator<_InputIterator>::result,
        typename iterator_traits<_InputIterator>::difference_type  >::type
    distance(_InputIterator __first, _InputIterator __last)
    {
        .....
    }
}


> Giovanni, are you convinced that this isn't really a gcc problem?

Well, let's put in this way: when I look at the testcase for this bug, it
strikes me VERY ODD that it is not valid C++ code. If you name distance as
distance2 it works. The name clash between ::distance() and std::distance()
should not happen, it is not meant to happen (this is why there is a namespace
std in the first place!), and it is just a side effect of the complexity of the
language rules.

So yes, I think it is a problem in GCC, and I reckon it can be fixed pretty
much like I have shown.

Giovanni Bajo




-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15910