Neal Becker skrev:
I'm a bit confused.
I'm working with blitz++. It has
template<typename T, int N> class Array ...
which already has a begin(), end(), but I can't use them. I'm specializing boost::begin,end,etc to hide them (I don't want to modify the blitz Array class)
If I specialize range_begin:
namespace boost {
template<typename T, int N> inline typename blitz::array_iterator< blitz::Array<T,N> >::type range_begin (blitz::Array<T,N> & a) { return blitz::range_begin (a); } }
Then this doesn't work, it seems that boost::begin (blitz::Array<T,N>&) finds the original begin(), but if I do:
namespace boost {
template<typename T, int N> inline typename blitz::array_iterator< blitz::Array<T,N> >::type begin (blitz::Array<T,N> & a) { return blitz::range_begin (a); } }
it works as desired. My questions are:
1) Is this reasonable?
Compared to what?
2) What is the advantage of using e.g., range_begin() instead of just specializing boost::begin()?
i guess that would work too ... I think this is what is done for Shims in the STLSoft libs. I can't remember the details of the differences between the two approaches, but I'm sure there is a someone on the list that can. I guess the ADL version emplasizes putting stuff in the same interface as your class, instead of messing around into another namespace. -Thorsten