[boost] Another, potentially nastier problem with dereferenceable and the iterator_helpers

From: jsiek_at_[hidden]
Date: 2000-06-13 12:32:46

You are right, it should be the pointer type instead of V*.  As is
definitely causes problems... fails reverse iterator test for
ListReverseConstIterator.

David Abrahams writes:
 > template <class T, class V, class D = boost::detail::empty_base>
 > struct dereferenceable : D
 > {
 >   V* operator->() const
 >   {
 >     return &*static_cast<const T&>(*this);
 >   }
 > };
 > 
 > Notice that the return type is V*: plain, unadorned V*.
 > The standard quoth:
 > 
 > [iterator traits is specialized] for pointers to const as:
 > 
 > template<class T> struct iterator_traits<const T*> {
 >   typedef ptrdiff_t difference_type;
 >   typedef T value_type;
 >   typedef const T* pointer;
 >   typedef const T& reference;
 >   typedef random_access_iterator_tag iterator_category;
 > };
 > 
 > Notice that the value_type is plain, unadorned T. I guess that means that
 > the standard library expects value_type, even for const iterators, to be
 > non-const. Exhaustively searching the standard for reliance on this "fact"
 > turns up nothing of any consequence (look at uninitialized_copy and decide
 > for yourself). This means that when we create the following class:
 > 
 > struct X : boost::random_access_iterator_helper<X, Y, std::ptrdiff_t, const
 > Y* const Y&>
 > {
 >   ...
 > };
 > 
 > We get an iterator through which we can access Y as non-const, but whose
 > iterator_traits look just like the iterator_traits for const Y*. Problem?
 > I guess there shouldn't be one, (since we can ask the user to pass const Y
 > as the value_type) but I am not confident...
 > 
 > I suppose we could change the interface to dereferenceable to take the
 > pointer type instead of the value_type.
 > 
 > -Dave
 > 
 > 
 > ------------------------------------------------------------------------
 > IT Professionals: Match your unique skills with the best IT projects at
 > http://click.egroups.com/1/3381/3/_/9351/_/960869339/
 > ------------------------------------------------------------------------
 > 
 > 
 >