[RFC] libstdc++/8780

[Martin Sebor]

Joe Buck wrote:
> I wrote:
> 
>>>The short answer is that, given
>>>    std::vector<T>::iterator iter_T;
>>>
>>>then
>>>    &*iter_T
>>>
>>>has type T* and points to the same object.  As the elements of a vector
>>>must be contiguous, this works portably.
>>
> 
> Martin writes:
> 
>>Unfortunately, it will not work for an iterator pointing at the end
>>of a container.
> 
> 
> You are of course correct that there are no guarantees that 
> &*some_vector.end() will not produce a one-past-the-end pointer;
> debugging implementations of vector<T>::iterator will very likely
> throw an exception or abort.
> 
> Still, note that I said "has type T* and points to the same object":
> in most cases, users that complain about this are trying to interface
> data in a vector to a function that wants a pointer, and for them
> the &*iter_T mechanism will do.
> 
> The most common case where I've needed to use the &* construction is to
> pass data in a vector<T> to some function that wants a const T* and
> perhaps a count.

Over a year ago, there was a discussion on c++std-lib@research.att.com
(see c++std-lib-7863) about adding vector::data() member function that
would be analogous to basic_string::data(). A function like that would
take care of the problem but, strangely, I don't see an issue about it
on the issues list or a formal proposal. Any takers? :)

Martin