Question

I’m not sure if I’m suffering more from a documentation error or a headache, so…

What I want to do is create a shared_ptr that shares ownership with another, but which references a member of the object instead of the whole object. Simple example, starting point…

struct s
{
  int a, b;
};

shared_ptr<s> s1 (new s);  //  pointing to whole object

From en.cppreference.com, constructor (8) of shared_ptr is…

template< class Y >
shared_ptr( const shared_ptr<Y>& r, T *ptr );

The description mentions “Constructs a shared_ptr which shares ownership information with r, but holds an unrelated and unmanaged pointer ptr … such as in the typical use cases where ptr is a member of the object managed by r”.

So… Was T just accidentally missed from the template in that constructor, or am I missing something? In fact, Y looks like it’s wrong to me too, so just generally is that constructor described correctly?

What I’m hoping I can do is something like this…

shared_ptr<int> s2 (s1, &(s1.get ()->a));

s2 points to member a (an int), but shares ownership of the whole object with s1.

Is that sane?

Answer 1

The T parameter is a template parameter on the shared_ptr itself, whereas the Y parameter is a template parameter on that particular shared_ptr constructor. Something like this:

template< class T >
class shared_ptr
{
     template< class Y >
     shared_ptr( const shared_ptr<Y>& r, T *ptr );
}

As for the example code you’ve posted, that looks fine to me.