Question

All the STL containers that implement resize use copies to populate the new elements even if the source of the copy is a default constructed object?

Why is it done this way?

I see no advantage and some cost.

---

As context, I ran across this while looking for a random access container for elements that can’t be copied.

Answer 1

It saves on complexity. We certainly need the copy-construction case, and default-construction can be modeled as replicating a default-constructed object.

The performance penalty is negligible. Writing zeroes is about the same speed as copying zeroes. The compatibility penalty is nil since all containers require copyability anyway. Default-construction on the other hand isn’t required.

If you really want to use standard containers with non-copyable objects, look into C++0x and in-place construction with emplace. However, there is no method to emplace multiple elements at once. (If you’re using deque, there shouldn’t be much performance penalty to an emplace loop vs. resize though.)