C++ Move Semantics – Why Call Move on an Rvalue Reference?

c++move-semantics

In the code below, why doesn't the first call mkme = mvme_rv dispatch to T& operator=(const T&&)?

#include <iostream>
#include <string>
#include <vector>

using namespace std;
using T = vector<int>;

int main()
{
  T mvme(10, 1), mkme;
  T&& mvme_rv = move(mvme); // rvalue ref?
  mkme = mvme_rv;           // calls T& operator=(const T&)?
  cout << mvme.empty();     // 0
  mkme = move(mvme_rv);     // calls T& operator=(const T&&)?
  cout << mvme.empty();     // 1
}

Best Answer

As skypjack correctly comments, accessing an object through its name always results in an lvalue reference.

This is a safety feature and if you think it through you will realise that you are glad of it.

As you know, std::move simply casts an l-value reference to an r-value reference. If we use the returned r-value reference immediately (i.e. un-named) then it remains an r-value reference.

This means that the use of the r-value can only be at the point in the code where move(x) is mentioned. From a code-reader's perspective, it's now easy to see where x's state became undefined.

so:

 1: auto x = make_x();
 2: auto&& r = std::move(x);
 3: // lots of other stuff
35: // ...
54: // ...
55: take_my_x(r);

does not work. If it did, someone maintaining the code would have a hard time seeing (and remembering) that x (defined on line 1) enters an undefined state on line 55 through a reference taken on line 2.

This is a good deal more explicit:

 1: auto x = make_x();
 2: //
 3: // lots of other stuff
35: // ...
54: // ...
55: take_my_x(std::move(x));
Related Question