静态转换到右值引用和 std::move 之间有什么区别吗

Question

The description for static cast says

If new_type is an rvalue reference type, static_cast converts the value of expression to xvalue. This type of static_cast is used to implement move semantics in std::move.(since C++11)

Does this confirm that the following are equivalent ?

(A)

X x1;
X x2 = static_cast<X&&>(x1); 

(B)

X x1;
X x2 = std::move(x1);

Answer 1

Yes there is a very important difference: std::move documents what you want to do. In addition the cast is prone to writing errors like a forgotten & or wrong type X .

As it can be seen, std::move is even less to type.

Answer 2

In C++11, T&& is an rvalue reference. They behave like lvalue references from C++ 98/03. Their goal - to be a candidate for moving. In C++98 this construct can appear in reference collapsing.

std::move - turn expression into an rvalue. It could have been called rvalue_cast , but wasn't.

Explicit cast to type T&& is possible in principle. The official standard costs some money, but in the ISO/IEC 14882:2011 draft there's this:

5.2.9 Static cast

8)

The lvalue-to-rvalue (4.1), array-to-pointer (4.2), and function-to-pointer (4.3) conversions are applied to the operand....

From a practical point of view, it is more convenient to use std::move .

Consider this example:

#include <stdio.h>
#include <utility>

class A
{
public:
A () {printf ("A ()" "\n");}
A (const A &) {printf ("A (&)" "\n");}
A (A &&) {printf ("A (&&)" "\n");}
A (const A &&) {printf ("A (const &&)" "\n");}
~ A () {printf ("~ A ()" "\n");}
};


int main ()
{
const A obj;
A obj2 (std::move (obj)); // 1-st approach
A obj3 (static_cast <const A&&> (obj));  // 2-nd approach
}

For me, the first approach is:

• more convenient (should you perform static_cast to const A&& , or to A&& ?)
• more explicitly (I can use search in text editor to find std::move in the project)
• less error-prone.

Answer 3

They are not strictly equivalent. std::move() 's implementation relies on static_cast :

template<typename _Tp>
constexpr typename std::remove_reference<_Tp>::type&&
move(_Tp&& __t) noexcept
{ return static_cast<typename std::remove_reference<_Tp>::type&&>(__t); }

They differ in the sense that std::move() has remove_reference to deal with reference collapse . An example where (A) and (B) are not strictly equivalent:

// Set up different behavior for lvalue and rvalue.
class T {};
void func(T&& t) { std::cout << "rvalue captured.\n"; }
void func(T& t) { std::cout << "lvalue captured.\n"; }
// Example:
Using X = T&;
X x1;
X x2 = static_cast<X&&>(x1); // (A) "lvalue captured."
X x3 = std::move(x1);        // (B) "rvalue captured."

Answer 4

You can use static_cast<A &&>(a) when a is an rvalue, but you shouldn't use std::move(a) .
When you use A && a = std::move(A()) , you get a dangling reference.

The basic idea is that the lifetime of a temporary cannot be further extended by "passing it on": a second reference, initialized from the reference to which the temporary was bound, does not affect its lifetime.

std::move 's implementation is somewhat like

template <typename T>
constexpr decltype(auto) move(T && __t) noexcept  // when used in std::move(A()),
                                                  // the lifetime of the temporary object is extended by __t
{
    return static_cast<typename std::remove_reference<T>::type &&>(__t);  // a xvalue returned, no lifetime extension
}

auto && a = std::move(A());  // the anonymous object wiil be destructed right after this line