right-value reference
left-value, right-value
lvalue
: have an address (can use&
operator)rvalue
: no address, a temporary value (cannot use&
operator)
// a is left, 5 is right
int a = 5;
struct A {};
// a is left, A() is right
A a = A();
left-value reference
The most common reference we use.
Mostly used in lieu of pointer to make code cleaner.
int a = 5;
int& r1 = a; // OK! left-value can be referenced.
int& r2 = 5; // Error! right-value cannot be referenced.
const int& r2 = 5; // OK! right-value can be const referenced. This is necessary in cases such as vector::push_back(const T& val), we want both v.push_back(complex_obj) and v.push_back(1) work.
right-value reference
rvalue
can be further categorized:
prvalue
, pure-right-value: the initial value, or expression, or function returns.xvalue
, expiring value: a temporary value instantiated from aprvalue
.
int&& r1 = 5; // OK! standard right-value reference.
r1 = 6; // can be modified (not a const).
int a = 5;
int&& r2 = a; // Error! left-value cannot be right-referenced.
int&& r3 = std::move(a); // OK! std::move cast left value to right value.
int&& r4 = static_cast<int&&>(a); // ditto.
// note &a == &r2 == &r3 == &r4
In fact, right-value reference is just a wrapper of left-value reference with a temp variable.
int &&r = 5;
// eqauls
int tmp = 5;
int &&r = std::move(tmp);
right-value can also be used to pass reference:
void f(int& x) { x++; }
void g(int&& x) { x++; }
int main() {
int x = 1;
f(x); // x is now 2
g(std::move(x)); // x is now 3
// we prefer to use f(x) here, since this is not a useful application of rvalue reference.
}
右值引用既可以是左值也可以是右值,如果有名称则为左值,否则是右值。
int a = 5;
std::move(a); // no name, a right value.
int&& r = std::move(a); // with name, so r is a left value.
void f(int&& r) {}
void g1(int&& r) { f(r); } // wont compile.
void g2(int&& r) { f(std::move(r)); }
f(r); // Error! r is a left value.
f(std::move(r)); // OK!
f(std::move(a)); // OK! equals last line.
// Once passed in a function and got named, the right value is no longer a right value.
g1(std::move(r)); // Error! (no matching function to call f)
g2(std::move(r)); // OK!
More Examples:
// 形参是个右值引用
void change(int&& right_value) {
right_value = 8;
}
int main() {
int a = 5; // a是个左值
int &ref_a_left = a; // ref_a_left是个左值引用,ref_a_left本身也是左值
int &&ref_a_right = std::move(a); // ref_a_right是个右值引用,但ref_a_right本身也是左值
change(a); // 编译不过,a是左值,change参数要求右值
change(ref_a_left); // 编译不过,左值引用ref_a_left本身也是个左值
change(ref_a_right); // 编译不过,右值引用ref_a_right本身也是个左值
change(std::move(a)); // 编译通过
change(std::move(ref_a_right)); // 编译通过
change(std::move(ref_a_left)); // 编译通过
change(5); // 当然可以直接接右值,编译通过
cout << &a << ' ';
cout << &ref_a_left << ' ';
cout << &ref_a_right;
// 打印这三个左值的地址,都是一样的
}
Applications
右值引用主要在函数传参时用移动代替拷贝,即需要拷贝但被拷贝者之后又不再需要的场景,例如临时变量。因为std::move
只改变地址指向,而不会物理上的移动(即拷贝)数据。
需要注意的是,类的默认移动构造函数(A(A&& a){}
)就是默认拷贝构造函数(A(A& a){}
)。为了使得自己定义的类能够利用移动语义,我们需要自己定义移动构造函数:
// ref: https://stackoverflow.com/questions/11572669/move-with-vectorpush-back
// ref: https://stackoverflow.com/questions/3106110/what-is-move-semantics
class mystring {
char* data;
// the big THREE to manage raw pointers: destructor, copy constructor, copy assignment operator.
// from c++11, we have the big FIVE, plus: move constructor, move assignment operator.
// with the copy-and-swap idiom, we only need one assignment operator, so it's also called big 4.5
public:
// default constructor
mystring(const char* p) {
size_t s = std::strlen(p) + 1;
data = new char[s];
std::memcpy(data, p, s);
}
// destructor
~mystring() {
delete[] data;
}
// copy constructor (deep copy)
mystring(const mystring& that) {
size_t s = std::strlen(that.data) + 1;
data = new char[s];
std::memcpy(data, that.data, s);
}
// move constructor (no copy)
mystring(mystring&& that) {
data = that.data;
that.data = nullptr;
}
// (copy and move) assignment operator
// ref: https://stackoverflow.com/questions/3279543/what-is-the-copy-and-swap-idiom
// we can use copy-and-swap idiom to avoid duplicated code and self-assignment test.
// this handles both copy and move, based on whether that is lvalue(copy) or rvalue(move).
// e.g., assume x, y are both mystring.
// mystring s = x; // copy, since x is lvalue. the copy constructor is called.
// mystring s = x + y; // move, since x + y is rvalue (assume we overload operator+). first create a temp object, then the move constructor is called.
mystring& operator=(mystring that) {
std::swap(data, that.data);
return *this;
}
}
STL容器大多实现了各种方法的移动语义,例如
// just an illustration
vector(vector&& tmp_vector) {
data = tmp_vector.data;
tmp_vector.data = nullptr;
}
void push_back(const T& x) {/* copy semantic */}
void push_back(T&& x) {/* move semantic */}
Use case:
vector<mystring> v;
mystring s("a_long_string");
v.push_back(s); // copied
v.push_back(move(s)); // moved
v.push_back(mystring("blahblah")); // moved
struct A {
/* naive flat class, no raw pointer, no move constructor */
int x;
A(int _x) : x(_x) {}
};
vector<A> v;
A a(1);
v.push_back(a); // traditional. a is copied. slow!
v.push_back(move(a)); // a is still copied, no difference from the last line.
v.push_back(A(1)); // A(1) is still copied.
其它语言中的类似情况:
c++默认的对象传值方式是拷贝,例如
vec.push_back(obj); auto vec2 = vec1;
等都会触发拷贝行为。引用传值必须通过&
显式显式实现,例如vec.push_back(move(obj)); auto vec2 = &vec1;
python默认的对象传值方式就是引用,例如
l.append(obj), obj2 = obj1
均为引用。相反,如果需要拷贝,则要通过deepcopy
显式实现。
通用引用 Universal Reference
表现形式为T && (模板+右值引用),利用编译器对模板的推导多义性,可以同时绑定左值与右值。
引用折叠:编译器特性,实际语法不存在引用的引用int& &
,但模板类型推导支持对引用的引用进行折叠,具体规则为:
T& &, T& &&, T&& & --> T&
T&& && --> T&&
利用此特性,通用引用得以实现:
template <typename T> void f(T&& x) {}
int x = 1;
f(x); // x is lvalue, T = int won't match, so try T = int&, and use f(int& x) which matches!
f(1); // 1 is rvalue, T = int matches, so use f(int&& x).
事实上,move
的实现就利用了通用引用:
template <typename T>
typename remove_reference<T>::type&& move(T&& t) {
return static_cast<typename remove_reference<T>::type&&>(t);
}
完美转发问题
完美转发指的是在函数内部将参数包括类型原封不动的传递给内部的另一函数。但由于右值引用传入函数后会变成左值,简单的参数转发是错误的!
// inner function
void f(int& x) { x++; }
void g(int&& x) { x++; }
// sender
template<typename F, typename T>
void sender(F f, T t) { f(t); }
int x = 1;
f(x); // x is now 2.
sender(f, x); // not work, x is still 2.
sender(g, x); // wont compile, T = int, t is a lvalue
sender(g, move(x)); // wont compile, T = int&&, but t is a lvalue
通用引用只能解决f(int& x)
的问题:
// inner function
void f(int& x) { x++; }
void g(int&& x) { x++; }
// sender
template<typename F, typename T>
void sender(F f, T&& t) { f(t); }
int x = 1;
f(x); // x is now 2.
sender(f, x); // x is now 3. T = int&
sender(g, x); // wont compile, T = int&, but t is still a lvalue
sender(g, move(x)); // wont compile, T = int, but t is still a lvalue
从而,我们需要动态判断应该转发左值还是右值。仅仅通过move是不够的,所以我们要使用std::forward<T>(u)
实现左右值转化:
// impl for lvalue
template <typename T>
T&& forward(typename remove_reference<T>::type& t) noexept {
return static_cast<T&&>(t);
}
// impl for rvalue
template <typename T>
T&& forward(typename remove_reference<T>::type&& t) noexept {
return static_cast<T&&>(t);
}
// examples
std::forward<int>(x); // cast to right-value, equals `std::move(x)`
std::forward<int&&>(x); // also cast to right-value, equals `std::move(x)`
std::forward<int&>(x); // cast to left-value
More Examples:
#include <iostream>
void change2(int&& ref_r) {}
void change3(int& ref_l) {}
void change4(const int& ref) {}
void change(int&& ref_r) {
change2(5);
// change2(ref_r);
change2(std::move(ref_r));
change2(std::forward<int>(ref_r));
change2(std::forward<int &&>(ref_r));
// change3(5);
change3(ref_r);
change3(std::forward<int &>(ref_r));
change4(5);
change4(ref_r);
change4(std::forward<int &>(ref_r));
}
int main() {
int a = 5;
change(std::move(a));
}
完美转发的实现(结合forward与通用引用):
// inner function
void f(int& x) { x++; }
void g(int&& x) { x++; }
// sender
template<typename F, typename T>
void sender(F f, T&& t) { f(forward<T>(t)); }
int x = 1;
f(x); // x is now 2.
sender(f, x); // x is now 3. T = int&, forward cast t to lvalue.
sender(g, move(x)); // x is now 4, T = int, forward cast t to rvalue.
sender(g, x); // wont compile, T = int&, forward cast t to lvalue.