C ++将operator =()赋值转换为构造的规则究竟是什么?例如Foo foo = bar
实际上会调用Foo的构造函数接受bar作为参数,如果它存在的话。我用谷歌搜索了它是如何工作但似乎找不到任何东西。
我遇到一个问题,弄清楚为什么下面的任务试图采用一个构造函数,但没有采用明显正确的一个:HandlePtr(TYPE& resource)。使用实际构造语法的构造工作正常,但不能使用赋值运算符。
代码(显然是为了简洁而编辑):
template< typename TYPE >
class HandlePtr {
public:
HandlePtr( void ) = default;
HandlePtr( HandlePtr< TYPE >& other ) = default;
HandlePtr( TYPE& resource ) {} // generally I would make this explicit, but for testing purposes I took it out
~HandlePtr( void ) = default;
public:
HandlePtr<TYPE>& operator=( TYPE& resource ) { return *this; }
HandlePtr<TYPE>& operator=( HandlePtr<TYPE>& other ) { return *this; }
};
int main ( void ) {
int x = 5;
HandlePtr< int > g( x ); // works
HandlePtr< int > i;i = x; // works
HandlePtr< int > h = x; // doesn't work
// also tried this just out of curiosity:
HandlePtr< int > h = HandlePtr< int >( x ); // also does not work
return 0;
}
错误:
shit.cpp: In function ‘int main()’:
try.cpp:19:24: error: no matching function for call to ‘HandlePtr<int>::HandlePtr(HandlePtr<int>)’
HandlePtr< int > h = x; // doesn't work
^
try.cpp:19:24: note: candidates are:
try.cpp:7:3: note: HandlePtr<TYPE>::HandlePtr(TYPE&) [with TYPE = int]
HandlePtr( TYPE& resource ) {} // generally I would make this explicit, but for testing purposes I took it out
^
try.cpp:7:3: note: no known conversion for argument 1 from ‘HandlePtr<int>’ to ‘int&’
try.cpp:6:3: note: HandlePtr<TYPE>::HandlePtr(HandlePtr<TYPE>&) [with TYPE = int]
HandlePtr( HandlePtr< TYPE >& other ) = default;
^
try.cpp:6:3: note: no known conversion for argument 1 from ‘HandlePtr<int>’ to ‘HandlePtr<int>&’
try.cpp:5:3: note: HandlePtr<TYPE>::HandlePtr() [with TYPE = int]
HandlePtr( void ) = default;
^
try.cpp:5:3: note: candidate expects 0 arguments, 1 provided
try.cpp:20:20: error: redeclaration of ‘HandlePtr<int> h’
HandlePtr< int > h = HandlePtr< int >( x ); // also does not work
^
try.cpp:19:20: error: ‘HandlePtr<int> h’ previously declared here
HandlePtr< int > h = x; // doesn't work
答案 0 :(得分:2)
您在声明:
中忽略了这一点T t = u;
这不是赋值运算符。 t = u;
不是声明的子表达式。这里唯一的表达是u
;评估表达式u
的结果用作声明的对象t
的初始值。
如果u
的类型为T
,则t
是u
复制构建的。{/ p>
如果u
没有T
类型,则首先需要将u
转换为T
类型。这会创建T
类型的 rvalue 。
您没有任何接受右值的构造函数,因此T t = u;
和相同的T t = T(u);
都会失败。但是,T t(u)
成功,因为没有创建右值;值u
用作构造函数T(U &)
的参数。
简化代码示例:
struct T
{
T(int &);
T(T&);
T();
T &operator=(int &);
};
int main()
{
int x = 5;
T g(x); // OK, T(int &)
T g2(5); // fail, looks for T(int const &)
T i; // OK, T()
i = x; // OK, T::operator=(int&)
T h3 = i; // OK, T(T&)
T h1 = T(x); // fail, looks for T(T const &)
T h2 = x; // fail, identical to previous line
}
通常你应该使用const &
作为copy-constructors和赋值运算符的参数;然后所有这些“失败”的情况变为“OK”,因为rvalue可以绑定到const引用。