explicit in details

If you ask a C ++ programmer about the meaning of the keyword explicit, the majority will answer that this keyword is placed before the constructor declaration with one parameter (or with a large number of parameters, but when all parameters, starting from the second, have default values) and prevent implicit conversion types during initialization.

class Simple { public: Simple(int a) : a_(a) {} private: int a_; }; class SimpleExplicit { public: explicit SimpleExplicit(int a) : a_(a) {} private: int a_; }; template <typename S> void someFunc(const S& s) { } int main(int, char**) { Simple s3 = 11; // SimpleExplicit se3 = 11; - COMPILE ERROR SimpleExplicit se3 = SimpleExplicit(11); someFunc<Simple>(11); // someFunc<SimpleExplicit>(11); - COMPILE ERROR someFunc<SimpleExplicit>(SimpleExplicit(11)); return 0; } 

In the good old C ++ 03, the application scenarios for this keyword ended there, however, since C ++ 11, the scope of explicit has expanded: now it makes sense not only in constructors with one parameter, and even not only in constructors.

In 2011, a standard initialization (uniform initialization) was added to the Standard, which should bring order to the zoo in ways to initialize objects, inherited C ++ from language C. I will not discuss in detail here about universal initialization, there are many detailed articles on this topic. easy to find by keywords. In a nutshell: objects are proposed to be initialized using curly brackets, in fact, this extension is so-called. aggregate initialization (aggregate initialization), inherited from the time of C.

With the advent of universal initialization, explicitly found meaning for designers with 0.2.3 or more parameters:

 class Simple { public: Simple() : a_(0), b_(0) {} Simple(int a) : a_(a), b_(0) {} Simple(int a, int b) : a_(a), b_(b) {} private: int a_, b_; }; class SimpleExplicit { public: explicit SimpleExplicit() : a_(0), b_(0) {} explicit SimpleExplicit(int a) : a_(a), b_(0) {} explicit SimpleExplicit(int a, int b) : a_(a), b_(b) {} private: int a_, b_; }; template <typename S> void someFunc(const S& s) { } int main(int, char**) { Simple s4 = {}; someFunc<Simple>({}); // SimpleExplicit se4 = {}; - COMPILE ERROR SimpleExplicit se4 = SimpleExplicit{}; // someFunc<SimpleExplicit>({}); - COMPILE ERROR someFunc<SimpleExplicit>(SimpleExplicit{}); Simple s5 = {11}; someFunc<Simple>({11}); // SimpleExplicit se5 = {11}; - COMPILE ERROR SimpleExplicit se5 = SimpleExplicit{11}; // someFunc<SimpleExplicit>({11}); - COMPILE ERROR someFunc<SimpleExplicit>(SimpleExplicit{11}); Simple s6 = {11, 22}; someFunc<Simple>({11, 22}); // SimpleExplicit se6 = {11, 22}; - COMPILE ERROR SimpleExplicit se6 = SimpleExplicit{11, 22}; // someFunc<SimpleExplicit>({11, 22}); - COMPILE ERROR someFunc<SimpleExplicit>(SimpleExplicit{11, 22}); return 0; } 

In addition, starting with C ++ 11, the explicit keyword can also be applied to type conversion operators, also prohibiting their implicit call:

 class Simple { public: Simple() {} operator bool() const { return true; } }; class SimpleExplicit { public: explicit SimpleExplicit() {} explicit operator bool() const { return true; } }; int main(int, char**) { Simple s7{}; bool b7 = s7; SimpleExplicit se7{}; // bool be7 = se7; - COMPILE ERROR bool be7 = static_cast<bool>(se7); return 0; } 

In conclusion, I would like to recommend using universal initialization in any new C ++ code, and also explicitly declare explicit constructors at all times, unless the implicit conversion is semantically justified.