Once I loved C very much and I thought it’s the best programming language in the world. I used to argue that C can be used to implement the object-oriented design with struct. I used to argue that private and public is useless in C++. However, now I know I was wrong.

Why we need private

It’s related to a very important concept in C++, that is, encapsulation.

Encapsulation is not invisibility. It’s maintainability.

Suppose a class is very popular and is used by many projects. For example, the string class. If every members in this class is public, the clients can use many interfaces to manipulate string. One day, when the string class is going to be modified(some members’ name change), then lots of clients have to modify there code. This is painful! If we declare some of the members private and change the name of the private members, only the implementations in the public member functions should be modified. None of the clients need to change their codes.

Therefore, public means unchangable and public means unencapsulated.

The public member functions should not be changed in the future. Since you don’t know how your clients will use your class, don’t expose the members that may be changed in the future to the clients.

Prefer non-member non-friend functions to member functions

Consider the following class.

class Person {
public:
    explicit Person()
    { }
    ~Person()
    { }
    void
    say_age() const
    { printf("%d\n", age); }
    void
    say_name() const
    { printf("%s\n", name.c_str()); }
private:
    Person(const Person&);
    const Person&
    operator=(const Person&);
    int age;
    std::string name;
};

Somebody may want to use the functions together, so we add another member function.

class Person {
public:
    explicit Person()
    { }
    ~Person()
    { }
    void
    say_age() const
    { printf("%d\n", age); }
    void
    say_name() const
    { printf("%s\n", name.c_str()); }
    void
    say() const
    {
        say_age();
        say_name();
    }
private:
    Person(const Person&);
    const Person&
    operator=(const Person&);
    int age;
    std::string name;
};

This is straightforward, especially for those Java and C# programmers. Every functions should be in a class and everything is object. It’s obvious, right?

However, e have another option, that is, using the non-member function.

class Person {
public:
    explicit Person()
    { }
    ~Person()
    { }
    void
    say_age() const
    { printf("%d\n", age); }
    void
    say_name() const
    { printf("%s\n", name.c_str()); }
private:
    Person(const Person&);
    const Person&
    operator=(const Person&);
    int age;
    std::string name;
};

void
person_say(const Person& p)
{
    p.say_age();
    p.say_name();
}

Which is better? It seems that they are the same.

We should use the non-member function instead of the member function.

As we all know, **the less the public member functions and the friend ** functions, the greater encapsulation is, since member functions and friend functions are the only interface that can access the private members.

In this example, if we add another public member function, we decrease the encapsulation of the class since this member function can access the private members of the class. But if we use the non-member non-friend function, it won’t have impact with the class because it’s the function outside.

Does it explain that C++ is not so object-oriented? From Java or C#, those programmers will say that everything is object and classes are everywhere. However, they are wrong. Object-oriented princiles state that data should be encapsulated as possible. Therefore, prefering non-member non-friend functions is more object-oriented.

Until now, I can say that I really understand the meaning of encapsulation and why it’s so important in object-oriented programming.