Containment or Private Inheritance?

Given that you can model a has-a relationship either with containment or with private inheritance, which should you use? Most C++ programmers prefer containment. First, it’s easier to follow. When you look at the class declaration, you see explicitly named objects representing the contained classes, and your code can refer to these objects by name. Using inheritance makes the relationship appear more abstract. Second, inheritance can raise problems, particularly if a class inherits from more than one base class. You may have to deal with issues such as separate base classes having methods with the same name or of separate base classes sharing a common ancestor. All in all, you’re less likely to run into trouble using containment. Also containment allows you to include more than one subobject of the same class. If a class needs three string objects, you can declare three separate string members by using the containment approach. But inheritance limits you to a single object. (It is difficult to tell objects apart when they are all nameless.)

However, private inheritance does offer features beyond those provided by containment. Suppose, for example, that a class has protected members, which could either be data members or member functions. Such members are available to derived classes but not to the world at large. If you include such a class in another class by using composition, the new class is part of the world at large, not a derived class. Hence it can’t access protected members. But using inheritance makes the new class a derived class, so it can access protected members.

Another situation that calls for using private inheritance is if you want to redefine virtual functions. Again, this is a privilege accorded to a derived class but not to a containing class. With private inheritance, the redefined functions would be usable just within the class, not publicly.