Application of object-oriented programming

Answer 1

Answer

A virtual function must be declared as a non-static member method of a class that you expect to act as a base class. Declaring a virtual function adds some overhead as a result of creating the v-table (virtual method table), but most of that overhead is paid with the first virtual function (subsequent virtual functions just add a new entry to the already-existing v-table). However, do not declare a virtual function unless you expect that function to be overridden. Bear in mind that overriding an overloaded, non-virtual function "hides" all the overloads in the base class.

If a virtual function must be overridden, declare it as pure-virtual instead. You do not need to implement the method in the base class, but you will be reminded to provide an implementation in the derived class at compile time if one does not exist, even if you provide a default implementation in the base class. Bear in mind that base classes with one or more pure-virtual methods become abstract -- they cannot be instantiated.

If there is any virtual function or pure-virtual function, there must also be a virtual destructor, as well as a public or protected default constructor (a constructor with no arguments). When a derived class is constructed, it calls the base class constructor, which calls its base class constructor. Derived classes are constructed in sequence, beginning with the least-derived class. Destruction is the reverse -- the most-derived class is destroyed before the base classes are destroyed.

Virtual functions can be invoked just like any other class member method, both via an object reference's member operator (.), and the indirection operator (-->) for pointers to objects. It does not matter whether the reference or pointer refers to a base class or a derived class; the v-table decides which override (where one is provided) will actually execute, starting from the most-derived override and working back towards the base class, the least-derived. With appropriate use of virtual functions, dynamic casting can be avoided completely (dynamic casting should never be employed as it completely defeats the point of having a v-table in the first place).

Answer 2

The ability to create large programs while maximising code reuse and allowing design and reasoning about a problem to be fully contained on different levels of abstraction.