9.13. static Variables And Functions

Generally, each object instantiated from a class maintains its own distinct copy of each member variable declared in the class. So, in a sense, the object "owns" the member variables, and programs must access them through their owners. However, sometimes programs need to manage data that is accessible by various combinations of many objects (the Pouring Puzzle example presented later in the chapter illustrates this situation). We could implement the data as global variables, but we've seen that globals are error-prone, so we seek a more secure, object-oriented solution. The object-oriented solution we seek is called class variables. Whereas member variables belong to specific objects, class variables belong to all instances of a given class. Like Java, C++ implements class variables with the static keyword (please see footnote 1).

static And UML Diagrams

Non-static or class members (attributes and operations) belong to individual instances of the class (i.e., objects). On the other hand, static or class features "belong" to the class. The contrast between the two ownerships constitutes a significant difference between static and non-static features. The UML provides additional notation to differentiate between the kinds of ownership. Unless otherwise specified, features in a UML class diagram are non-static and have instance (i.e., object) ownership. Alternatively, the UML denotes static features with class ownership by underlining the feature in the class diagram. Programmers translate the underlining to the static keyword when they translate the UML class diagram to C++.

static Variables And Functions

To help us understand the role static features play in C++ programs, imagine that we want to write a program that counts the total number of objects it instantiates from class. Our approach is simple: define a variable named count to count each object as the program creates it. Consider the following problems that arise if we define count as a non-static or member variable:

• The program will duplicate the data throughout all the instantiated objects, which wastes space and makes maintaining the correct count in all the objects difficult at best.
• The value saved in count is lost when the program destroys all the counted objects.
• count is an accumulator. So, where should we initialize it? If we initialize it in the constructor, the program will reset the value in count whenever the program creates a new object.

Although programs can access static variables through objects, these variables "belong" to a class. Consequently, they exist even when the program hasn't instantiated any objects from the class or destroyed them. This observation implies that we need a class-level notation to manage static variables independently of any object. Furthermore, that notation must be sufficiently robust to support any functions that access the variables, including getter and setter functions. C++ bases the notation on the scope resolution operator, ::. When programmers use the scope resolution operator to define or use a variable or function, the code looks very much like it did when they used the operator with namespaces.

Foo.h	Foo.cpp
class Foo { private: int var1; int var2; static int count; public: Foo(); static int get_count(); };	#include "Foo.h" int Foo::count = 0; // define and initialize Foo::Foo() { count++; } int Foo::get_count() { return count; }
(a)	(b)
Client	Memory Layout
#include <iostream> #include "Foo.h" using namespace std; int main() { Foo f0; Foo f1; Foo f2; cout << f0.get_count() << endl; // works not preferred cout << Foo::get_count() << endl; // preferred return 0; }
(c)	(d)

Class Constants

A symbolic constant is an unchangeable value named for clarity or convenience. C++ inherits two ways of creating symbolic constants from the C programming language: create macros with #define or use enumerations. The first technique makes it difficult to restrict the constant to class scope, and the second technique is limited to integer constants. But we can also create constants with const keyword, and these can be any data type, follow all scoping rules, and can represent any valid data type. We can easily make these into class constants whose scope is limited to member functions by placing them inside a class.

class foo
{
    public:
        const static double MAGIC = 2.7;
	void function();
};

class bar
{
    public:
        const static double MAGIC = 3.14;
	void function();
};

void foo::function()
{
    ... MAGIC ...
}

void bar::function()
{
    ... MAGIC ...
}

void application1()
{
    ... foo::MAGIC ...
}

void application2()
{
    ... bar::MAGIC ...
}