next up previous contents
Next: 2.2 Design of the Up: 2. The Algorithm Library Previous: 2. The Algorithm Library

2.1 Object-Oriented Programming

Object-oriented programming has been a very heavily (and often misused) catchword in recent days coming up with quite a number of different and partially contradicting interpretations of the involved concepts. This fuzziness is the result of the complex roots of the object-oriented programming concepts in conjunction with the huge variety of competing programming languages putting it into practice [2,10].

The first programming language utilizing objects is Simula [11] which was aimed at supporting simulation activities. Procedures could be attached to a type (a class in Simula's terminology) to represent the behavior of an instance. Amongst many others the most popular object-oriented programming languages Smalltalk, Eiffel, Objective-C, and C++ evolved, each introducing own specific nomenclatures and interpretations of the key features of object-orientation.

Therefore the basic ideas and most important notions shall be summarized in this chapter clarifying their exact interpretation in the remaining parts of this text. Note that these interpretations are closely related and compatible to the corresponding definitions used by B. Stroustrup for C++ [58,59,1]. The central concept in this context is the object. Summarizing different implicit and explicit definitions in [10,7,48] a object represents an identifiable entity able to span a state space and a number of operations (behavior) to either examine or affect its state or the state of peer objects. Based on this definition object-oriented programming comprehends the following programming paradigms and notions:

data encapsulation: Semantically interdependent data items (members) are grouped into an aggregate (class) to form a single data item. The grouping instructions are summarized in the class definition. Implementations of these class definitions which allocate certain system resources per item are commonly addressed as (class) instances or objects.
function encapsulation: Algorithms are coded as a sequence of statements grouped together forming functions. While the concept of data and function encapsulation is available in most of the procedural third-generation languages, object-oriented programming languages provide the concept of
data and function association: Definite relations between data items and functions processing them are generated. In this context the functions associated with the class definition are called (class) methods. Consequently special functions can be used to define operators working on data items of specific user defined classes. Thus user defined data types can be defined which behave almost like the built-in data types of the language itself. This concept is often referred to as data abstraction.
inheritance: Structural relationships between classes in the form ``class A'' is a kind of (implementation of) ``class B'' can be expressed by means of data and function inheritance. In these terms inheritance means, that the derived class A inherits data members and methods from the parent class B with a single declaration of their relationship. Inherited members and methods are indistinguishable from intrinsic members and methods defined on the class.
overloading: To implement different flavors of certain concepts, object-oriented languages provide the method of overloading method and data members inherited from a base class in the derived class. Thereby it is possible to define an interface class providing method definitions which are common to a certain group of classes. Classes which are derived from this interface class can either implement own versions of these methods or rely on the implementation in the parent class. All classes which are defined by inheriting from the interface class or from other classes which are derived from the interface class in this manner are called implementations of the interface class. The derived classes also are denominated as classes of the type of the interface class.
runtime type identification: Each instance of the different class types can be identified at runtime. This identification works also for references to the object whether the reference points to a class with the actual object type or to one of its parent classes. This feature allows for containers of references to base classes without loosing the information about the actual type of the stored objects, which is a big advantage over statically typed languages like C or PASCAL.
polymorphism: The availability of runtime type information for object references enables the concept of consistently using references to a common interface class of a group of classes instead of references to the actual class. Such a reference represents all different implementations of the interface and is therefore entitled as polymorphic reference. Two types of methods can be defined for such classes. When normal methods of the interface class are invoked the method definition of the interface class is evaluated. When virtual methods of the interface class are invoked the method definition of the actual class definition represented by the reference to its interface class are computed.

next up previous contents
Next: 2.2 Design of the Up: 2. The Algorithm Library Previous: 2. The Algorithm Library
Robert Mlekus