The widespread use of the object-oriented (OO) paradigm to 
develop software systems has resulted in the class becoming 
the basic test unit.  The class structure is more complex 
than the traditional test unit, the "subprogram", due to 
the possible combinations of the OO properties exhibited by 
the features (data items and methods) of the class.  These
properties include: typing, accessibility, polymorphism, 
dynamic binding, exception handling, and concurrency.
In addition, the relationships between classes in a system 
component have become more complex.  As a result of this 
complexity in both the class structure and inter-class 
relationships, testers are spending more time analyzing 
the source code before the actual testing process begins.  
This analysis is required so that the tester can decide on 
the appropriate class integration test order and which testing 
techniques are most appropriate to adequately testing each 
class in the application.
 
In this dissertation I present a new class abstraction technique, 
"A Taxonomy of Object-Oriented (OO) Classes", and show how 
this class abstraction technique supports the testing of classes.   
My taxonomy of OO classes catalogs an OO class based on the 
characteristics exhibited by that class.  The characteristics 
of a class include the properties of its features (data items 
and methods) and its relationships with other classes.  Using 
the information captured by my taxonomy I extend an existing 
class integration test strategy to: (1) order concrete derived 
classes of an abstract class, hence reducing the testing effort, 
and (2) order instances of parameterized classes based on the 
actual types of their parameters.  Finally, I develop an algorithm 
to map implementation-based testing techniques to a "class under
test".  This mapping process identifies those implementation-based
testing techniques that can be used to "suitably test" a class 
based on the "characteristics" of the class under test.