State-based testing - a new method for testing object-oriented programs

Abstract

State-based testing is a new method for testing object-oriented programs. The information stored in the state of an object is of two kinds: control-information and data-storage. The control-information transitions are modelled as a finite state automaton. Every operation of the class under test is considered as a mapping from starting states to a finishing states dependent upon the parameters passed. The possible parameter values are analysed for significant values which combined with the invocation of an operation can be used to represent stimuli applied to an object under test. State-based testing validates the expected transformations that can occur within a class. Classes are modelled using physical values assigned to the attributes of the class. The range of physical values is reduced by the use of a technique based on equivalence partitioning. This approach has a number of advantages over the conceptual modelling of a class, in particular the ease of manipulation of physical values and the independence of each operation from the other operations provided by an object. The technique when used in conjunction with other techniques provides an adequate level of validation for object-oriented programs. A suite of prototype tools that automate the generation of state-based test cases are outlined. These tools are used in four case studies that are presented as an evaluation of the technique. The code coverage achieved with each case study is analysed for the factors that affect the effectiveness of the state-based test suite. Additionally, errors have been seeded into 2 of the classes to determine the effectiveness of the technique for detecting errors on paths that are executed by the test suite. 92.5% of the errors seeded were detected by the state-based test-suite.