ROCS: An object-oriented class-level testing system based on the Relevant Observable ContextS technique

Abstract

Given an algebraic specification of a class of objects, we define a fundamental pair as two equivalent terms generated by substituting all the variables on both sides of an axiom by normal forms. For any implementation error in the class, if it can be revealed by two equivalent terms in general, it can also be revealed by a fundamental pair. Hence, we need only select test cases from the set of fundamental pairs instead of equivalent pairs in general. We highlight an algorithm for selecting a finite set of fundamental pairs as test cases. Further, by using the relevant observable contexts technique, we highlight another algorithm to determine whether the objects resulting from executing a fundamental pair are observationally equivalent. If not, it reveals an implementation error. Using these algorithms, we have constructed a system to test object-oriented programs at class-level. We describe in detail the implementation of a prototype system, including the data structure of a Data member Relevant Graph (DRG) for the class, the procedures for the construction and path traversal of the DRG, the generation and execution of relevant observable contexts on the objects under test, and the reporting of implementation errors. The implementation illustrates an innovative idea of embedding testing algorithms into an interpreter to facilitate software testing.