Prioritization of Program Elements Based on Their
Testing Requirements

Abstract

Even after thorough testing of a program, usually a few bugs still remain. These residual bugs are usually uniformly distributed throughout the code. It is observed that bugs in some parts of a program can cause more frequent and more severe failures compared to those in other parts. It should, then be possible to prioritize the statements, methods and classes of an object-oriented program according to their potential to cause failures. Once the program elements have been prioritized, the testing effort can be apportioned so that the elements causing most frequent failure are tested more. Based on this idea, in this paper we propose a program metric called the influence of program elements. Influence of a class indicates the potential of class to cause failures. In this approach, we have used the intermediate graph representation of a program. The influence of a class is determined through a forward slicing of the graph. Our proposed program metric can be useful in applications such as coding, debugging, test case design and maintenance etc.