Extracting implicit programming rules: comparing static and dynamic approaches

Abstract

Programs often follow implicit programming rules, such as, function call A must be followed by function call B. Rules of such kinds are rarely documented by developers. Nevertheless, programming rules play an important role in software testing and maintenance. For example, the rules can be used as test oracles to detect violations. If a programmer can be notified of these rules before updating the source code, the chances of generating defects due to rule violations might be minimized. Prior works have used static and dynamic analysis techniques to extract implicit programming rules, but none compares the effectiveness of the two techniques. In this paper, we have undertaken an empirical study to compare the two techniques when they are being used for extracting programming rules. Our results indicate that the performance of the dynamic analysis technique depends on the number and the diversity of the traces. Moreover, the dynamic analysis technique generates more precise rules than the static analysis technique if a diverse and sufficient number of test cases are provided.