Heuristics for the Identification of Class Integration Order in Object-Oriented Software Testing

Abstract

Dependency cycles among components (classes) represent a practical challenge when identifying the class integration order in object-oriented software testing. Classical approaches, such as “top-down” or “bottom-up” integration (or their combination) become less useful due to their acyclic characteristics. Some OO integration testing strategies explore the principle of breaking these dependency cycles, but broken dependencies imply that the target class will have to be stubbed when integrating and testing the source class, which increases the testing effort. This paper describes an integration testing strategy that guides software engineers to identify the class integration order with the minimum number of implemented stubs. This strategy can be directly applied to high level OO design (UML class) diagrams allowing the reduction of extra construction efforts related to additional classes needed in other strategies. The results from experimental studies have indicated its feasibility and effectiveness. The use of such integration testing strategy can be supported by a case tool, also described in this

paper.