Modification-Impact based Test Prioritization for Process-Driven Applications

Abstract

Automated regression tests are essential, especially when practicing continuous software engineering techniques. In the case of Process-Driven Applications (PDA), the tests must consider executable process models typically using the BPMN notation and external software services as crucial parts of the application. The complexity of processes is constantly increasing and hence also the need for more tests to ensure the correctness of the PDA. Running all implemented tests in an arbitrary order is time-consuming and, causes high costs. It is more promising in terms of early fault detection to start with those tests that are more likely affected by the changes which have been implemented for the upcoming release under test. For this purpose, we have identified the different types of modifications at the control flow and activity level and evaluated their impact for the intended test prioritization. Furthermore, our concept considers the impact of adaptions at the data-flow level. The approach has been implemented prototypically for the Camunda BPM platform. Experimental results through various case studies proved an earlier detection of errors compared to traditional process coverage techniques.