Regression Test Case Prioritization – A Contribution-Based Approach

Regression test case prioritization techniques have traditionally been studied as a function of objective metrics such as code coverage and fault proneness, which require much data analysis and computation from software release to release. Moreover, such techniques have traditionally been evaluated as a function of fault detection effectiveness. Consequently, not only applying the techniques is an expensive exercise, the results focus on fault detection rates rather than release-readiness, a more practical interpretation of software quality. We therefore propose to approach regression test case prioritization with a completely different tack: We propose a novel approach to prioritize based on subjective but inexpensive, one-off human assessment of test cases’ contribution to predicting release-readiness, thereby eliminating the intensive data analysis and computation required to begin the prioritization process. Further, to align closer to industry software processes, we define the effectiveness of a prioritization as the time required to pronounce a software application release-ready. We believe these novel prioritization and evaluation methods will bring a new perspective to the state-of-theart.