Comparison analysis of two test case prioritization approaches with the core idea of adaptive

摘要

Test case prioritization problem (TCP) has been widely discussed. It aims to controlling the test case execution sequence to improve the effectiveness of software testing. The key issue of TCP is to identify which test cases can provide useful information for failure detection and fault localization. So far, many TCP approaches have been proposed. Among them, Adaptive Random Testing (ART) and Dynamic Random Testing (DRT) are two of the most popular approaches to solve TCP with a basic idea borrowed from Cybernetics: adaptive. Both ART and DRT has been widely explored and observed with good performances in experimental studies. Nevertheless, although they are proposed by two related research groups, they are developed independently and in parallel. In fact, their mechanisms have many similarities and differences and, for the completeness of the domains of Adaptive Testing and Software Cybernetics, many issues concerning the comparison between these two approaches should be further explored. In this paper, we specifically explores the relationship between these two adaptive TCP approaches. Their mechanisms are described respectively with explorations of their distinctions, similarities, and respective characteristics. Moreover, based on these explorations, we analyse their advantages from the aspects of failure detection and fault understanding. During the analysis, a symbolic-graphic combination method is applied. Finally simulation based on real-life programs is conducted to observe our analysis. Our comparison analysis can support the selection of a proper testing approach according to various practical environments with different targets. Furthermore, the clarification of the two easily confused concepts is also a complement for the framework of Adaptive Testing and Software Cybernetics.