PEQtest: Testing Functional Equivalence

Abstract

AbstractRefactoring a program without changing the program’s functional behavior is challenging. To prevent that behavioral changes remain undetected, one may apply approaches that compare the functional behavior of original and refactored programs. Difference detection approaches often use dedicated test generators and may be inefficient (i.e., execute (some of) the non-modified code twice). In contrast, proving functional equivalence often requires expensive verification. Therefore, we propose PEQtest, which aims at localized functional equivalence testing thereby relying on existing tests or test generators. To this end, PEQtest derives a test program from the original program by replacing each code segment being refactored with program code that encodes the equivalence of the original and its refactored code segment. The encoding is similar to program encodings used by some verification-based equivalence checkers. Furthermore, we prove that the test program derived by PEQtest indeed checks functional equivalence. Moreover, we implemented PEQtest in a prototype and evaluate it on several examples. Our evaluation shows that PEQtest successfully detects refactored programs that change the program behavior and that it often performs better than the state-of-the-art equivalence checker PEQcheck.