Automated Identification of Uniqueness in JUnit Tests

Abstract

In the context of testing, descriptive test names are desirable because they document the purpose of tests and facilitate comprehension tasks during maintenance. Unfortunately, prior work has shown that tests often do not have descriptive names. To address this limitation, techniques have been developed to automatically generate descriptive names. However, they often generated names that are invalid or do not meet developer approval. To help address these limitations, we present a novel approach to extract the attributes of a given test that make it unique among its siblings. Because such attributes often serve as the basis for descriptive names, identifying them is an important first step towards improving test name generation approaches. To evaluate the approach, we created a prototype implementation for JUnit tests and compared its output with human judgment. The results of the evaluation demonstrate that the attributes identified by the approach are consistent with human judgment and are likely to be useful for future name generation techniques.