What’s in a Display Name? An Empirical Study on the Use of Display Names in Open-Source JUnit Tests

Readability is an important aspect of any production or test code artefact. During development, testing and maintenance, the readability of a unit test can be a key contributor to its usefulness for a range of tasks, including refactoring, debugging, and test augmentation. Several strategies have been proposed both in academia and in industry to build readability into unit tests, e.g., test code summarisation and test naming conventions. Display names constitute an industry-led effort to incorporate natural language descriptions into unit tests. In this paper, we investigate the use of display names in a large dataset of open-source Java projects. Our study reveals that despite being a stable feature of the JUnit framework for at least five years, display names are not widely used in open-source projects yet. We analyse existing display names in terms of length, language and grammatical structure, explore the use of a large language model to generate display names similar to those open-source developers write, and develop a taxonomy of display name smells aimed at fostering a more cohesive and coherent use of the feature by developers.