Towards reliable rule mining about code smells: The McPython approach (Invited Lecture — Extended Abstract)

Code smell is a risky code pattern impacting code maintenance. Some of the code smells are defined by metrics (e.g., lines of code). Unfortunately, it is not clear how to set these thresholds for them. Goal: To propose a smell description language that allows querying code repositories to empirically determine impact of metric thresholds on severity of smells. Method: We propose a language, called McPython, that allows defining metric-based smells. We evaluate the expressiveness of the language by specifying some popular code smells. Results: McPython is a functional domain-specific language that allows defining smells as parameterized logical propositions with auxiliary functions. McPython code is translated to Python and executed on object-oriented representation of a code repository. Its current version is capable of expressing 7 code smells. Conclusion: Despite its limitations, McPython has the potential to help in investigating the impact of code smell parameters on their severity.

---

1. Main Track Invited Contributions

---

2. Opracowanie rekordu ze środków MEiN, umowa nr SONP/SP/546092/2022 w ramach programu "Społeczna odpowiedzialność nauki" - moduł: Popularyzacja nauki i promocja sportu (2024).