No less than three papers from the Software Lab (SOLA) at the Institut für Software Engineering (ISTE) have made it onto the List of accepted papers for ESEC/FSE 2022 (The ACM Joint European Software Engineering Conference and Symposium on the Foundations of Software Engineering). The following research convinced the program committee representatives:
- "DynaPyt: A Dynamic Analysis Framework for Python" (Aryaz Eghbali, Michael Pradel),
- "Generating Realistic Vulnerabilities via Neural Code Editing: An Empirical Study" (Yu Nong, Yuzhe Ou, Michael Pradel, Feng Chen, Haipeng Cai) as well as
- "The Evolution of Type Annotations in Python: An Empirical Study" (Luca Di Grazia, Michael Pradel).
Empirical Study on the Development of Type Annotations
The latter paper, "The Evolution of Type Annotations in Python: An Empirical Study", received an ACM SIGSOFT Distinguished Paper Award. In their paper, Luca Di Grazia and Prof. Pradel provide results of a large-scale, empirical survey on type annotations. The researchers from the department investigated, whether developers are actually adding type annotations to the data types in their code more and more frequently and how this affects error rates. The research bases upon 1.4 million changes to Python code in 9,655 repositories hosted on GitHub. The result: The importance of type annotations is increasing; while only 3.8 percent of the examined commits contained one or more type annotations in 2015, the proportion already reached a value of 10.9 percent in 2021.
At the same time, however, the scientists still see considerable potential in the approach. Automated processes, such as neural type prediction, could add type annotations to existing code. Furthermore, the scientists recommend raising awareness among developers and relying more on gradual type checkers.
CrystalBLEU compares four and a half times more accurate
Back in October, Aryaz Eghbali and Prof. Pradel received an ACM SIGSOFT Distinguished Paper Award at the 37th IEEE/ACM International Conference on Automated Software Engineering (ASE 2022). The winning paper, "CrystalBLEU: Precisely and Efficiently Measuring the Similarity of Code", describes an improved approach to comparing the similarity of code. The researchers optimize the "BLEU score" intended for natural language, which already recognizes matches of completely different program fragments based on syntactic specifications and programming paradigms. Compared to this, the proposed CrystalBLEU approach works up to four and a half times more accurately.
Also at ASE 2022, Islem Bouzenia won the ACM Student Research Competition für sich entscheiden with his work "Detecting Inconsistencies in If-Condition-Raise Statements".