Mid Sweden University

Scientific software is vital for research across domains and can be reused when it is open-source. To promote this reuse, it is generally beneficial to adopt software engineering best practices to improve accessibility and popularity. However, given the distinct properties of scientific software, there is no consensus on how these practices are being or should be used in scientific software. Since previous evidence on this topic is primarily anecdotal or qualitative, this study used repository mining to quantitatively examine best practices and their relationship with popularity in 90 software engineering artifacts, which are examples of scientific software. The data varied significantly but showed that the studied artifacts generally did not prioritize software engineering best practices, and no significant relationships were found between these aspects and popularity. The results may suggest that scientific software developers prioritize scientific quality over software quality and that traditional software quality measures may not be suitable quality benchmarks in scientific software. However, accessibility issues were identified, highlighting potential societal concerns. Based on these findings, we offer practical advice for quality improvements from a software engineering perspective. Further research is needed to obtain more conclusive and general results.