Author:

Abstract:

In the mathematical sciences, researchers increasingly turn to computational experiments, both when validating new theoretical results and when modeling physical processes. The software used for these experiments is often developed by researchers with temporary positions, whose main objective is producing publications. Once the code produces the required figures, then there is often little motivation to spend time documenting the code and ensuring long-term experiment reproducibility. In addition, external libraries change over time, endangering long-term reproducibility if software versions are not well-documented. In a collaboration between the NUMA section at the Department of Computer Science, the Research Data Management Competence Centre, and the Research Coordination Office, a set of workflows is being developed, integrating iRODS, Docker, and GitLab, which automatically document the links between research results, code, and input data during the research process, with minimal effort from researchers themselves. The end goal is to ensure that computational experiments remain portable, well-documented and reproducible, both during the research process and years later.