Exploring Sustainability in Scientific Software through Code Quality & Test Coverage Metrics
Only sustainable scientific software projects maintain consistent test coverage above 50%...
A new study from Sheikh Md. Mushfiqur Rahman, Gregory R. Watson, and Nasir U. Eisty, accepted for the Platform for Advanced Scientific Computing (PASC) 2026 conference in Bern, Switzerland, tackles the overlooked challenge of sustainability in scientific open-source software (SciOSS). By analyzing the CASS Software Portfolio—a collection of scientific software projects—the team classified projects as sustainable or unsustainable and compared their code structure, test coverage, and the relationship between code quality and testing.
The results reveal stark differences: sustainable projects maintain significantly higher and more consistent test coverage, with clear correlations between code quality metrics and test effectiveness. In contrast, unsustainable projects show weaker patterns, and overall test coverage across scientific software remains alarmingly low. The authors also found that high code complexity and coupling directly reduce testability, making it harder to maintain long-term software health. This work provides a practical, data-driven framework for evaluating and monitoring sustainability in scientific software, with direct implications for quality assurance and funding decisions.
- Sustainable SciOSS projects show >50% higher and more consistent test coverage vs. unsustainable ones.
- High code complexity and coupling in scientific software reduce testability and long-term maintainability.
- Study accepted at PASC 2026; analyzed CASS Software Portfolio from Univ. of Alabama & Oak Ridge National Lab.
Why It Matters
Data-driven sustainability metrics enable research labs to prioritize funding and maintenance for software that will last.