Background
With continuing urbanization, the challenges associated with the urban environment (such as air quality, pedestrian thermal comfort, wind loads on tall buildings - to name just a few) are increasingly relevant. Our ability to realistically capture processes such as the transport of heat, moisture, momentum and pollutants, and those of radiative transfer in urban environments, is key to understanding and facing these challenges. The turbulent nature of urban flows, along with the inherent heterogeneity and wide range of scales associated with the urban environment, results in a complex modelling problem. To address this, researchers within the Department of Civil and Environmental Engineering have developed uDALES: an open-source, multi-physics, microscale urban modelling tool, capable of performing simulations of urban airflow, heat transfer and pollutant dispersion. To coincide with the release of a new version of uDALES (v2.0.0), the group requested an overall code audit, a clean-up of their documentation and installation guide, and to implement software engineering best practices related to unit testing and continuous integration.
Our Contribution
Two members of our group took part in a week-long sprint with the research group. Using GitHub projects, we facilitated the sprint planning process, curating and assigning tasks to be tracked and discussed in the daily stand-ups. We reviewed the installation and "getting started" guides, as well as the user documentation; as fresh pairs of eyes we were well placed to suggest improvements in this regard. This included an audit of the build scripts for local and remote cluster installation. We assessed the overall state of the code with an aim to identify and isolate parts of the code for unit-testing. We also reviewed the current continuous integration workflows being run as GitHub actions, looking for ways to improve and streamline these.
Outcomes
By the end of the sprint, the group were able to publish a new release of uDALES (v2.1.0), containing significant updates to the documentation related to installation and getting started using the code; a more intuitive user interface; a range of improvements to the build process across various architectures and compiler toolchains, enabling specific build configurations; and an enforced, modern Fortran language standard. We provided the group with working examples of units tests for their core Fortran code using pFUnit, and for their pre- and post-processing MATLAB code, running in GitHub actions. The group were also better equipped with git and GitHub skills, in particular with regard to the feature branch workflow, as well as guidance on best practices for continuous integration workflows.