Santhosh Sundaram — Google Summer of Code 2024 Final Report

pybamm-cookie: A CLI for generating PyBaMM based projects

Google Summer of Code NumFOCUS PyBaMM pybamm-cookie PyPI Copier

Blogs Project page on the Google Summer of Code website Proposal on Google Docs


View my GitHub profile View my LinkedIn profile

This page serves as a permalink to the final report that provides details of Santhosh Sundaram’s project in the Google Summer of Code 2024 program.

📖 TL;DR#

A brief of the tasks I completed throughout the GSoC 2024 program are listed below.

  • A CLI package for generating PyBaMM-based projects using copier templating engine published at PyPI.
  • Improved entry points API to accommodate third-party models as well as parameter sets.
  • Automated testing workflows and testing infrastructure for the CLI and end users using pytest.
  • Implemented support for various build-backends such as setuptools and hatch, versionings like hatch-vcs and setuptools-scm, packaging, and for repository hosting services like GitHub and Gitlab.
  • Support for documentation using sphinx for the generated project for end users.

🔋 About PyBaMM#

PyBaMM (Python Battery Mathematical Modeling) is an open-source Python module for the simulation of battery models. PyBaMM aims to provide a flexible, user-friendly, and high-performance platform for researchers, engineers, and developers working in the field of battery modeling and electrochemical simulations. It provides inbuilt battery models and parameter sets to visualise, simulate and analyse the overall interpretability of a battery model. It is fiscally sponsored by NumFOCUS and The Faraday Institution in the U.K

📄 Project abstract and motivation 🚀#

The project aims to build and publish pybamm-cookie as a jinja template for new PyBaMM-based projects. A copier template is designed to simplify the setup process of the development environment for researchers and scientists interested in utilizing PyBaMM for battery modeling who may lack familiarity with managing Python environments or repositories. The project intends to enhance the accessibility and usability of PyBaMM for newbies and experienced users alike. Providing a standardized template with best practices and automation tools would lower the barrier to entry for adopting PyBaMM in battery modeling projects. This would in turn make battery modeling more accessible, efficient, and collaborative for the research community. The project also intends to implement Model entry points, allowing community contributors to create and share models of their repositories using the copier template without directly adding them upstream. This would not only let community contributors retain ownership and choose license terms but also grant flexibility to the PyBaMM team in supporting models, Including all of GitHub’s functionality and infrastructure contained within the template.

🍀 Benefits to the community#

This project aims to lower the barrier to adopting PyBaMM in battery model projects indirectly increasing the userbase of the community by automating the development environments quickly and efficiently making battery modeling more accessible, efficient, and collaborative. Additionally, the support for Model entry points allows community contributors to create and share models from their repositories without directly adding them upstream, enabling them to retain ownership, choose license terms, and leverage GitHub’s functionality. This flexibility not only empowers contributors but also allows the PyBaMM team to focus on supporting models that best align with their goals, ultimately fostering a more dynamic and inclusive research community.

🧑‍💻 Work done#

🍪 pybamm-cookie: A CLI for generating PyBaMM-based projects using copier#

Due to the lack of extensive programming knowledge, especially in the area of the Python development ecosystem, researchers using PyBaMM find it difficult to set up a Python package and add their entry points to the project. That’s where my project comes in, you could set up a PyBaMM-based Python package using a single CLI command that accommodates the features of PyBaMM and you could add your third-party parameter sets and models in that package and test it! Without having to go through the overwhelming hassles of setting up a Python package for your third-party parameter sets and models. A user can define parameter sets and models that can be used for different battery chemistries and configurations and simulations. These can then be added to the generated project using pybamm-cookie CLI and can be manually added by mapping it in pyproject.toml under the [project.entry-points.pybamm_parameter_sets] group. This would then load the parameter set through the parameter_sets or models instance within the generated Python package which could be then packaged, published and accessed by a third-party user.

🔗 Unification of parameter sets and model entry points#

Parameter sets and model entry points in both these APIs had the same behaviour, while one was getting the entry points from the parameter_sets group under pyproject.toml, the other one was getting them from the models group. While it was possible to have two different APIs in the form of different files for them, I chose to unify them by initialising the entry point API with the group name in its constructor method, which gets loaded only once after the main import of the package. This reduced the complexity of code and avoided duplication of code. This would create a doubleton instance, one for parameter sets and the other for model entry points which get initialised through their group name while pybamm gets imported. This feature is planned to upstream to PyBaMM once they are done and tested in the pybamm-cookie template. While the PyBaMM project would in the future try to accommodate solvers as an entry point as well, it would be beneficial in the code base to have one single source of API to manage all the entry points.

🔩 Testing infrastructure and documentation support for generated projects#

Using github-actions and nox I automated the tests and separated out the sessions for testing the generated project using the CLI to ensure proper generation of the project against multiple inputs, and a session for testing the units inside the project like the entry point APIs and the user-defined tests. The API tests load the defined models and parameter sets through the entry points and ensure that they return the same object as the ones loaded without the entry points. The documentation is user-faced and basic sphinx configuration is set up in the project inside the template. A user could use either Numpy or Google styles for documenting using doctstrings which gets rendered to HTML using sphinx. A separate nox session for buildling the website is also present and could be locally initialised by executing nox -s docs. The documentation builds are also tested in CI for the generated project to ensure better coverage.

👷 Pull requests#

🛠️ Issues#

✍️ Blogs#

I wrote a series of blogs on my journey throughout the coding period. They can be accessed on my personal blog site https://santacodes.github.io/series/gsoc/

🔮 Future work#

  • Adding model entry points API upstream to PyBaMM - Replicating the model and parameter sets entry points API in upstream PyBaMM to accommodate PyBaMM’s predefined models and parameter sets using entry points.
  • Separating IDAKLU solver from PyBaMM pure Python package - Packaging IDAKLU solver as a separate dependency with its own versioning, and releases to make PyBaMM a pure Python package. This would speed up workflows, build-times, and better packaging to make IDAKLU optional, and would potentially reduce wheel sizes.
  • Add uv backend for nox sessions in CI for faster package resolution times - Supporting uv by astral-sh backend in nox would decrease the CI times as package resolutions would be faster.
  • Add documentation on pybamm-cookie to pybamm.org - Document about entry points and pybamm-cookie usage in the PyBaMM documentation site.

Acknowledgments#

I would like to extend my deepest gratitude to the Google Summer of Code program, NumFOCUS, and the entire PyBaMM team for providing me with an incredible and enriching experience this summer. This journey marked my first significant contribution to the development and maintenance of open-source scientific software on such a large scale, and the lessons I’ve learned have been invaluable.

I am truly honored to have been mentored by, Saransh Chopra, Ferran Brosa Planella, Agriya Khetarpal and Arjun Verma. Their guidance and unwavering support, both in technical aspects and in fostering a positive, collaborative environment, have been immensely uplifting.

I would also like to sincerely thank Valentin Sulzer for their invaluable reviews and guidance throughout various stages of my contributions to PyBaMM and pybamm-cookie. The collective efforts of the entire PyBaMM team, their insights, and their collaborative spirit have been the driving forces behind the successes achieved during this project.

🎬 Footnotes#

I am grateful to have completed the Google Summer of Code 2024 program and I hope to contribute to PyBaMM and its thriving codebase and community further in the future and help ameliorate the development and success of the PyBaMM ecosystem.

I am eager to incorporate the technical skills I’ve gained into my academic, personal, and professional projects in the future.

🔭 References#

On this page