Contributing to the user guide
MDAnalysis maintains two kinds of documentation:
This user guide: a map of how MDAnalysis works, combined with tutorial-like overviews of specific topics (such as the analyses)
The documentation generated from the code itself. Largely built from code docstrings, these are meant to provide a clear explanation of the usage of individual classes and functions. They often include technical or historical information such as in which version the function was added, or deprecation notices.
This guide is about how to contribute to the user guide. If you are looking to add to documentation of the main code base, please see Working with the code documentation.
ipython directive lets you put code in the documentation which will be run
during the doc build. For example:
.. ipython:: python x = 2 x**3
will be rendered as:
In : x = 2 In : x**3 Out: 8
Many code examples in the docs are run during the doc build. This approach means that code examples will always be up to date, but it does make the doc building a bit more complex.
Here is an overview of the development workflow for the user guide, as expanded on throughout the rest of the page.
Fork the MDAnalysis repository from the mdanalysis account into your own account
Set up an isolated virtual environment for your documentation
Add your new documentation.
Forking and cloning the User Guide
Go to the MDAnalysis project page and hit the Fork button. You will want to clone your fork to your machine:
git clone https://github.com/your-user-name/UserGuide.git cd UserGuide git remote add upstream https://github.com/MDAnalysis/UserGuide
This creates the directory UserGuide and connects your repository to the upstream (main project) MDAnalysis repository.
Creating a development environment
Create a new virtual environment for the user guide. Install the required dependencies, and activate the
nglview extension. We use
nglview for visualizing molecules in Jupyter notebook tutorials.
conda or similar (
micromamba), create a new environment with all the dependencies:
cd UserGuide/ conda env create --file environment.yml --quiet conda activate mda-user-guide jupyter-nbextension enable nglview --py --sys-prefix
Building the user guide
Navigate to the
doc/ directory and run
cd doc/ make html
The HTML output will be in
doc/build/, which you can open in your browser of choice. The homepage is
If rebuilding the documentation becomes tedious after a while, install the sphinx-autobuild extension.
Saving state in Jupyter notebooks
One of the neat things about
nglview is the ability to interact with molecules via the viewer. This ability can be preserved for the HTML pages generated from Jupyer notebooks by
nbsphinx, if you save the notebook widget state after execution.
Test with pytest and nbval
Whenever you add or modify notebook cells, you should make sure they run without errors and that their outputs are consistent, since they are part of the documentation as well.
We use a pytest plugin for this called nbval, it takes advantage of the saved notebook state and re-runs the notebook to determine if its output is still identical to the saved state. Thus, cells not only have to work (no errors), but also must give the same output they gave when they were saved.
To test all notebooks, just cd into
UserGuide/tests and run
If you want to test a particular notebook, check the the contents of pytest.ini, this file
defines flags that you can also pass directly to pytest.
For example, if you wanted to test the hole2 notebook:
pytest --nbval --nbval-current-env --nbval-sanitize-with ./sanitize_output.cfg ../doc/source/examples/analysis/polymers_and_membranes/hole2.ipynb
--nbval tells pytest to use nbval to test Jupyter notebooks,
to use the currently loaded python environment (make sure you actually loaded your environment)
instead of trying to use the one that was used when the notebook was saved and
--nbval-sanitize-with to point pytest to a file full of replacement rules like this one
regex: (.*B \[0.*B/s\])
This tells pytest to scan the outputs of all cells and replace the matching string with the word DOWNLOAD. This is called sanitization.
Exactly matching cell outputs between runs is a high bar for testing and tends to give false errors
– otherwise correct cells may give different outputs each time they are run (e.g. cells with code
that outputs memory locations).
To alleviate this, before testing each cell pytest will match its output against the regular
sanitize_output.cfg. This file contains replacements for strings that we know will vary.
Pytest will replace the dynamic output with these constant strings, which won’t change between runs and hence prevent spurious failures.
If your code correctly outputs variable strings each time its run, you should add a replacement
rule to the
sanitize_output.cfg file and try to make it as specific as possible.
On the hole2 notebook
The hole2 notebook is special in that it requires installation of extra software to run,
namely the hole2 program. If you test all the notebooks you may therefore run
into errors if hole2 is not installed. These errors can be generally ignored unless
you do specifically want to test the hole2 notebook. Of course, you should take
note of other errors that occur if hole2 is installed!
To run the hole2 notebook you’ll have to download hole2, compile it, and make sure your system can find
the hole2 executable. In UNIX-based systems this implies adding its path to the
environmental variable like this:
Adding changes to the UserGuide
As with the code, commit and push your code to GitHub. Then create a pull request. The only test run for the User Guide is: that your file compile into HTML documents without errors. As usual, a developer will review your PR and merge the code into the User Guide when it looks good.
It is often difficult to review Jupyter notebooks on GitHub, especially if you embed widgets and images. One way to make it easier on the developers who review your changes is to build the changes on your forked repository and link the relevant sections in your pull request. To do this, create a
gh-pages branch and merge your new branch into it.
# the first time
git checkout -b gh-pages
git merge origin/my-new-branch
Fix any merge conflicts that arise. Then edit
UserGuide/doc/source/conf.py and change the URL of the site, which is set to
site_url = "https://www.mdanalysis.org/UserGuide". Change it to your personal site, e.g.
site_url = "https://www.my_user_name.github.io/UserGuide"
Now you can build your pages with the
make github macro in the
UserGuide/doc/ directory, which builds the files and copies them to the top level of your directory.
You should be able to open one of these new HTML files (e.g.
UserGuide/index.html) in a browser and navigate your new documentation. Check that your changes look right. If they are, push to your gh-pages branch from the
git add .
git commit -m 'built my-new-branch'
git push -f origin gh-pages
On GitHub, navigate to your fork of the repository and go to Settings. In the GitHub Pages section, select the “gh-pages branch” from the Source dropdown. Check that your website is published at the given URL.
For each time you add changes to another branch later, just merge into gh-pages and rebuild.
git checkout gh-pages
git merge origin/my_branch
Automatically building documentation
Constantly rebuilding documentation can become tedious when you have many changes to make. Use sphinx-autobuild to rebuild documentation every time you make changes to any document, including Jupyter notebooks. Install
pip install sphinx-autobuild
Then, run the following command in the
python -m sphinx_autobuild source build
This will start a local webserver at http://localhost:8000/, which will refresh every time you save changes to a file in the documentation. This is helpful for both the user guide (first navigate to
UserGuide/doc) and the main repository documentation (navigate to
Using pre-commit hooks
Manually editing files can often lead to small inconsistencies: a whitespace here, a missing carriage return there. A tool called pre-commit can be used to automatically fix these problems, before a git commit is made. To enable the pre-commit hooks, run the following:
To perform the pre-commit checks on all the files, run the following:
pre-commit run --all-files
To remove the pre-commit hooks from your .git directory, run the following: