Skip to content

pyiron/actions

Repository files navigation

pyiron actions

A centralized location for our GitHub actions to perform CI on python modules. It includes both custom actions, which we strive to make reusable everywhere (but whose default values are optimized for our repositories), and custom reusable workflows that are tuned specifically for our other pyiron repositories.

Note that for both security (i.e. use in pr-target triggered workflows) and flexibility, none of the actions call actions/checkout. You will need to do this yourself in the calling workflow in order for these actions to be productive.

The Actions

add-to-python-path

Adds an arbitrary number of local repo directories to the python path, pre-pending the path to that directory as it is found on the local runner. This is sent to the $GITHUB_ENV so the expanded path is available in all further steps of the same job.

In a pyiron context, this is important for repos that use pympipool in order to be able to use patterns like python -m unittest discover tests instead of cd tests; python -m unittest discover . by making sure tests (and/or sub-directories) are available in the python path.

build-docs

Combines your code's environment file with a separate environment file for building documentation (.support/environment-docs.yml from this repo by default) and uses sphinx to build a set of HTML documentation.

These docs wind up getting discarded when the virtual machine (VM) for your job shuts down, so the purpose of this action is to make sure that an external builder, e.g. readthedocs will also succeed.

build-notebooks

Uses papermill to make sure that selected notebooks in your repository build and execute OK. You can exclude a notebook from this check by naming it in an exclusion file (this is empty by default, but for repos inside the pyiron organization using the workflow here that calls this action, the default location is .ci_support/exclude). Notebooks are found in all sub directories of /notebooks. The files listed in the exclusion file must be given as relative paths to that folder.

cached-miniforge

A wrapper combining conda-incubator/setup-miniconda and actions/cache along with our own write-environment that allows you to easily make a cached conda environment from any number (>=1) of conda environment yaml files.

pip-check

Builds your environment with the cached-minforge action and then runs pip check.

pyiron-config

Necessary only if you plan to execute pyiron code -- perhaps as part of a build-notebooks call. It configures the pyiron environment for the VM.

unit-tests

Uses coverage to run python's unittest.

For pyiron repos, more detailed coveralls and codacy reports are generated in the same job that calls this action over in our reusable workflows. However, those elements require secrets and these actions are secret-free, so if you're not operating inside the pyiron organization you'll need to set that part up yourself.

update-env-files

Updates the environment file based on patterns in the PR title. Intended to be used in conjuction with Dependabot.

write-docs-env

Combines your environment file(s) and an additional environment file(s) and writes them, by default, to docs/environment.yaml. Intended to be committed to the repo and later read by, e.g., readthedocs.

write-environment

Another extremely versatile action: merges an arbitrary number of conda environment yaml files and writes them to a single environment.yml.

Smartly handles multiple channels and avoids duplicate dependencies, but doesn't do any environment pre-solving, so don't expect it to save you from incompatible version demands or anything like that. Still, extremely useful for common use cases like adding nbsphinx to your environment before building docs, etc.

Expected repository structure and defaults

To make full use of default values with these actions, your repository should have the following structure:

your_module/
|-- .binder/  # MyBinder config files (a workflow writes an env file here)
|-- .ci_support/
|     |-- environment.yml  # The conda environment file for your module
|-- docs/  # Necessary files for sphinx to built HTML documentation
|     |-- conf.py
|     |-- index.rst
|-- !environment.yml  # You MUST NOT have 'environment.yml' in your repo root -- our actions write to this location
|-- tests/  # The directory containing your test files
|     |-- benchmark/  # A total test suite under a time limit
|     |-- integration/  # Expensive tests run on a single system
|     |-- unit/  # To run on a variety of systems 
|-- your_module/  # The directory for your python module
|     |-- _version.py  # We will assume you are using versioneer and will ignore this when calculating coverage

By default, some of our actions add to the environment, e.g. build-docs uses the file .support/environment-docs.yml in this repository to add nbsphinx to the environment with the standard-docs-env-file input argument. In these situations you can always add additional files to the environment in the env-files argument, or stop these "standard" modules from getting added by overriding these arguments with an empty string.

Similarly, other defaults like where the docs or notebooks directories are, or even where the final environment file resides can always be overriden for each action.

Workflows (pyiron-organization specific)

Unlike the actions, reusable workflows in the .github/workflows directory enforce expectations about repository structure and lean heavily on the default values. Some properties, such as which env files to use, are still exposed, although in most cases you can continue to rely on the default values.

These workflows are (mostly) named based on the expected on values for triggering the workflow and ought to be applicable to any pyiron module.

Further, various workflows rely on the following repository secrets, all of which are organization wide inside the pyiron group (or generated by GitHub by default), but you'll need to specify them to use these workflows outside pyiron:

  • CODACY_PROJECT_TOKEN (Organization-wide)
  • DEPENDABOT_WORKFLOW_TOKEN (Organization-wide)
  • PYPI_PASSWORD (Organization-wide)
  • GITHUB_TOKEN (Don't worry about this, GitHub will produce it automatically)

Thankfully, if you're running these workflows for a pyiron-organization module, all you need to do in your calling workflow is set secrets: inherit.

Putting it all together, we get caller scripts that look like this:

# This runs jobs which pyiron modules should run on pushes or PRs to main

name: Push-Pull-main

on:
  push:
    branches: [ main ]
  pull_request:
    branches: [ main ]

jobs:
  pyiron:
    uses: pyiron/actions/.github/workflows/[email protected]
    secrets: inherit

Developing these actions

The actions here which are composed of other actions here and all of the reusable workflows all target @main on the main branch. If you want to develop these, and test those developments, it's not enough to have some other repo target your development branch, all these internal targets need to be redirected as well! For this we have a simple helper script: .support/update_actions_tag.sh. A typical use pattern should look like this:

  • Check out your branch ("foo", say)
  • Then from inside the actions home directory, invoke .support/update_actions_tag.sh to automatically rename all your targets to your branch name (@foo)
  • Target this branch from wherever for testing
  • Before merging REMEMBER to run .support/update_actions_tag.sh main to revert all your special targets!