If you experience any problems with the package, feel free to put up a Github issue giving a detailed explanation the problem. This description should mention:
- What the problem is
- Why it is a problem
- If the problem is with the code itself, enough information to reproduce the behaviour
- (Optional) Ideas or suggestions for fixing the issue
To contribute to the code, clone the repository to your machine, and create a new branch for your contribution
git clone git@github.com:neffint/neffint.git
git checkout -b your-branch-name
Branch names should start with a category, followed by a /
and a descriptive name. The categories used are:
feature
- for new code featuresbugfix
- for bugfixes to existing codemisc
- for other changes, e.g. stuctural changes, documentation changes, addition of tests, etc. As a general rule these should not modify the behaviour of the code itself.
An example of a branch name: feature/add-asymptotic-correction
If desirable, sub-branches can be created to segment development into smaller work chunks. For large features this can be a good way to ensure more managable code review sizes, and thus improve the quality of the review.
The main
branch is used for stable releases. This is a protected branch, meaning all changes to this branch must go through a pull request with review. A pull request is submitted on Github, under the "Pull requests" tab on the repository home page.
When making contributions to the code, keep in mind that the code should be understandable to future readers of the code (including reviewers). Try therefore to make the code as readable as possible. Some concrete suggestions include, but are not limited to:
- Use type hints in function declarations
- Write docstrings for functions, classes and modules
- Use descriptive names for variables, functions and classes
As far as possible and practical, the code should be split into functions and classes with a single responsibility, with corresponding unit tests checking that it performs that task correctly.
For publishing a release and uploading the release to PyPI, use the following procedure:
- Press the
"Create a new release"
link on the sidebar on Github. - Create a new tag with the name
vX.Y.Z
onmain
, whereX
,Y
andZ
are the semantic version number of the release. This should be the same version number as__version__
inneffint/_version.py
. Give the release the same name. - Press
"Generate release notes"
to generate release notes. No additional notes are needed. - Press publish release. This will push a git tag to the latest commit to main, and run a CI job to build and publish the package to PyPI.
Alternatively, one can checkout to main
and run release.sh
, which will also push a git tag and build and publish the release. Using this alternative procedure, one should still preferably make a Github release from the pushed tag as well. The main differences are:
- When creating a Github release, the Github release and the git tag will automatically be the same, while one must manually verify that the version number in
neffint/_version.py
matches. When using the shell script,neffint/_version.py
and the git tag are automatically the same, but one must manually check that the Github release matches. release.sh
relies on usingsetup.py
, a format setuptools is moving away from. The Github release workflow does not rely on thesetup.py
format.- When using
release.sh
, one must have an authentication token locally on one's machine to upload to PyPI. The Github action uses a repository secret to authenticate with PyPI, so one only needs write access to the Github repository (for creating the release).