Development of a feature for this repository should follow the workflow described by Vincent Driessen.
Here are the minimal procedure you should follow :
Create an issue on the github repository, describing the problem you will then address with your feature/fix. This is an important step as it forces one to think about the issue (to describe an issue to others, one has to think it through first).
-
Create a separate branch from
dev
, to work ongit checkout -b feature/myfeature dev
The convention is to always have
feature/
in the branch name. Themyfeature
part should describe shortly what the feature is about (separate words with-
). If you are fixing and error you can replacefeature/
byfix/
. -
Try to follow these conventions for commit messages:
- Keep the subject line short (i.e. do not commit more than a few changes at the time)
- Use imperative for commit messages
- Do not end the commit message with a period
You can use
to edit the commit message of your latest commit (provided it is not already pushed on the remote server). With
git commit --amend
--amend
you can even add/modify changes to the commit.
-
Push your local branch on the remote server
origin
git push
If your branch does not exist on the remote server yet, git will provide you with instructions, simply follow them.
To run tests locally, install the dependencies
pip install -r tests/test_requirements.txt
- Integration/unit tests
pytest tests
- Linting tests
and
flake8
You can fix the linting errors either manually or with the packagespylint
autopep8
orblack
for example.
Follow the steps of the github help to create the PR.
Please note that you PR should be directed from your branch (for example myfeature
) towards the branch dev
.
Add a line Fix #<number of the issue created in Step 2.0>
in the
description of your PR, so that when it is merged, it automatically
closes the issue once your code gets merged into the default branch of
your repository.
Here
is a list of other keywords you can use to automatically close the
issues
To have the test run automatically everytime you push to a pull request
you can add the bash commands under # command to run tests
of the
.travis.yml
file. For this you need to have an account by
Travis and link your repo to their service.
Soon GitHub action might take
care of this directly within github.
In the .travis.yml
file are many options commented out, you can have
very complexe schemes to test on many different python versions etc. For
more information look at Travis
doc for python.
Once you are ready to publish a release, branch off from dev
bash git checkout -b release/vX.Y.Z dev
For meaning of X, Y and Z version numbers, please refer to this semantic versioning guidelines.
In this branch, you should normally only update the version number in the CHANGELOG.md
and setup.py
files.
Your CHANGELOG.md
file could look like this before the release
## [unreleased]
### Added
- feature 1
- feature 2
### Changed
- thing 1
- thing 2
### Removed
- some stuff
Simply replace unreleased
by X.Y.Z
and add the date of release in ISO format, then add the structure for a new unreleased
version
## [unreleased]
### Added
-
### Changed
-
### Removed
-
## [X.Y.Z] - 20**-**-**
### Added
- feature 1
- feature 2
### Changed
- thing 1
- thing 2
### Removed
- some stuff
After pushing these changes, create a pull request from release/vX.Y.Z
towards master
and merge it in master
.
Locally, merge release/vX.Y.Z
into dev
git checkout release/vX.Y.Z
git pull
git checkout dev
git merge release/vX.Y.Z
And push your these updates to the remote
git push
The idea behind this procedure is to avoid creating a merge commit in dev
(because master
would otherwise have two merge commit for this release once you merge the next release).
Finally, create a release on github. Please choose master as the target for the tag and format the tag as vX.Y.Z
. In the description field simply copy-paste the content of the CHANGELOG
descriptions for this release and you're done!