Contributing to pytest-django

Like every open-source project, pytest-django is always looking for motivated individuals to contribute to it’s source code. However, to ensure the highest code quality and keep the repository nice and tidy, everybody has to follow a few rules (nothing major, I promise :) )

Community

The fastest way to get feedback on contributions/bugs are usually to open an issue in the issue tracker.

Discussions also happends via IRC in #pylib on irc.freenode.org. You may also be interested in following @andreaspelme on Twitter.

In a nutshell

Here’s what the contribution process looks like, in a bullet-points fashion:

  1. pytest-django is hosted on GitHub, at https://github.com/pelme/pytest-django
  2. The best method to contribute back is to create an account there, then fork the project. You can use this fork as if it was your own project, and should push your changes to it.
  3. When you feel your code is good enough for inclusion, “send us a pull request”, by using the nice GitHub web interface.

Contributing Code

Getting the source code

  • Code will be reviewed and tested by at least one core developer, preferably by several. Other community members are welcome to give feedback.
  • Code must be tested. Your pull request should include unit-tests (that cover the piece of code you’re submitting, obviously)
  • Documentation should reflect your changes if relevant. There is nothing worse than invalid documentation.
  • Usually, if unit tests are written, pass, and your change is relevant, then it’ll be merged.

Since we’re hosted on GitHub, pytest-django uses git as a version control system.

The GitHub help is very well written and will get you started on using git and GitHub in a jiffy. It is an invaluable resource for newbies and old timers alike.

Syntax and conventions

We try to conform to PEP8 as much as possible. A few highlights:

  • Indentation should be exactly 4 spaces. Not 2, not 6, not 8. 4. Also, tabs are evil.
  • We try (loosely) to keep the line length at 79 characters. Generally the rule is “it should look good in a terminal-base editor” (eg vim), but we try not be [Godwin’s law] about it.

Process

This is how you fix a bug or add a feature:

  1. fork the repository on GitHub.
  2. Checkout your fork.
  3. Hack hack hack, test test test, commit commit commit, test again.
  4. Push to your fork.
  5. Open a pull request.

Tests

Having a wide and comprehensive library of unit-tests and integration tests is of exceeding importance. Contributing tests is widely regarded as a very prestigious contribution (you’re making everybody’s future work much easier by doing so). Good karma for you. Cookie points. Maybe even a beer if we meet in person :)

Generally tests should be:

  • Unitary (as much as possible). I.E. should test as much as possible only one function/method/class. That’s the very definition of unit tests. Integration tests are interesting too obviously, but require more time to maintain since they have a higher probability of breaking.
  • Short running. No hard numbers here, but if your one test doubles the time it takes for everybody to run them, it’s probably an indication that you’re doing it wrong.

In a similar way to code, pull requests will be reviewed before pulling (obviously), and we encourage discussion via code review (everybody learns something this way) or IRC discussions.

Running the tests

There is a Makefile in the repository which aids in setting up a virtualenv which can be used to invoke py.test to run pytest-django’s tests when developing.:

$ make

This will install a virtualenv with py.test and the latest stable version of Django. The virtualenv can then be activated with:

$ source bin/activate

Then, simply invoke py.test to run the test suite:

$ py.test --ds=tests.settings_sqlite

tox can be used to run the test suite under different configurations by invoking:

$ tox

There is a huge number of unique test configurations (63 at the time of writing), running them all will take a long time. All valid configurations can be found in tox.ini. To test against a few of them, invoke tox with the -e flag:

$ tox -e python3.3-1.7-postgres,python2.7-1.5-sqlite

This will run the tests on Python 3.3/Django 1.7/PostgeSQL and Python 2.7/Django 1.5/SQLite.

The tox and Travis CI configuration is generated by the script generate_configurations.py in the root directory. To add tests for a new Python or Django version, simply update the script and run it to regenerate the configuration files.

Measuring test coverage

Some of tests are executed in subprocesses. Because of that regular coverage measurements (using pytest-cov plugin) are not reliable.

If you want to measure coverage you’ll need to create .pth file as described in subprocess section of coverage documentation. If you’re using “setup.py develop” thing you should uninstall pytest_django (using pip) for the time of measuring coverage.

You’ll also need mysql and postgres databases. There are predefined settings for each database in tests directory. You may want to modify these files but please don’t include them in your pull requests.

After this short initial setup you’re ready to run tests:

$ COVERAGE_PROCESS_START=`pwd`/.coveragerc COVERAGE_FILE=`pwd`/.coverage PYTHONPATH=`pwd` py.test --ds=tests.postgres_settings

You should repeat above step for sqlite and mysql before next step. This step will create a lot of .coverage files with additional suffix for every process.

The final step is to combine all files created by different processes and generate html coverage report:

$ coverage combine
$ coverage html

Your coverage report is now ready in htmlcov directory.

Continous integration

Travis is used to automatically run all tests against all supported versions of Python, Django and different database backends.

The pytest-django Travis page shows the latest test run. Travis will automatically pick up pull requests, test them and report the result directly in the pull request.

Contributing Documentation

Perhaps considered “boring” by hard-core coders, documentation is sometimes even more important than code! This is what brings fresh blood to a project, and serves as a reference for old timers. On top of this, documentation is the one area where less technical people can help most - you just need to write a semi-decent English. People need to understand you. We don’t care about style or correctness.

Documentation should be:

  • We use Sphinx/restructuredText. So obviously this is the format you should use :) File extensions should be .rst.
  • Written in English. We can discuss how it would bring more people to the project to have a Klingon translation or anything, but that’s a problem we will ask ourselves when we already have a good documentation in English.
  • Accessible. You should assume the reader to be moderately familiar with Python and Django, but not anything else. Link to documentation of libraries you use, for example, even if they are “obvious” to you (South is the first example that comes to mind - it’s obvious to any Django programmer, but not to any newbie at all). A brief description of what it does is also welcome.

Pulling of documentation is pretty fast and painless. Usually somebody goes over your text and merges it, since there are no “breaks” and that GitHub parses rst files automagically it’s really convenient to work with.

Also, contributing to the documentation will earn you great respect from the core developers. You get good karma just like a test contributor, but you get double cookie points. Seriously. You rock.

Note

This very document is based on the contributing docs of the django CMS project. Many thanks for allowing us to steal it!