Running and writing tests¶
Good code needs tests.
A project like django CMS simply can’t afford to incorporate new code that doesn’t come with its own tests.
Tests provide some necessary minimum confidence: they can show the code will behave as it expected, and help identify what’s going wrong if something breaks it.
Not insisting on good tests when code is committed is like letting a gang of teenagers without a driving license borrow your car on a Friday night, even if you think they are very nice teenagers and they really promise to be careful.
We certainly do want your contributions and fixes, but we need your tests with them too. Otherwise, we’d be compromising our codebase.
So, you are going to have to include tests if you want to contribute. However, writing tests is not particularly difficult, and there are plenty of examples to crib from in the code to help you.
There’s more than one way to do this, but here’s one to help you get started:
# create a virtual environment virtualenv test-django-cms # activate it cd test-django-cms/ source bin/activate # get django CMS from GitHub git clone firstname.lastname@example.org:divio/django-cms.git # install the dependencies for testing # note that requirements files for other Django versions are also provided pip install -r django-cms/test_requirements/django-X.Y.txt # run the test suite # note that you must be in the django-cms directory when you do this, # otherwise you'll get "Template not found" errors cd django-cms python manage.py test
It can take a few minutes to run.
When you run tests against your own new code, don’t forget that it’s useful to repeat them for different versions of Python and Django.
Problems running the tests¶
We are working to improve the performance and reliability of our test suite. We’re aware of certain problems, but need feedback from people using a wide range of systems and configurations in order to benefit from their experience.
Please report any issues on our GitHub repository.
If you can help improve the test suite, your input will be especially valuable.
OS X users¶
In some versions of OS X,
gettext needs to be installed so that it is
available to Django. If you run the tests and find that various tests in
cms.tests.frontend raise errors, it’s likely that you have this problem.
A solution is:
brew install gettext && brew link --force gettext
(This requires the installation of Homebrew)
ERROR: test_copy_to_from_clipboard (cms.tests.frontend.PlaceholderBasicTests)¶
You may find that a single frontend test raises an error. This sometimes happens, for some users, when the entire suite is run. To work around this you can invoke the test class on its own:
manage.py test cms.PlaceholderBasicTests
and it should then run without errors.
ERROR: zlib is required unless explicitly disabled using --disable-zlib, aborting¶
If you run into that issue, make sure to install zlib using Homebrew:
brew install libjpeg zlib && brew link --force zlib
Advanced testing options¶
manage.py test --help for the full list of advanced options.
--parallel to distribute the test cases across your CPU cores.
--failed to only run the tests that failed during the last run.
--retest to run the tests using the same configuration as the last run.
--vanilla to bypass the advanced testing system and use the built-in
Django test command.
To use a different database, set the
DATABASE_URL environment variable to a
dj-database-url compatible value.
Running Frontend Tests¶
In order to be able to run them you need to install necessary dependencies as outlined in frontend tooling installation instructions.
Linting runs against the test files as well with
gulp tests:lint. In order
to run linting continuously, do:
Unit tests can be run like this:
If your code is failing and you want to run only specific files, you can provide
--tests parameter with comma separated file names, like this:
gulp tests:unit --tests=cms.base,cms.modal
If you want to run tests continuously you can use the watch command:
This will rerun the suite whenever source or test file is changed. By default the tests are running on PhantomJS, but when running Karma in watch mode you can also visit the server it spawns with an actual browser.
INFO [karma]: Karma v0.13.15 server started at http://localhost:9876/
On Travis CI we are using SauceLabs integration to run tests in a set of
different real browsers, but you can opt out of running them on saucelabs using
[skip saucelabs] marker in the commit message, similar to how you would skip
the build entirely using
We’re using Jasmine as a test framework and Istanbul as a code coverage tool.
In order to run integration tests you’ll have to install at least the version of django CMS from the current directory and django-app-helper into into your virtualenv. All commands should be run from the root of the repository. If you do not have virtualenv yet, create and activate it first:
virtualenv env . env/bin/activate
Then install minimum required dependencies:
pip install -r test_requirements/django-1.8.txt pip install -e .
Now you’ll be able to run a tests with this command:
The command will start a server, wait for a minute for the migrations to run
and will run integration tests against it. It will use
testdb.sqlite as the
database. If you want to start with a clean state you could use
Some tests require different server configuration, so it is possible that the server will stop, and another variation will start with different arguments. Take a look inside testserver.py if you need to customise the test server settings.
While debugging you can use the
--tests parameter as well in order to run test
gulp tests:integration --tests=pagetree gulp tests:integration --tests=loginAdmin,toolbar
If specified tests require different servers they will be grouped to speed things up, so the order might not be the same as you specify in the argument.
When running locally, it sometimes helps to visualise the tests output. For that
you can install casperjs visual debugging utility,
and run the tests with additional
--visual argument. It will try to
communicate with the server and display the progress of the test, which you then
can also rewind.
It might sometimes be useful not to restart the server when creating the tests,
for that you can run
python testserver.py with necessary arguments in one
gulp tests:integration --no-server in another. However you would
need to clean the state yourself if the test you’ve been writing fails.
Contributing tests is widely regarded as a very prestigious contribution (you’re making everybody’s future work much easier by doing so). We’ll always accept contributions of a test without code, but not code without a test - which should give you an idea of how important tests are.
What we need¶
We have a wide and comprehensive library of unit-tests and integration tests with good coverage.
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.
- Easy to understand. If your test code isn’t obvious, please add comments on what it’s doing.