Nosedjango brings the goodness of Nose and its plugin ecosystem to the world of Django testing. Nose already has plugins for multiprocessing, coverage, tagging, profiling, skipping, xunit plugin and most everything else you could need. Nosedjango means you don't have to re-invent those wheels.
Perhaps the most compelling case for using nosedjango is the performance gains seen when using the multiprocess module. (both runs on a core i7 laptop with a fixture-intensive test suite)
Normal Django testrunner using sqlite
$ ./manage.py test account ... Ran 65 tests in 507.930s OK
NoseDjango with 8 processes
$ nosetests --with-doctest --with-django --django-settings pstat.settings --with-django-testfs --with-django-sqlite --processes 8 pstat.account ... Ran 65 tests in 35.731s
Nose is ~14x faster.
In addition, Nosedjango provides its own plugin system to hook in to the low level django-specific testing operations. Included with Nosedjango are plugins to do things like:
- Create an isolated file storage location for testing.
- Use an in-memory sqlite database.
- Start a cherrpy server for integration-style tests.
- Make it easier to test Celery.
- Create and use a Sphinx search index for fulltext search tests.
- Open an SSH tunnel for things like Selenium that might need outside resources.
- Run Selenium2 functional tests in a headless virtual frame buffer.
- Selectively switch out settings from the command line for different kinds of tests.
The plugin takes care of finding your applications settings.py file and creating/tearing down test database. It also has support for fixtures and it has experimental mechanism that wraps the tests in transactions to speed up testing.
This plugin works with Django versions 1.0 through django 1.3.
Unit tests can be run with following command:
nosetests --with-django [nose-options]
In addition to default nose command line options, nosedjango offers following options:
|Specify a custom Django settings MODULE. The specified MODULE needs to be found in sys.path.|
An easy win for Nosedjango out of the box is the ability to safely distribute tests across multiple processes, dramatically speeding up test runs on multicore machines.
In the simplest case, the following will run your tests distributed across two cores using in-memory sqlite databases and separate file storage locations to minimize file collision conflicts:
nosetests --with-django --with-django-sqlite --with-django-testfs --processes=2 <your_project_module>
For very small test suites or test suites that don't use fixtures, the overhead from starting multiple processes can result in the full test run actually being slower with multiple processes than with a single process.
Complex test suites might require some adaptation to support parallel test running. If tests rely on things like hardcoded file paths or shared external resources, these will need to be made generic. Usually this is as easy as using a a NamedTemporaryFile instead of a hardcoded path.
Installation via Pip is straightforward:
$ pip install -e git+git://github.com/winhamwr/nosedjango.git#egg=nosedjango
An easy way to try out your nosedjango installation is to try the test suite yourself:
$ python setup.py nosetests
Nosedjango makes a few decisions differently than Django's normal testrunner and depending on your project, you might need to make adjustments for all of your tests to run properly.
By default, Nose only runs doctests if the --with-doctest option is included. Nosedjango respects this default rather than the Django default, so if you'd like to run your doctests, add --with-doctest to your options.
Nosedjango relies on Nose's test discovery method, which means that Nose might find some tests that weren't being run by Django.
For performance reasons, the database schema is only created once. If you have tests that alter the schema (migration tests for example), you'll need to add a rebuild_schema attribute to those tests.
class LargerUsernameTestCase(TestCase): rebuild_schema = True def setUp(self): if settings.DATABASES['default']['ENGINE'] == 'django.db.backends.mysql': from django.db import connection # pylint: disable=W0404 cursor = connection.cursor() cursor.execute("ALTER TABLE `auth_user` CHANGE COLUMN `username` " "`username` VARCHAR(130) " "COLLATE utf8_unicode_ci NOT NULL") def test_long_username(self): # test some stuff
Nose supports module-level fixtures, and so does Nosedjango. This means that if you have a fixtures variable floating around in a test module, Nosedjango will load it.
fixtures = ['cheese.json', 'cakes'] def test_cheesecake(): # do something...
The cache is cleared between each test run, as is the case with newer versions of Django. If you have tests that depend on other tests modifying the cache (tsk tsk tsk), then you will need to modify those tests for them to work under Nosedjango.
Nosedjango's plugin system is heavily inspired by Nose's own system and provides loads of hooks in to the Django test-running process. Nosedjango plugins are actually just Nose plugins themselves that have access to extra hooks. To see available hooks, check out nosedjango.plugins.base_plugin.Plugin. Plugins should extend that class.
Better documentation is hopefully forthcoming, but reading the source for the included file_storage_plugin and sqlite_plugin should provide clues along with Nose's documentation on writing plugins. One example of solving very project-specific testing needs is the NoseDjango plugin located at https://github.com/jlward/nosedjango-pstat
- Multiprocess testing only currently works with in-memory sqlite. This is very fixable though and pull requests are welcome.
- Nosedjango is broken with Nose 1.0 and higher due to changes in Nose's Multiprocessing module. This is currently being investigated.
Original plugin courtesy of Victor Ng <firstname.lastname@example.org> who rewrote Jason Pellerin's original nose-django plugin.
This software is licensed with GNU LESSER GENERAL PUBLIC LICENSE version 3 or (at your option) any later version. See COPYING for more details.