Shuffling tests

By default, every time you launch the testrunner it will run the tests in a specific order. However, if you want to ensure that your tests are well isolated then running them in a varying order can be helpful. (Proper isolation meaning your tests don’t depend on each others outcomes or side effects and thus the setup and tear down methods are written properly.)

The --shuffle option tells the test runner to shuffle the list of tests before running them:

>>> import os.path, sys
>>> directory_with_tests = os.path.join(this_directory, 'testrunner-ex')
>>> defaults = [
...     '--path', directory_with_tests,
...     '--tests-pattern', '^sampletestsf?$',
...     ]
>>> from zope import testrunner
>>> default_argv = 'test -u -m sample1 -t test_y0 --list-tests '

Running shuffled tests

When specifying the --shuffle option tests are ordered differently each time you run the tests:

>>> argv = (default_argv + '--shuffle').split()
>>> testrunner.run_internal(defaults, argv)
Tests were shuffled using seed number ...
Listing zope.testrunner.layer.UnitTests tests:
...
False

Note

The runner prints out a new piece of information which is the seed number used to generate the shuffled list of tests. This seed number can later be used to re-run the tests in exactly the same order to support debugging.

Specifying a seed number to control tests shuffling

Along with the --shuffle option comes the --shuffle-seed option which takes a seed number as an argument. If you want to reproduce test failures that happened during a randomly shuffled test run then simply write down the seed number that was printed out and run it again using the --shuffle-seed option. The order is guaranteed to be the same.

Note

There was an issue with the guaranteed order in zope.testrunner before 4.0.2: the order could change depending on dictionary iteration order, if you had multiple test layers. This bug was fixed in 4.0.2, but a side effect of the fix means that the new guaranteed order is likely to be different from what you used to get with an older zope.testrunner.

For example, using the seed number 0 will give us the following, stable, list of tests:

>>> argv = (default_argv + '--shuffle --shuffle-seed 0').split()
>>> testrunner.run_internal(defaults, argv)
Tests were shuffled using seed number 0.
Listing zope.testrunner.layer.UnitTests tests:
  test_y0 (sample1.sampletestsf.TestA)
  test_y0 (sample1.sampletests.test_one)
  test_y0 (sample1.sampletests.test1.TestA)
  test_y0 (sample1.sampletestsf)
  test_y0 (sample1.sampletests.test_one.TestA)
  test_y0 (sample1.sample13.sampletests)
  test_y0 (sample1.sample13.sampletests.TestA)
  test_y0 (sample1.sample11.sampletests)
  test_y0 (sample1.sample11.sampletests.TestA)
  test_y0 (sample1.sampletests.test1)
False
>>> testrunner.run_internal(defaults, argv)
Tests were shuffled using seed number 0.
Listing zope.testrunner.layer.UnitTests tests:
  test_y0 (sample1.sampletestsf.TestA)
  test_y0 (sample1.sampletests.test_one)
  test_y0 (sample1.sampletests.test1.TestA)
  test_y0 (sample1.sampletestsf)
  test_y0 (sample1.sampletests.test_one.TestA)
  test_y0 (sample1.sample13.sampletests)
  test_y0 (sample1.sample13.sampletests.TestA)
  test_y0 (sample1.sample11.sampletests)
  test_y0 (sample1.sample11.sampletests.TestA)
  test_y0 (sample1.sampletests.test1)
False

Whereas using the seed number 42 will give us the following, different but stable, list of tests:

>>> argv = (default_argv + '--shuffle --shuffle-seed 42').split()
>>> testrunner.run_internal(defaults, argv)
Tests were shuffled using seed number 42.
Listing zope.testrunner.layer.UnitTests tests:
  test_y0 (sample1.sample13.sampletests.TestA)
  test_y0 (sample1.sample13.sampletests)
  test_y0 (sample1.sampletests.test1)
  test_y0 (sample1.sampletests.test1.TestA)
  test_y0 (sample1.sample11.sampletests.TestA)
  test_y0 (sample1.sampletestsf)
  test_y0 (sample1.sampletests.test_one)
  test_y0 (sample1.sample11.sampletests)
  test_y0 (sample1.sampletestsf.TestA)
  test_y0 (sample1.sampletests.test_one.TestA)
False
>>> testrunner.run_internal(defaults, argv)
Tests were shuffled using seed number 42.
Listing zope.testrunner.layer.UnitTests tests:
  test_y0 (sample1.sample13.sampletests.TestA)
  test_y0 (sample1.sample13.sampletests)
  test_y0 (sample1.sampletests.test1)
  test_y0 (sample1.sampletests.test1.TestA)
  test_y0 (sample1.sample11.sampletests.TestA)
  test_y0 (sample1.sampletestsf)
  test_y0 (sample1.sampletests.test_one)
  test_y0 (sample1.sample11.sampletests)
  test_y0 (sample1.sampletestsf.TestA)
  test_y0 (sample1.sampletests.test_one.TestA)
False

Selecting a seed number without --shuffle

Note that the --shuffle-seed option must be used along with --shuffle option or tests will not be re-ordered:

>>> argv = (default_argv + '--shuffle-seed 42').split()
>>> testrunner.run_internal(defaults, argv)
Listing zope.testrunner.layer.UnitTests tests:
  test_y0 (sample1.sampletestsf.TestA)
  test_y0 (sample1.sampletestsf)
  test_y0 (sample1.sample11.sampletests.TestA)
  test_y0 (sample1.sample11.sampletests)
  test_y0 (sample1.sample13.sampletests.TestA)
  test_y0 (sample1.sample13.sampletests)
  test_y0 (sample1.sampletests.test1.TestA)
  test_y0 (sample1.sampletests.test1)
  test_y0 (sample1.sampletests.test_one.TestA)
  test_y0 (sample1.sampletests.test_one)
False