Skip to main content

Helpers for better testing.

Project description


Exam is a Python toolkit for writing better tests. It aims to remove a lot of the boiler plate testing code one often writes, while still following Python conventions and adhering to the unit testing interface.


A simple pip install exam should do the trick.


Aside from the obvious “does the code work?”, writings tests has many additional goals and benefits:

  1. If written semantically, reading tests can help demonstrate how the code is supposed to work to other developers.

  2. If quick running, tests provide feedback during development that your changes are working or not having an adverse side effects.

  3. If they’re easy to write correctly, developers will write more tests and they will be of a higher quality.

Unfortunately, the common pattern for writing Python unit tests tends to not offer any of these advantages. Often times results in inefficient and unnecessarily obtuse testing code. Additionally, common uses of the mock library can often result in repetitive boiler-plate code or inefficiency during test runs.

exam aims to improve the state of Python test writing by providing a toolkit of useful functionality to make writing quick, correct and useful tests and as painless as possible.


Exam features a collection of useful modules:


Exam has some useful decorators to make your tests easier to write and understand. To utilize the @before, @after, @around and @patcher decorators, you must mixin the exam.cases.Exam class into your test case. It implements the appropriate setUp() and tearDown() methods necessary to make the decorators work.

Note that the @fixture decorator works without needing to be defined inside of an Exam class. Still, it’s a best practice to add the Exam mixin to your test cases.

All of the decorators in exam.decorators, as well as the Exam test case are available for import from the main exam package as well. I.e.:

from exam import Exam
from exam import fixture, before, after, around, patcher

The @fixture decorator turns a method into a property (similar to the @property decorator, but also memoizes the return value). This lets you reference the property in your tests, i.e. self.grounds, and it will always reference the exact same instance every time.

from exam.decorators import fixture
from exam.cases import Exam

class MyTest(Exam, TestCase):

    def user(self):
        return User(name='jeff')

    def test_user_name_is_jeff(self):
        assert == 'jeff'

As you can see, self.user was used to reference the user property defined above.

If all your fixture method is doing is constructing a new instance of type or calling a class method, exam provides a shorthand inline fixture syntax for constructing fixture objects. Simply set a class variable equal to fixture(type_or_class_method) and exam will automatically call your type or class method.

from exam.decorators import fixture
from exam.cases import Exam

class MyTest(Exam, TestCase):

    user = fixture(User, name='jeff')

    def test_user_name_is_jeff(self):
        assert == 'jeff'

Any *args or **kwargs passed to fixture(type_or_class_method) will be passed to the type_or_class_method when called.


The @before decorator adds the method to the list of methods which are run as part of the class’s setUp() routine.

from exam.decorators import before
from exam.cases import Exam

class MyTest(Exam, TestCase):

    def reset_database(self):

@before also hooks works through subclasses - that is to say, if a parent class has a @before hook in it, and you subclass it and define a 2nd @before hook in it, both @before hooks will be called. Exam runs the parent’s @before hook first, then runs the childs’. Also, if your override a @before hook in your child class, the overridden method is run when the rest of the child classes @before hooks are run.

For example, with hooks defined as such:

from exam.decorators import before
from exam.cases import Exam

class MyTest(Exam, TestCase):

    def reset_database(self):
        print 'parent reset_db'

    def parent_hook(self):
        print 'parent hook'

class RedisTest(MyTest):

    def reset_database(self):
        print 'child reset_db'

    def child_hook(self):
        print 'child hook'

When Exam runs these hooks, the output would be:

"prent hook"
"child reset_db"
"child hook"

As you can see even though the parent class defines a reset_database, because the child class overwrote it, the child’s version is run instead, and also at the same time as the rest of the child’s @before hooks.

@before hooks can also be constructed with other functions in your test case, decorating actual test methods. When this strategy is used, Exam will run the function @before is constructed with before running that particular test method.

from exam.decorators import before, fixture
from exam.cases import Exam

from myapp import User

class MyTest(Exam, TestCase):

    user = fixture(User)

    def create_user(self):

    def confirm_user(self):

    def test_confirmed_users_have_no_token(self):

    def test_user_display_name_exists(self):

In the above example, the confirm_user method is run immediately before the test_confirmed_users_have_no_token method, but not the test_user_display_name_exists method. The @before globally decorated create_user method still runs before each test method.

@before can also be constructed with multiple functions to call before running the test method:

class MyTest(Exam, TestCase):

    @before(func1, func2)
    def test_does_things(self):

In the above example, func1 and func2 are called in order before test_does_things is run.


The compliment to @before, @after adds the method to the list of methods which are run as part of the class’s tearDown() routine. Like @before, @after runs parent class @after hooks before running ones defined in child classes.

from exam.decorators import after
from exam.cases import Exam

class MyTest(Exam, TestCase):

    def remove_temp_files(self):

Methods decorated with @around act as a context manager around each test method. In your around method, you’re responsible for calling yield where you want the test case to run:

from exam.decorators import around
from exam.cases import Exam

class MyTest(Exam, TestCase):

    def run_in_transaction(self):
        yield  # Run the test

@around also follows the same parent/child ordering rules as @before and @after, so parent @arounds will run (up until the yield statement), then child @around``s will run. After the test method has finished, however, the rest of the child's ``@around will run, and then the parents’. This is done to preserve the normal behavior of nesting with context managers.


The @patcher decorator is shorthand for the following boiler plate code:

from mock import patch

 def setUp(self):
     self.stats_patcher = patch('mylib.stats', new=dummy_stats)
     self.stats = self.stats_patcher.start()

 def tearDown(self):

Often, manually controlling a patch’s start/stop is done to provide a test case property (here, self.stats) for the mock object you are patching with. This is handy if you want the mock to have default behavior for most tests, but change it slightly for certain ones – i.e absorb all calls most of the time, but for certain tests have it raise an exception.

Using the @patcher decorator, the above code can simply be written as:

from exam.decorators import patcher
from exam.cases import Exam

class MyTest(Exam, TestCase):

   def stats(self):
       return dummy_stats

Exam takes care of starting and stopping the patcher appropriately, as well as constructing the patch object with the return value from the decorated method.

If you’re happy with the default constructed mock object for a patch (MagicMock), then patcher can simply be used as an inline as a function inside the class body. This method still starts and stops the patcher when needed, and returns the constructed MagicMock object, which you can set as a class attribute. Exam will add the MagicMock object to the test case as an instance attribute automatically.

from exam.decorators import patcher
from exam.cases import Exam

class MyTest(Exam, TestCase):

    logger = patcher('coffee.logger')

The patcher.object decorator provides the same features as the patcher decorator, but works with patching attributes of objects (similar to mock’s mock.patch.object). For example, here is how you would use patcher to patch the objects property of the User class:

from exam.decorators import patcher
from exam.cases import Exam

from myapp import User

class MyTest(Exam, TestCase):

    manager = patcher.object(User, 'objects')

As with the vanilla patcher, in your test case, self.manager will be the mock object that User.objects was patched with.


The helpers module features a collection of helper methods for common testing patterns:


The track helper is intended to assist in tracking call orders of independent mock objects. track is called with kwargs, where the key is the mock name (a string) and the value is the mock object you want to track. track returns a newly constructed MagicMock object, with each mock object attached at a attribute named after the mock name.

For example, below track() creates a new mock with` as the ``cool_mock and tracker.heat as the heat_mock.

from exam.helpers import track

def test_roasting_heats_then_cools_beans(self, cool_mock, heat_mock):
    tracker = track(heat=heat_mock, cool=cool_mock)

This is a simple helper that just removes all folders and files at a path:

from exam.helpers import rm_f


Removes most of the boiler plate code needed to mock imports, which usually consists of making a patch.dict from sys.modules. Instead, the patch_import helper can simply be used as a decorator or context manager for when certain modules are imported.

from exam.helpers import mock_import

with mock_import('os.path') as my_os_path:
    import os.path as imported_os_path
    assert my_os_path is imported_os_path

mock_import can also be used as a decorator, which passed the mock value to the testing method (like a normal @patch) decorator:

from exam.helpers import mock_import

def test_method(self):
    import os.path as imported_os_path
    assert my_os_path is imported_os_path

Helper class that is itself callable, whose return values when called are configured via the tuples passed in to the constructor. Useful to build side_effect callables for Mock objects. Raises TypeError if called with arguments that it was not configured with:

>>> from exam.objects import call, effect
>>> side_effect = effect((call(1), 'with 1'), (call(2), 'with 2'))
>>> side_effect(1)
'with 1'
>>> side_effect(2)
'with 2'

Call argument equality is checked via equality (==) of the call` object, which is the 0th item of the configuration tuple passed in to the effect constructor. By default, call objects are just objects.

If you would like to customize this behavior, subclass effect and redefine your own call_class class variable. I.e.

class myeffect(effect):
    call_class = my_call_class


Exam has a subclass of the normal mock.Mock object that adds a few more useful methods to your mock objects. Use it in place of a normal Mock object:

from exam.mock import Mock

mock_user = Mock(spec=User)

The subclass has the following extra methods:

  • assert_called() - Asserts the mock was called at least once.

  • assert_not_called() - Asserts the mock has never been called.

  • assert_not_called_with(*args, **kwargs) - Asserts the mock was not most recently called with the specified *args and **kwargs.

  • assert_not_called_once_with(*args, **kwargs) - Asserts the mock has only every been called once with the specified *args and **kwargs.

  • assert_not_any_call(*args, **kwargs) - Asserts the mock has never been called with the specified *args and **kwargs.


Helpful fixtures that you may want to use in your tests:

  • exam.fixtures.two_px_square_image - Image data as a string of a 2px square image.

  • exam.fixtures.one_px_spacer - Image data as a string of a 1px square spacer image.


Useful objects for use in testing:

exam.objects.noop - callable object that always returns None. no matter how it was called.


The asserts module contains an AssertsMixin class, which is mixed into the main Exam test case mixin. It contains additional asserts beyond the ones in Python’s unittest.


Used when you want to assert that a section of code changes a value. For example, imagine if you had a function that changed a soldier’s rank.

To properly test this, you should save that soldier’s rank to a temporary variable, then run the function to change the rank, and then finally assert that the rank is the new expected value, as well as not the old value:

    old_rank = self.soldier.rank
    promote(self.soldier, 'general')
    self.assertEqual(self.soldier.rank, 'general')
    self.assertNotEqual(self.soldier.rank, old_rank)

Checking the old rank is not the same is the new rank is important. If, for some reason there is a bug or something to where self.soldier is created with the rank of general, but promote is not working, this test would still pass!

To solve this, you can use Exam’s assertChanges:

def test_changes_rank(self):
    with self.assertChanges(getattr, self.soldier, 'rank', after='general'):
        promote(self.soldier, 'general')

This assert is doing a few things.

  1. It asserts that the rank once the context is run is the expected general.

  2. It asserts that the context changes the value of self.soldier.rank.

  3. It doesn’t actually care what the old value of self.soldier.rank was, as long as it changed when the context was run.

The definition of assertChanges is:

def assertChanges(thing, *args, **kwargs)
  1. You pass it a thing, which which be a callable.

  2. assertChanges then calls your thing with any *args and **kwargs additionally passed in and captures the value as the “before” value.

  3. The context is run, and then the callable is captured again as the “after” value.

  4. If before and after are not different, an AssertionError is raised.

  5. Additionally, if the special kwarg before or after are passed, those values are extracted and saved. In this case an AssertionError can also be raised if the “before” and/or “after” values provided do not match their extracted values.


Similar to assertChanges, assertDoesNotChange asserts that the code inside the context does not change the value from the callable:

def test_does_not_change_rank(self):
    with self.assertDoesNotChange(getattr, self.soldier, 'rank'):

Unlike assertChanges, assertDoesNotChange does not take before or after kwargs. It simply asserts that the value of the callable did not change when the context was run.


Exam is MIT licensed. Please see the LICENSE file for details.

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

exam-0.10.6.tar.gz (15.0 kB view hashes)

Uploaded source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page