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.
pip install exam should do the trick.
Aside from the obvious "does the code work?", writings tests has many additional goals and bennefits:
- If written semantically, reading tests can help demostrate how the code is supposed to work to other developers.
- If quick running, tests provide feedback during development that your changes are working or not having an adverse side effects.
- 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 ineffecient and unnessarily obtuse testing code. Additionally, common uses of the mock library can often result in repetitive boiler-plate code or ineffeciency 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 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
@patcher decorators, you must mixin the
exam.cases.Exam class into your test case. It implements the appropriate
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
@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): @fixture def user(self): return User(name='jeff') def test_user_name_is_jeff(self): assert self.user.name == 'jeff'
As you can see,
self.user was used to reference the
user property defined above.
If all your fixture method is doing is contructing 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 witll 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 self.user.name == 'jeff'
**kwargs passed to
fixture(type_or_class_method) will be passed to the
type_or_class_method when called.
@before decorator adds the method to the list of methods which are run as part of the class's
from exam.decorators import before from exam.cases import Exam class MyTest(Exam, TestCase): @before def reset_database(self): mydb.reset()
@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 child class's
@before hook first, then runs the parents'.
The compliment to
@after adds the method to the list of methods which are run as part of the class's
tearDown() routine. Like
@after runs child class
@after hooks before running their parents'
from exam.decorators import after from exam.cases import Exam class MyTest(Exam, TestCase): @after def remove_temp_files(self): myapp.remove_temp_files()
@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): self.stats_patcher.stop()
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 defaut 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.
@patcher decorator, the above code can simply be written as:
from exam.decorators import patcher from exam.cases import Exam class MyTest(Exam, TestCase): @patcher('mylib.stats') 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 (
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')
helpers module features a collection of helper methods for common testing patterns:
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
tracker.cool` as the ``cool_mock and
tracker.heat as the
from exam.helpers import track @mock.patch('coffee.roast.heat') @mock.patch('coffee.roast.cool') def test_roasting_heats_then_cools_beans(self, cool_mock, heat_mock): tracker = track(heat=heat_mock, cool=cool_mock) roast.perform() tracker.assert_has_calls([mock.call.heat(), mock.call.cool()])
This is a simple helper that just removes all folders and files at a path:
from exam.helpers import rm_f rm_f('/folder/i/do/not/care/about')
Removes most of the boiler plate code needed to mock imports, which usually consists of making a
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
from exam.helpers import mock_import @mock_import('os.path') def test_method(self): import os.path as imported_os_path assert my_os_path is imported_os_path
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
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
assert_not_called_once_with(*args, **kwargs)- Asserts the mock has only every been called once with the specified
assert_not_any_call(*args, **kwargs)- Asserts the mock has never been called with the specified
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 objectgs for use in testing:
exam.objects.noop - callable object that always returns
None. no matter how it was called.
Exam is MIT licensed. Please see the
LICENSE file for details.