A flyweight mocking helper for Python
Posted by Tom Moertel Mon, 07 Nov 2011 05:05:00 GMT
Recently, I needed a lightweight mocking solution when adding unit tests to some existing Python code. I could have used one of the many Python mocking libraries, but I only had to replace a few module functions during testing, so I just wrote a tiny mocking helper rather than add a dependency to the project.
The helper is simple and surprisingly versatile:
import contextlib
import functools
@contextlib.contextmanager
def mocked(func,
expecting=None, returning=None, raising=None, # specs
replacement_logic=None, called=1):
"""Stub out and mock a function for a yield's duration."""
if (returning, raising, replacement_logic).count(None) < 2:
raise ValueError("returning, raising, and replacement_logic "
"are incompatible with each other")
# default logic for implementing mock fn: return or raise per specs
def default_logic(*_args, **_kwds):
if raising:
raise raising
return returning
# prepare wrapper to replace mocked function for duration of yield
invocations = [0]
@functools.wraps(func)
def replacement(*args, **kwds):
if expecting is not None:
assert expecting == (args, kwds) # did we get expected args?
invocations[0] += 1
return (replacement_logic or default_logic)(*args, **kwds)
# replace mocked function, yield to test, and then check & clean up
module = sys.modules.get(func.__module__)
setattr(module, func.__name__, replacement)
try:
yield # give control back to test for a while
assert invocations[0] == called # was mock called enough?
finally:
setattr(module, func.__name__, func)
def uncalled(func):
"""Require that a function not be called for a yield's duration."""
return mocked(func, called=0)
The idea is to wrap test code that requires mocked external functions with a mocking helper. Here’s an example:
def test_ml_errors_must_be_reported(self):
"""When an error occurs, it must be reported; nothing must be sent."""
data = self._request_data()
exception = mls.MlError()
exception.exc = '[[[error description]]]'
with mocked(mls.subscriber_count, raising=exception):
with uncalled(sms.send):
resp = self.app.request("/mailings", method='POST', data=data)
self.assertEqual(resp.status, '200 OK')
self.assertIn(exception.exc, resp.data) # must be reported
In this test method, I’m making a simulated POST request to a web service that’s supposed to check how many users are in a mailing list and then, possibly, send some SMS messages. In this case, however, I want to simulate that an error occurs when talking to the mailing-list service. The test must verify that the error is reported and, crucially, that no messages are sent.
The library modules for talking to the mailing-list service and the
SMS service are called mls and sms. So,
for the duration of the simulated request, I’m replacing two functions
in these modules with mock versions.
The mock version of mls.subscriber_count, when called,
will raise the exception I’m trying to simulate. The mock version of
sms.send, however, must not be called; if it is, the
uncalled mocking helper will alert me by raising an
exception.
So the mocking helpers not only temporarily install mock implementations of functions but also assert that those mock implementations are (or are not) called as expected. In the following code, for example, I use this capability to make sure that the mailing-list service is asked to get the subscriber count for the mailing list having the right key. I also simulate the service returning a subscriber count of 123.
def test_unconfirmed_msgs_must_be_confirmed(self):
"""An unconfirmed msg must be confirmed and not sent."""
mlkey = 'TEST_ML_KEY'
body = 'Test message!'
data = self._request_data(mlkey=mlkey, body=body)
with mocked(mls.subscriber_count, expecting=((mlkey,),{}), returning=123):
with uncalled(sms.send):
resp = self.app.request("/mailings", method='POST', data=data)
self.assertEqual(resp.status, '200 OK')
self.assertIn('Reply CONFIRM', resp.data) # we must ask for confirmation
self.assertIn('123 subscribers', resp.data) # and supply subscriber count
self.assertIn(body, resp.data) # and the msg to be sent
The mocked helper lets you do a few more things, too, but
you get the idea: Sometimes a short helper function can take you a
long way.
Tags mocks, python, testing
no comments
no trackbacks
