I am trying to write a unittest for a function named search_ldap(), which searches an LDAP server given a particular username. Here is the function definition in utils.py (note: I'm using Python 3):
from ldap3 import Server, Connection
def search_ldap(username):
result = ()
baseDN = "o=Universiteit van Tilburg,c=NL"
searchFilter = '(uid={})'.format(username)
attributes = ['givenName', 'cn', 'employeeNumber', 'mail']
try:
server = Server('ldap.example.com', use_ssl=True)
conn = Connection(server, auto_bind=True)
conn.search(baseDN, searchFilter, attributes=attributes)
for a in attributes:
result += (conn.response[0]['attributes'][a][0], )
except Exception:
raise LDAPError('Error in LDAP query')
return result
Of course I don't want to actually connect to ldap.example.com during testing, so I've decided to use Python's mock object library to mock the Server() and Connection() classes in my unittests. Here is the test code:
from unittest import mock
from django.test import TestCase
class LdapTest(TestCase):
#mock.patch('ldap3.Server')
#mock.patch('ldap3.Connection')
def test_search_ldap(self, mockConnection, mockServer):
from .utils import search_ldap
search_ldap('username')
self.assertTrue(mockServer.called)
self.assertTrue(mockConnection.called)
This test simply asserts that the mocked Server and Connection objects are instantiated. However, thay don't, because when I run the tests with ./manage.py test I receive the following error:
Creating test database for alias 'default'...
F.
======================================================================
FAIL: test_search_ldap (uvt_user.tests.LdapTest)
----------------------------------------------------------------------
Traceback (most recent call last):
File "/usr/lib/python3.4/unittest/mock.py", line 1142, in patched
return func(*args, **keywargs)
File "/home/jj/projects/autodidact/uvt_user/tests.py", line 28, in test_search_ldap
self.assertTrue(mockServer.called)
AssertionError: False is not true
----------------------------------------------------------------------
Ran 2 tests in 0.030s
FAILED (failures=1)
Destroying test database for alias 'default'...
Why are my tests failing? How can I successfully mock ldap3's Server and Connection classes?
To mock class you should provide it's fake implementation with required methods. For example:
class FakeServer:
def call():
pass
class LdapTest(TestCase):
#mock.patch('ldap3.Server', FakeServer)
def test_search_ldap(self):
<do you checks here>
With patch() it is important that you patch objects in the namespace where they are looked up. This is explained in the Where to patch section of the documentation. There is a fundamental difference between doing
from ldap3 import Server
server = Server()
and
import ldap3
server = ldap3.Server()
In the first case (also the case in the original question), the name "Server" belongs to the current module. In the second case, the name "Server" belongs to the ldap3 module where it is defined. The following Django unittest patches the correct "Server" and "Connection" names and should work as intended:
from unittest import mock
from django.test import TestCase
class LdapTest(TestCase):
#mock.patch('appname.utils.Server')
#mock.patch('appname.utils.Connection')
def test_search_ldap(self, mockConnection, mockServer):
from .utils import search_ldap
search_ldap('username')
self.assertTrue(mockServer.called)
self.assertTrue(mockConnection.called)
Related
I am trying to mock GetDatabaseConnection but it is still running the code within it.
class GetDatabaseConnection:
resp_dict = json.loads(get_secret())
endpoint = resp_dict.get('host')
username = resp_dict.get('username')
password = resp_dict.get('password')
database_name = resp_dict.get('dbname')
port = resp_dict.get('port')
connection = pymysql.connect(host=endpoint, user=username, passwd=password, db=database_name, port=port)
cursor = connection.cursor()
Here is the test I have written to try to mock the class.
#mock.patch("lambda_function.GetDatabaseConnection")
def test_mock_simple_class(mock_class):
mock_class.return_value = "test"
But I get the following error
test_lambda_function.py::TestPreSignUp::test_mock_simple_class FAILED [100%]
test_lambda_function.py:151 (TestPreSignUp.test_mock_simple_class)
/Library/Frameworks/Python.framework/Versions/3.9/lib/python3.9/unittest/mock.py:1334: in patched
with self.decoration_helper(patched,
/Library/Frameworks/Python.framework/Versions/3.9/lib/python3.9/contextlib.py:117: in __enter__
return next(self.gen)
/Library/Frameworks/Python.framework/Versions/3.9/lib/python3.9/unittest/mock.py:1316: in decoration_helper
arg = exit_stack.enter_context(patching)
/Library/Frameworks/Python.framework/Versions/3.9/lib/python3.9/contextlib.py:429: in enter_context
result = _cm_type.__enter__(cm)
/Library/Frameworks/Python.framework/Versions/3.9/lib/python3.9/unittest/mock.py:1389: in __enter__
self.target = self.getter()
/Library/Frameworks/Python.framework/Versions/3.9/lib/python3.9/unittest/mock.py:1564: in <lambda>
getter = lambda: _importer(target)
/Library/Frameworks/Python.framework/Versions/3.9/lib/python3.9/unittest/mock.py:1236: in _importer
thing = __import__(import_path)
../lambda_function.py:42: in <module>
class GetDatabaseConnection():
../lambda_function.py:43: in GetDatabaseConnection
resp_dict = json.loads(get_secret())
You are experiencing the same problem as in this other question about importing without executing the class - python. Since you designed your class to have attributes that would execute the call to pymysql, then they would be executed right away even when the file is just imported (e.g. during mock patching) without even creating an instance of GetDatabaseConnection.
src.py
import pymysql
class GetDatabaseConnection:
connection = pymysql.connect(host="127.0.0.1", user='username', passwd="password", db='database_name', port=80)
test_src.py
from unittest import mock
# This will read your class. And this will run your <pymsql> commands even without a running patch yet.
import src # or <from src import GetDatabaseConnection>
# This will also read your class with or without the import above. And this will run your <pymsql> commands even without the patch yet.
#mock.patch('src.GetDatabaseConnection')
def test_try1():
assert True
Output
$ pytest -q -rP
================================================================================================= ERRORS ==================================================================================================
______________________________________________________________________________________ ERROR collecting test_samp.py
E pymysql.err.OperationalError: (2003, "Can't connect to MySQL server on '127.0.0.1' ([Errno 111] Connection refused)")
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Interrupted: 1 error during collection !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
1 error in 0.22s
Solution 1:
Redesign you class to not execute class-level logic by putting them inside class methods.
class GetDatabaseConnection:
def __init__(self):
self.connection = pymysql.connect(host="127.0.0.1", user='username', passwd="password", db='database_name', port=80)
from unittest import mock
#mock.patch('src.GetDatabaseConnection')
def test_try(mock_class):
assert True
$ pytest -q -rP
================================================================================================= PASSES ==================================================================================================
1 passed in 0.03s
Solution 2:
Transform your GetDatabaseConnection into an ordinary function:
def getDatabaseConnection():
return {
"connection": pymysql.connect(host="127.0.0.1", user='username', passwd="password", db='database_name', port=80),
}
from unittest import mock
#mock.patch('src.getDatabaseConnection')
def test_try(mock_func):
assert True
Output same as Solution 1
Solution 3:
Not recommendable. Don't import the file directly without a running patch. Thus, don't patch the class GetDatabaseConnection too to avoid reading from the file and executing pymysql. First, you need to patch pymysql before importing the file. This is very hard to maintain and would break if one of your source code files import the file containing the class GetDatabaseConnection.
from unittest import mock
#mock.patch('pymysql.connect')
def test_try(mock_pymsql):
import src # Or <from src import GetDatabaseConnection>
print(f"{src.GetDatabaseConnection.connection=}") # This will be the patched version
assert True
Output:
$ pytest -q -rP
================================================================================================= PASSES ==================================================================================================
________________________________________________________________________________________________ test_try _________________________________________________________________________________________________
------------------------------------------------------------------------------------------ Captured stdout call -------------------------------------------------------------------------------------------
src.GetDatabaseConnection.connection=<MagicMock name='connect()' id='139647536711760'>
1 passed in 0.03s
I'm writing integration tests for an Alexa app.
Our application uses a controller-request-response pattern. The controller receives a request with a specified intent and session variables, routes the request to functions that do some computation with the session variables, and returns a response object with the results of that computation.
We get the right behavior from UnhandledIntentTestCase as far as test_for_smoke is concerned. However, test_returning_reprompt_text
never fires, because returns_reprompt_text is never overwritten.
Can someone explain how I can overwrite it in the parent class and/or
how the correct intent name is passed to the request object in setUpClass?
intent_base_case.py
import unittest
import mycity.intents.intent_constants as intent_constants
import mycity.mycity_controller as mcc
import mycity.mycity_request_data_model as req
import mycity.test.test_constants as test_constants
###############################################################################
# TestCase parent class for all intent TestCases, which are integration tests #
# to see if any changes in codebase have broken response-request model. #
# #
# NOTE: Assumes that address has already been set. #
###############################################################################
class IntentBaseCase(unittest.TestCase):
__test__ = False
intent_to_test = None
returns_reprompt_text = False
#classmethod
def setUpClass(cls):
cls.controller = mcc.MyCityController()
cls.request = req.MyCityRequestDataModel()
key = intent_constants.CURRENT_ADDRESS_KEY
cls.request._session_attributes[key] = "46 Everdean St"
cls.request.intent_name = cls.intent_to_test
cls.response = cls.controller.on_intent(cls.request)
#classmethod
def tearDownClass(cls):
cls.controller = None
cls.request = None
def test_for_smoke(self):
self.assertNotIn("Uh oh", self.response.output_speech)
self.assertNotIn("Error", self.response.output_speech)
def test_correct_intent_card_title(self):
self.assertEqual(self.intent_to_test, self.response.card_title)
#unittest.skipIf(not returns_reprompt_text,
"{} shouldn't return a reprompt text".format(intent_to_test))
def test_returning_reprompt_text(self):
self.assertIsNotNone(self.response.reprompt_text)
#unittest.skipIf(returns_reprompt_text,
"{} should return a reprompt text".format(intent_to_test))
def test_returning_no_reprompt_text(self):
self.assertIsNone(self.response.reprompt_text)
test_unhandled_intent.py
import mycity.test.intent_base_case as base_case
########################################
# TestCase class for unhandled intents #
########################################
class UnhandledIntentTestCase(base_case.IntentBaseCase):
__test__ = True
intent_to_test = "UnhandledIntent"
returns_reprompt_text = True
output
======================================================================
FAIL: test_correct_intent_card_title (mycity.test.test_unhandled_intent.UnhandledIntentTestCase)
----------------------------------------------------------------------
Traceback (most recent call last):
File "/Users/wdrew/projects/alexa_311/my_city/mycity/mycity/test/intent_base_case.py", line 44, in test_correct_intent_card_title
self.assertEqual(self.intent_to_test, self.response.card_title)
AssertionError: 'UnhandledIntent' != 'Unhandled intent'
- UnhandledIntent
? ^
+ Unhandled intent
? ^^
======================================================================
FAIL: test_returning_no_reprompt_text (mycity.test.test_unhandled_intent.UnhandledIntentTestCase)
----------------------------------------------------------------------
Traceback (most recent call last):
File "/Users/wdrew/projects/alexa_311/my_city/mycity/mycity/test/intent_base_case.py", line 56, in test_returning_no_reprompt_text
self.assertIsNone(self.response.reprompt_text)
AssertionError: 'So, what can I help you with today?' is not None
----------------------------------------------------------------------
This is because of execution order. The SkipIf decorators are executed once during the parsing of the IntentBaseCase class. They aren't re-executed for each class or for each call to the test function.
The decorator pattern for SkipIf is designed for use with fixed global variables such as versions of dependent modules, operating system or some other external resource who's availability can be calculated or known in the global context.
Skipping tests is also something that should be done for external reasons, not for internal ones such as the needs of a sub-class. A skip is still a kind of failing test which is indicated in the report so you can see your test suite isn't exercising the whole of the functional scope of the project.
You should redesign your base class structure so functions are only available to run if the sub-class and skip using Skip for this. My recommendation would be:
class IntentBaseCase(unittest.TestCase):
...
class RepromptBaseCase(IntentBaseCase):
def test_returning_reprompt_text(self):
self.assertIsNotNone(self.response.reprompt_text)
class NoRepromptBaseCase(IntentBaseCase):
def test_returning_no_reprompt_text(self):
self.assertIsNone(self.response.reprompt_text)
You should also consider moving the response portion out of the setUp and put it into a test_ function of it's own and change these test_returning functions into a simpler assertReprompt and assertNoReprompt functions. It's a good idea to set up the tests in setUp, but not a good idea to run the actual code there.
I have the following code that I'm attempting to create a test (still work in progress):
from core.tests import BaseTestCase
from core.views import get_request
from entidades.forms import InstituicaoForm
from mock import patch
class InstituicaoFormTestCase(BaseTestCase):
def setUp(self):
super(InstituicaoFormTestCase, self).setUp()
#patch('get_request', return_value={'user': 'usuario_qualquer'})
def test_salva_instituicao_quando_informaram_convenio():
import pdb
pdb.set_trace()
form = InstituicaoForm()
it fails because when I try to create a InstituicaoForm, a get_request is called:
def get_request():
return getattr(THREAD_LOCAL, 'request', None)
and it trows this error
entidades/tests.py:11: in <module>
class InstituicaoFormTestCase(BaseTestCase):
entidades/tests.py:16: in InstituicaoFormTestCase
#patch('get_request', return_value={'user': 'usuario_qualquer'})
.tox/unit/local/lib/python2.7/site-packages/mock/mock.py:1670: in patch
getter, attribute = _get_target(target)
.tox/unit/local/lib/python2.7/site-packages/mock/mock.py:1522: in _get_target
(target,))
E TypeError: Need a valid target to patch. You supplied: 'get_request'
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> entering PDB >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
> /home/vinicius/telessaude/.tox/unit/local/lib/python2.7/site-packages/mock/mock.py(1522)_get_target()
-> (target,))
What am I doing wrong? How should mock this get_request() method?
I think the specific thing you're trying to do can be done like this:
#patch('core.views.get_request', return_value={'user': 'usuario_qualquer'})
But you should also look at the Django testing documentation, if you haven't already. You can use the testing client to fake the web request.
If you want to experiment with mock tests that don't access a database, check out Django Mock Queries. (I'm a small contributor to that project.) I've also tried mocking views, but it's fiddly.
I'm trying to test a Python method containing a with statement. The code inside the with statement can raise an RuntimeError. The test I'm talking about tests if the RuntimeError is raised.
The __enter__ and __exit__ methods are heavy (typically open and close SSH connections), and I mock them when testing.
Here is a simplified definition of the method I want to test:
# client.py
class Client():
def method_to_test():
with self:
raise RuntimeError()
For clarification purpose, I omitted the definition of __enter__ and __exit__, and removed all the code in method_to_test which was not involved in the current issue.
To test this method, I mock __enter__ and __exit__, and check if RuntimeError is raised:
# tests.py
from django.test import TestCase
import mock
from .client import Client
class ClientTestCase(TestCase):
#mock.patch('mymodule.client.Client.__enter__')
#mock.patch('mymodule.client.Client.__exit__')
def test_method_raises_Runtime(self, mock_exit, mock_enter):
mock_enter.return_value = None
client = Client()
with self.assertRaises(RuntimeError):
client.method_to_test()
This test fails with: AssertionError: RuntimeError not raised
If I do not mock __enter__, the RuntimeError is raised. Why does mocking __enter__ makes this test fail?
Seen in PEP343 https://www.python.org/dev/peps/pep-0343/
The exception is swallowed if exit() returns true
As self.__exit__ is a MagicMock in Client.method_to_test, self.__exit__ returns a MagicMock, evaluated as True. The RuntimeError is swallowed.
The fix is easy. self.__exit__() as to return None instead of a MagicMock:
# tests.py
from django.test import TestCase
import mock
from .client import Client
class ClientTestCase(TestCase):
#mock.patch('mymodule.client.Client.__enter__')
#mock.patch('mymodule.client.Client.__exit__')
def test_method_raises_Runtime(self, mock_exit, mock_enter):
# __exit__ returns None, evaluated as False
mock_exit.return_value = None
mock_enter.return_value = None
client = Client()
with self.assertRaises(RuntimeError):
client.method_to_test()
Having a weird problem with Python's unittest and PyMongo. The test randomly succeeds or fails:
import unittest
from pymongo import Connection
from tractor import Tractor
class TestTractor(unittest.TestCase):
def setUp(self):
self.tractor = Tractor(1)
self.mongo = Connection()
self.db = self.mongo.tractor
self.db.classes.remove({'name': {'$regex':'^test_'}})
self.action_class_id = self.db.classes.insert({'name': 'test_action',
'metaclass': 'action'})
self.object_class_id = self.db.classes.insert({'name': 'test_object',
'metaclass': 'object'})
def tearDown(self):
self.tractor = None
def test_create_class(self):
cid1 = self.tractor.create_action_class('test_create_action_class')
cid2 = self.tractor.create_object_class('test_create_object_class')
self.assertNotEqual(cid1, None)
self.assertNotEqual(cid2, None)
action_obj = self.db.classes.find_one({'_id': cid1})
object_obj = self.db.classes.find_one({'_id': cid2})
self.assertNotEqual(cid1, cid2)
self.assertEqual(action_obj['_id'], cid1)
self.assertEqual(object_obj['_id'], cid2)
self.assertEqual(action_obj['name'], 'test_create_action_class')
self.assertEqual(object_obj['name'], 'test_create_object_class')
Class being tested:
from pymongo import Connection
from pymongo.objectid import ObjectId
class Tractor(object):
def __init__(self, uid):
self.uid = uid
self.mongo = Connection()
self.db = self.mongo.tractor
# Classes
def create_action_class(self, name):
return self.db.classes.insert({'name': name,
'attributes': [],
'metaclass': 'action'})
def create_object_class(self, name):
return self.db.classes.insert({'name': name,
'attributes': [],
'metaclass': 'object'})
Random behavior:
silver#aregh-6930-lnx ~/projects/traction/tractor $ python -m unittest discover
......ssEssssssssss
======================================================================
ERROR: test_create_class (tests.test_tractor.TestTractor)
----------------------------------------------------------------------
Traceback (most recent call last):
File "/home/silver/projects/traction/tractor/tests/test_tractor.py", line 64, in test_create_class
self.assertEqual(action_obj['_id'], cid1)
TypeError: 'NoneType' object is not subscriptable
----------------------------------------------------------------------
Ran 19 tests in 0.023s
FAILED (errors=1, skipped=12)
...
silver#aregh-6930-lnx ~/projects/traction/tractor $ python -m unittest discover
......ss.ssssssssss
----------------------------------------------------------------------
Ran 19 tests in 0.015s
OK (skipped=12)
These two results randomly happen for the same test as I rerun the test without changing anything neither in the class nor in the test.
All of this runs on my machine and I know for sure that while running the test, nobody else tinkers neither with MongoDB nor with the code.
What gives?
I strongly suspect the problem here is that you are not using "safe" mode for your writes.
By default MongoDB uses "fire and forget" mode. This means that the insert command is sent to the server, but the driver doesn't check for any server responses.
When you switch to "safe" mode, the driver will send the insert command and it will then send a second command getLastError. This second command will return when the server has actually committed the write.
Again, by default you are running in "fire and forget" mode, so there is indeed a potential race condition here. For unit tests you will need to run with "safe" mode on.
The function signature for insert is defined here. However, you should also be able to make the change at the Connection level so that each connection to the DB uses "safe" mode by default.