I am using a simple unit test based test runner to test my Django application.
My application itself is configured to use a basic logger in settings.py using:
logging.basicConfig(level=logging.DEBUG)
And in my application code using:
logger = logging.getLogger(__name__)
logger.setLevel(getattr(settings, 'LOG_LEVEL', logging.DEBUG))
However, when running unittests, I'd like to disable logging so that it doesn't clutter my test result output. Is there a simple way to turn off logging in a global way, so that the application specific loggers aren't writing stuff out to the console when I run tests?
logging.disable(logging.CRITICAL)
will disable all logging calls with levels less severe than or equal to CRITICAL
. Logging can be re-enabled with
logging.disable(logging.NOTSET)
Since you are in Django, you could add these lines to your settings.py:
import sys
import logging
if len(sys.argv) > 1 and sys.argv[1] == 'test':
logging.disable(logging.CRITICAL)
That way you don't have to add that line in every setUp()
on your tests.
You could also do a couple of handy changes for your test needs this way.
There is another "nicer" or "cleaner" way to add specifics to your tests and that is making your own test runner.
Just create a class like this:
import logging
from django.test.simple import DjangoTestSuiteRunner
from django.conf import settings
class MyOwnTestRunner(DjangoTestSuiteRunner):
def run_tests(self, test_labels, extra_tests=None, **kwargs):
# Don't show logging messages while testing
logging.disable(logging.CRITICAL)
return super(MyOwnTestRunner, self).run_tests(test_labels, extra_tests, **kwargs)
And now add to your settings.py file:
TEST_RUNNER = "PATH.TO.PYFILE.MyOwnTestRunner"
#(for example, 'utils.mytest_runner.MyOwnTestRunner')
This lets you do one really handy modification that the other approach doesn't, which is to make Django just tests the applications that you want. You can do that by changing the test_labels
adding this line to the test runner:
if not test_labels:
test_labels = ['my_app1', 'my_app2', ...]
Is there a simple way to turn off logging in a global way, so that the application specific loggers aren't writing stuff out to the console when I run tests?
The other answers prevent "writing stuff out to the console" by globally setting the logging infrastructure to ignore anything. This works but I find it too blunt an approach. My approach is to perform a configuration change which does only what's needed to prevent logs to get out on the console. So I add a custom logging filter to my settings.py
:
from logging import Filter
class NotInTestingFilter(Filter):
def filter(self, record):
# Although I normally just put this class in the settings.py
# file, I have my reasons to load settings here. In many
# cases, you could skip the import and just read the setting
# from the local symbol space.
from django.conf import settings
# TESTING_MODE is some settings variable that tells my code
# whether the code is running in a testing environment or
# not. Any test runner I use will load the Django code in a
# way that makes it True.
return not settings.TESTING_MODE
And I configure the Django logging to use the filter:
LOGGING = {
'version': 1,
'disable_existing_loggers': False,
'filters': {
'testing': {
'()': NotInTestingFilter
}
},
'formatters': {
'verbose': {
'format': ('%(levelname)s %(asctime)s %(module)s '
'%(process)d %(thread)d %(message)s')
},
},
'handlers': {
'console': {
'level': 'DEBUG',
'class': 'logging.StreamHandler',
'filters': ['testing'],
'formatter': 'verbose'
},
},
'loggers': {
'foo': {
'handlers': ['console'],
'level': 'DEBUG',
'propagate': True,
},
}
}
End result: when I'm testing, nothing goes to the console, but everything else stays the same.
Why Do This?
I design code that contains logging instructions that are triggered only in specific circumstances and that should output the exact data I need for diagnosis if things go wrong. Therefore I test that they do what they are supposed to do and thus completely disabling logging is not viable for me. I don't want to find once the software is in production that what I thought would be logged is not logged.
Moreover, some test runners (Nose, for instance) will capture logs during testing and output the relevant part of the log together with a test failure. It is useful in figuring out why a test failed. If logging is completely turned off, then there's nothing that can be captured.
test_settings.py
file which sits next to my project's settings.py
. It is set to load settings.py
and make some changes like set TESTING_MODE
to True
. My test runners are organized so that test_settings
is the module loaded for the Django project settings. There are many ways this can be done. I usually go with setting the environment variable DJANGO_SETTINGS_MODULE
to proj.test_settings
.
assertLogs
was failing and it was due to logging.disable(logging.CRITICAL)
. Just turning off console output in some way is FAR better than totally breaking logging.
TESTING_MODE
(I just checked), so this need to be modified slightly. Could add the following into settings.py: TESTING_MODE = 'test' in sys.argv
.
I like Hassek's custom test runner idea. It should be noted that DjangoTestSuiteRunner
is no longer the default test runner in Django 1.6+, it has been replaced by the DiscoverRunner
. For default behaviour, the test runner should be more like:
import logging
from django.test.runner import DiscoverRunner
class NoLoggingTestRunner(DiscoverRunner):
def run_tests(self, test_labels, extra_tests=None, **kwargs):
# disable logging below CRITICAL while testing
logging.disable(logging.CRITICAL)
return super(NoLoggingTestRunner, self).run_tests(test_labels, extra_tests, **kwargs)
helpers
, which only has utils that don't import from anywhere else within the project.
I've found that for tests within unittest
or similar a framework, the most effective way to safely disable unwanted logging in unit tests is to enable/disable in the setUp
/tearDown
methods of a particular test case. This lets one target specifically where logs should be disabled. You could also do this explicitly on the logger of the class you're testing.
import unittest
import logging
class TestMyUnitTest(unittest.TestCase):
def setUp(self):
logging.disable(logging.CRITICAL)
def tearDown(self):
logging.disable(logging.NOTSET)
There is some pretty and clean method to suspend logging in tests with unittest.mock.patch
method.
foo.py:
import logging
logger = logging.getLogger(__name__)
def bar():
logger.error('There is some error output here!')
return True
tests.py:
from unittest import mock, TestCase
from foo import bar
class FooBarTestCase(TestCase):
@mock.patch('foo.logger', mock.Mock())
def test_bar(self):
self.assertTrue(bar())
And python3 -m unittest tests
will produce no logging output.
I am using a simple method decorator to disable logging only in a particular test method.
def disable_logging(f):
def wrapper(*args):
logging.disable(logging.CRITICAL)
result = f(*args)
logging.disable(logging.NOTSET)
return result
return wrapper
And then I use it as in the following example:
class ScenarioTestCase(TestCase):
@disable_logging
test_scenario(self):
pass
Sometimes you want the logs and sometimes not. I have this code in my settings.py
import sys
if '--no-logs' in sys.argv:
print('> Disabling logging levels of CRITICAL and below.')
sys.argv.remove('--no-logs')
logging.disable(logging.CRITICAL)
So if you run your test with the --no-logs
options you'll get only the critical
logs:
$ python ./manage.py tests --no-logs
> Disabling logging levels of CRITICAL and below.
It's very helpful if you want speedup the tests on your continuous integration flow.
manage.py
, not in settings.py
Disable logging for the entire module:
import logging
def setUpModule():
"""Disable logging while doing these tests."""
logging.disable()
def tearDownModule():
"""Re-enable logging after doing these tests."""
logging.disable(logging.NOTSET)
class TestFoo(unittest.TestCase):
def test_foo(self):
pass
If you're using pytest
:
Since pytest captures log messages and only displays them for failed tests, you typically don't want to disable any logging. Instead, use a separate settings.py
file for tests (e.g., test_settings.py
), and add to it:
LOGGING_CONFIG = None
This tells Django to skip configuring the logging altogether. The LOGGING
setting will be ignored and can be removed from the settings.
With this approach, you don't get any logging for passed tests, and you get all available logging for failed tests.
The tests will run using the logging that was set up by pytest
. It can be configured to your liking in the pytest
settings (e.g., tox.ini
). To include debug level log messages, use log_level = DEBUG
(or the corresponding command line argument).
In cases where I wish to temporarily suppress a specific logger, I've written a little context manager that I've found useful:
from contextlib import contextmanager
import logging
@contextmanager
def disable_logger(name):
"""Temporarily disable a specific logger."""
logger = logging.getLogger(name)
old_value = logger.disabled
logger.disabled = True
try:
yield
finally:
logger.disabled = old_value
You then use it like:
class MyTestCase(TestCase):
def test_something(self):
with disable_logger('<logger name>'):
# code that causes the logger to fire
This has the advantage that the logger is re-enabled (or set back to its prior state) once the with
completes.
You can put this in the top level directory for unit tests __init__.py
file. This will disable logging globally in the unit test suite.
# tests/unit/__init__.py
import logging
logging.disable(logging.CRITICAL)
Some of my tests contain assertions about log output, so changing the log level breaks them. Instead, I changed my Django LOGGING
settings to use a NullHandler when running tests:
if 'test' in sys.argv:
_LOG_HANDLERS = ['null']
else:
_LOG_HANDLERS = ['console']
LOGGING = {
'version': 1,
'disable_existing_loggers': False,
'formatters': {
'simple': {
'format': '%(levelname)s %(message)s'
},
},
'handlers': {
'null': {
'level': 'DEBUG',
'class': 'logging.NullHandler',
},
'console': {
'level': 'DEBUG',
'class': 'logging.StreamHandler',
'formatter': 'simple',
},
},
'loggers': {
'django': {
'handlers': _LOG_HANDLERS,
'propagate': True,
'level': 'INFO',
},
}
}
If you don't want it repeatedly turn it on/off in setUp() and tearDown() for unittest (don't see the reason for that), you could just do it once per class:
import unittest
import logging
class TestMyUnitTest(unittest.TestCase):
@classmethod
def setUpClass(cls):
logging.disable(logging.CRITICAL)
@classmethod
def tearDownClass(cls):
logging.disable(logging.NOTSET)
b.h.
for me works whell - in conftest.py:
import confing
# disable logging for tests
[logging.disable(level) for level in [logging.DEBUG,
logging.INFO,
logging.ERROR,
logging.CRITICAL]]
In my case I have a settings file settings/test.py
created specifically for testing purposes, here's what it looks like:
from .base import *
DATABASES = {
'default': {
'ENGINE': 'django.db.backends.sqlite3',
'NAME': 'test_db'
}
}
PASSWORD_HASHERS = (
'django.contrib.auth.hashers.MD5PasswordHasher',
)
LOGGING = {}
I put an environment variable DJANGO_SETTINGS_MODULE=settings.test
to /etc/environment
.
If you have different initaliser modules for test, dev and production then you can disable anything or redirect it in the initialser. I have local.py, test.py and production.py that all inherit from common.y
common.py does all the main config including this snippet :
LOGGING = {
'version': 1,
'disable_existing_loggers': False,
'formatters': {
'django.server': {
'()': 'django.utils.log.ServerFormatter',
'format': '[%(server_time)s] %(message)s',
},
'verbose': {
'format': '%(levelname)s %(asctime)s %(module)s %(process)d %(thread)d %(message)s'
},
'simple': {
'format': '%(levelname)s %(message)s'
},
},
'filters': {
'require_debug_true': {
'()': 'django.utils.log.RequireDebugTrue',
},
},
'handlers': {
'django.server': {
'level': 'INFO',
'class': 'logging.StreamHandler',
'formatter': 'django.server',
},
'console': {
'level': 'DEBUG',
'class': 'logging.StreamHandler',
'formatter': 'simple'
},
'mail_admins': {
'level': 'ERROR',
'class': 'django.utils.log.AdminEmailHandler'
}
},
'loggers': {
'django': {
'handlers': ['console'],
'level': 'INFO',
'propagate': True,
},
'celery.tasks': {
'handlers': ['console'],
'level': 'DEBUG',
'propagate': True,
},
'django.server': {
'handlers': ['django.server'],
'level': 'INFO',
'propagate': False,
},
}
Then in test.py I have this:
console_logger = Common.LOGGING.get('handlers').get('console')
console_logger['class'] = 'logging.FileHandler
console_logger['filename'] = './unitest.log
This replaces the console handler with a FileHandler and means still get logging but I do not have to touch the production code base.
In case you're using pytest, you can install the super useful pytest-mock plugin and define an auto-used, session-scoped fixture that can be triggered by an env var:
# conftest.py
import os
import pytest
@pytest.fixture(autouse=True, scope="session")
def _shut_logger(session_mocker):
if os.getenv("SHUT_LOGGER", None):
return session_mocker.patch("foo.logger")
You're likely asking the wrong question if you're wondering about this in 2021 or later
On modern versions of Django*, with the out of the box configuration, tests should not produce any logging on screen. Thus, if you're aksing this question the real answer may be "something is misconfigured". This is because (by default):
Printing of log messages to console requires debug to be True
Tests run with DEBUG=False
Thus, if the loggers you use match those you have defined in LOGGING['loggers']
and being handled by the "console"
handler, tests should not produce any logging on screen.
If you see something in tests anyway, you either
have mismatched the name of your logger with the thing you've defined in settings.LOGGING
are running tests with DEBUG=True (which requires an override)
have removed "require_debug_true" from your console handler's filters.
*Modern versions meaning: 2.1 and up, i.e. not ancient.
LOGGING
is to define it completely and that overwrites the default set up. A person would need to make sure they redefined the 'console'
handler in the same way so that it still has a filter to stop output when DEBUG=False
.
LOGGING
by doing a deepcopy of the default and only changing those parts that I want to have changed.
We use structlog, which is a bit more complicated to disable:
from structlog import DropEvent
def disable_logging_in_tests(_, __, event_dict):
if len(sys.argv) > 1 and sys.argv[1] == 'test':
raise DropEvent
return event_dict
structlog.configure(
processors=[
...
disable_logging_in_tests,
]
...
)
Success story sharing
logging.disable
(from the accepted answer) at the top oftests.py
in your application that is doing the logging.tearDown()
method:logging.disable(logging.NOTSET)
puts the logging back in place neatly.tests
module is very useful.