Question

In many cases unit-tests are significantly slowed down by the use of python’s logging package. Assuming logging isn’t essential to the test, how would you cleanly override logging per-test, so that log commands would be effectively skipped.

Assume the use of multiple loggers such as in:

logger1 = logging.getLogger('logger1')
logger2 = logging.getLogger('logger2')

Answer 1

Option 1:

Logging can be disabled by calling

logging.disable(logging.CRITICAL)

and turned back on with

logging.disable(logging.NOTSET)

However, even after disabling logging, a logging statement such as logger.info would still cause Python to do a few attribute lookups and function calls before reaching the isEnabledFor method. Still, this might be good enough.

Option 2:

Use mocking:

class MockLogger(object):
    def debug(msg, *args, **kwargs): pass
    def info(msg, *args, **kwargs): pass
    def warn(msg, *args, **kwargs): pass
    def error(msg, *args, **kwargs): pass
    def critical(msg, *args, **kwargs): pass

class Test(unittest.TestCase):
    def test_func(self):
        _logger1=testmodule.logger1
        _logger2=testmodule.logger2
        testmodule.logger1=MockLogger()
        testmodule.logger2=MockLogger()
        # perform test
        testmodule.logger1=_logger1
        testmodule.logger2=_logger2

This will reduce the time consumed by logging statements to the time it takes to do one attribute lookup and one (noop) function call. If that’s not satisfactory, I think the only option left is removing the logging statements themselves.