Django logging - custom filter works for handler, but does not work for logger - python-logging

Question

I have a custom filter like this:
proj/proj/log_utils/filters.py:
import logging
class OnlyThisSeverity(logging.Filter):
def filter(self, record):
print('levelname: {}'.format(record.levelname))
return record.levelno == logging.INFO
And Django settings look like this:
LOGGING = {
'version': 1,
'disable_existing_loggers': False,
'handlers': {
'console': {
'class': 'logging.StreamHandler',
},
},
'filters': {
'only_this_severity': {
'()': 'android_blend.log_utils.filters.OnlyThisSeverity',
}
},
'loggers': {
'my_app': {
'handlers': ('console',), # 'console'
'level': 'INFO',
'propagate': False,
'filters': ['only_this_severity']
},
}
}
Inside my_app/test.py, I have a function like this:
import logging
logger = logging.getLogger(__name__)
def handle(self, *args, **kwargs):
logger.info('Reporting info......')
logger.error('Reporting error......')
If my filter works correctly, I should see only Reporting info in the console, whereas I see both lines.
So the problem is that only_this_severity filter is ignored.
However, if I move the filters inside the console handler definition, then I achieve my goal. What I am doing wrong ? I want the filter to work
for the logger my_app, and not only on handler level.
My ultimate goal is to change the custom behavour of logging lib. That is, for handlers of level 'X', I want them to respect only records of level 'X', not 'X and higher'.

Answer 1

Your filter seems to work correctly even with the filter added at logger level.
This is my current output:
levelname: INFO
Reporting info......
levelname: ERROR
This is showing how the code is reaching the filter in both log operations, but filtering the error record.
I think that the problem might be in the logger hierarchy. When you use your test.py as an imported module, you're not getting the 'my_app' logger, but the 'my_app.test' which does not exist.
When this happens, the core retrieves the root logger and 'propagates' the default configuration to a new logger called 'my_app.test' including it on the log hierarchy. This logger doesn't have your filter engaged.
Try forcing to get 'my_app' logger by passing its name:
import logging
logger = logging.getLogger('my_app')
def handle(self, *args, **kwargs):
logger.info('Reporting info......')
logger.error('Reporting error......')
Or you can use the root logger to define your filters and then let the application create the hierarchy for you:
LOGGING = {
'version': 1,
'disable_existing_loggers': False,
'handlers': {
'console': {
'class': 'logging.StreamHandler',
},
},
'filters': {
'only_this_severity': {
'()': 'android_blend.log_utils.filters.OnlyThisSeverity',
}
},
'loggers': {
'root': {
'handlers': ('console',), # 'console'
'level': 'INFO',
'filters': ['only_this_severity']
},
}
}
I personally like this method because allows to safely call
logger = logging.getLogger(__name__)
in your modules in order to show (with the proper formatter) the module from which you are logging.
From https://docs.python-guide.org/writing/logging/
Best practice when instantiating loggers in a library is to only
create them using the __name__ global variable: the logging module
creates a hierarchy of loggers using dot notation, so using __name__
ensures no name collisions.