On Mar 14, 2013, at 3:05 AM, Sergey Gerasimov wrote:
I’m implementing some project based on twisted.
I would like to use some logging mechanism in my modules and see both twisted generated log records and log records from my modules and be able to filter log records by level and source (generated by twisted, or subset of my modules).
What should I do in this case? Log with python logging module in my code and send log records generated by twisted to PythonLoggingObserver? Or should I avoid using of python logging module and log only with twisted logging module? I didn’t find features like filtering logs in twisted logging.
Am I right that twisted based log observer uses blocking i/o? Example: observer = log.FileLogObserver(sys.stdout) # sys.stdio.write used in implementation
FileLogObserver will honor a 'system' kwarg that can be used to log the source of a message, so that: log.msg('Important message', system='MY_MODULE') yields: 2012-11-12 18:53:55-0700 [MY_MODULE] Important message Arbitrary message formatting can be accomplished by suppling a `format` kwarg in place of a message: log.msg(format='[%(level)s] %(msg)s', level='CATASTROPHIC', msg='Important Message', system='MY_MODULE') yields: 2012-11-12 18:53:55-0700 [MY_MODULE] [CATASTROPHIC] Important message That's a bit verbose, so creating a partial wrapper is useful: from functools import partial alert = partial(log.msg, format='[%(level)s] %(msg)s', level='CATASTROPHIC', system='MY_MODULE') alert(msg='Something terrible') alert(msg='Not so bad', level='INFO') # override the default level More generally, calls to log.msg() accept keyword arguments that can be inspected by any registered observer. You may want to write your own log observer and take advantage of this if you have specific filtering requirements.