Errordite exception logging.
This app provides integration between Python and Errordite, a centralised error logging and management service, not unlike Sentry, but with more functionality around the classification and management of errors.
The application is provided in the form of a standard Python logging handler. In order to log exceptions with Errordite, you simply use logging.error or logging.exception in your except block:
>>> import logging >>> import errordite >>> logger = logging.getLogger(__name__) >>> logger.addHandler(errordite.ErrorditeHandler('token')) >>> try: ... raise Exception() ... except: ... # handler uses sys.exc_info() so no need to pass ... # exception info explicitly. ... logging.error("Something went wrong") >>>
Details of the implementation are best found in the code itself - it’s fairly self-explanatory.
Important info regarding sys.exc_info() and sys.exc_clear()
This logging handler is explicitly designed to capture and publish Python exceptions. It is not a generic logger, and as such it relies on the use of sys.exc_info() to determine whether there is anything to report. This can have an unexpected effect if you are logging items as errors that have no explicit exception attached, but where a previous exception has been swallowed.
In the example below - we are catching and swallowing the DoesNotExist error because it’s a known code path:
try: .. do something that raises a known error - e.g. model.DoesNotExist except model.DoesNotExist: .. we half expected this, so just ignore it for now
However, the exception has not been cleared, and if you call sys.exc_info() further on it will still return the DoesNotExist error. A little further on in our example we are logging a business exception (trying to checkout a negative basket value) but not attaching any explicit python error:
.. continue on with the method .. some time later if basket_total < 0: logger.error("Someone tried to hack our checkout.")
In this case the wrong exception (DoesNotExist) information will be recorded.
The solution to this is to call sys.exc_clear() in the except block so that the exception is removed explicitly. The Python docs state that:
This function is only needed in only a few obscure situations.
Which suggest that this is not recommended, however, they go on to state:
These include logging and error handling systems that report information on the last or current exception.
It is the author’s opinion that this describes our exact predicament, and so the use of sys.exc_clear() is justified. (That said, you may also want to investigate why you are logging errors that aren’t exceptions if you see this situation occurring.)
The library is available at pypi as ‘errordite’, and can therefore be installed using pip:
$ pip install errordite
Once installed you can import the handler:
>>> import errordite >>> handler = errordite.ErrorditeHandler("your_errordite_token")
In order to set up a valid ErrorditeHandler you must pass in an Errordite API token, which you can get by signing up at http://www.errordite.com
There are tests in the package - they can be run using unittest:
$ python -m unittest errordite.tests
NB These tests do log real exceptions over the wire, so you will need to be connected to the web to run them. You will also need to set a local environment variable (ERRORDITE_TOKEN), which is picked up in the test suite.
If you are *nix you can pass this in on the command line:
$ ERRORDITE_TOKEN=123 python -m unittest errordite.tests
If you are on Windows you’ll need to set it up explicitly as an env var:
c:\> set ERRORDITE_TOKEN=123 c:\> python -m unittest errordite.tests
(This is a technique used to prevent having to have sensitive information in the public repo.)