Log message formatting options

The raven client had a defined log handler which let you use python’s built in logging configuration mechanisms to control how log messages looked when they were passed to sentry, however the new python-sdk does not appear to have any equivalent option. I have a number of apps where I have log messages managed by structlog, where structlog hooks into python’s logging and formatts everything nicely with pretty readable colours, nicely formatted timestamps, before it gets sent to stdout, however the combination of how python-sdk hooks into logging, and the formatting of these messages means that the new sentry-sdk logging integration/handling makes an absolute mess of things producing unhelpful results in sentry.io like dozens of variations of [2m2019-01-03T07:16:00.573607Z[0m [[31m[1merror [0m] [1mcan_send_message is Fals... each with one recorded error event, instead of a single can_send_message is False with the correct count of error events.

What is the mechanism for the sentry-sdk to work with python’s logging formatters and libraries such as structlog?

[odannyc]

@untitaker Thank you!

I actually got it sort of working now. Before I was using the structlog format… Which basically has a event key which contains the message. Now I changed it so [json-python-logger](https://github.com/madzak/python-json-logger) takes care of outputting the json - which has the message in a message key. This now displays correctly in Sentry BUT doesn’t display the exception stack trace. For example:

{
  "message": "something bad happened",
  "request_id": "asdf",
  "exception": "[an exception stacktrace in here]"
}

The sentry python lib will not look at exception key in this case and format it correctly in Sentry

[Keramblock]

Have same issue