Note: This is a question I got from Adam Easterling via Support.

One thing that I frequently find myself wanting is a stack trace that includes local variables at each stack frame. In Python, all that information is there -- however, it looks like your API doesn't have a place for this information.

I asked Codebase for your API documentation, and they directed me to version 2.3, https://help.airbrake.io/kb/api-2/notifier-api-version-23

*/notice/error/backtrace/line*

Required. This element can occur more than once. Each line element 
describes one code location or frame in the backtrace when the error 
occurred, and requires @file and @number attributes. If the location 
includes a method or function, the @method attribute should be used. 

Notice, there's not really a place to include local variables at each backtrace line.

@vmihailenco Could we add this into V3? https://help.airbrake.io/kb/api-2/notifier-api-v3

Among other things, this change adds a circle.yml file so we can start testing airbrake-python using CircleCI which is free for open source projects like this one.

The python versions I have configured for testing are:

• 2.7.9
• 2.7.10
• 3.3.3
• 3.4.3
• 3.5.0

I have a passing CircleCI build on my fork against the code in this pull request: https://circleci.com/gh/samstav/airbrake-python/12

Easier to create fake tracebacks for custom logging. If you want to create a traceback that has a custom / modified trace, airbrake will not treat it accordingly since it does a type check against types.TracebackType. This is overly cautionary since the object is just handed off to methods in the traceback module which work appropriately on duck-typed object

Inspired by this stack overflow question: http://stackoverflow.com/questions/13210436/get-full-traceback

When running from a non-git directory the process will print fatal: Not a git repository (or any of the parent directories): .git to stderr once every minute or so. In our case we use a carefully packaged Docker container which excludes the git directory to keep the sizes smaller. This pull request silences that warning and maintains the same functionality (returning None)

A test was also added to reproduce the issue while troubleshooting, though it was unable to capture the existing subprocess stderr so could not assert there will not be a regression. The new test will, however, display the fatal: Not a git repository (or any of the parent directories): .git warning when running test tests if there is a regression in the future.

Hi! It's me again.

I plugged the Airbrake handler to the logging of an existing django project, and noticed that airbrake notifications were not being sent after triggering errors in the webapp. The reason was that Django, when logging errors, sends along information in the extra argument that is not any of the basic JSON types (basically every object except string, int, list, dict, etc).

This was causing a TypeError('X is not JSON serializable') in notifier.py when executing json.dumps(payload)

The purpose of this PR, is to use repr(object) for any param that is not JSON serializable.

Notice that this scenario might be pretty common also outside the context of Django.

Tests and linters green: https://circleci.com/gh/argos83/airbrake-python/11

Some long needed updates. Improves usability and leverages more of the datapoints available through the airbrake error reporting API. Includes a helper for fetching a logger object, pre-configured with an AirbrakeHandler, and many other improvements.

Updated the README to reflect the simplified interface.

I set up Errbit in AWS and made the endpoint private for testing, then access the endpoint from EKS. I know we should always verify TLS, but as for especially testing, we should have an option to skip verification. Could you consider merging this feature?

How would i add airbrake code with logger i already added this

import airbrake
logger = airbrake.getLogger(api_key=myapikeyishiddenforprivacy:), project_id=234209)
now i at the end of my game starter i have this
autoRun = ConfigVariableBool('toontown-auto-run', 1)
if autoRun:
    try:
        base.run()
    except SystemExit:
        raise
    except:
        print describeException()
        raise

now how would i add logger.exception properly after describeexception so i report crashes to airbrake every time a person who plays my game crashes :)

Hi thanks for good tool and battery!

Asyncio/aiohttp stack is rapidly growing in usage and it will be really nice if library add support for async transport.

If you consider it as an addition i can create PR.

Hi

I'm rather new to logging with Python (and Pyramid in particular). How do I configure the logging facility to use Airbrake?

According to 16.8. logging.config of the Python 3 library a config file should contain this:

...
[handlers]
keys = ..., airbrake

[logger_myapp]
level = <something>
handlers = airbrake
...

[handler_airbrake]
class = 
args = 
level = 

Now, what value should class, args and level take? I could think of these but I have no clue if they'd even make sense syntactically:

[handler_airbrake]
# Because it implementes the Python handler interface?
class = airbrake.handler.AirbrakeHandler
# Don't know, just guessing
args = {'airbrake':None, 'level':NOTSET, 'project_id': 'myapp', 'api_key': myairbrakekey }
# Delegate decision over log level to logger. For convenience only
level = NOTSET

Before, the commit hash looks like "b'8cfe34106cc8ade9da14c932523b9c5e62c0c295'" Now it looks like '8cfe34106cc8ade9da14c932523b9c5e62c0c295'

In the following example flask app, we have a report that using capture() will not report the appropriate handlers that raised the error (when running on uwsgi). Using notify() solved this problem.

from flask import Flask, jsonify
from airbrake.notifier import Airbrake

app = Flask(__name__)
ab = Airbrake(project_id=2, api_key='5a00a5340e6cac4def2262adb4ded8a4', host="http://getexceptional.me.ab")

@app.errorhandler(Exception)
def handle_error(e):
    ab.notify(e)
    return jsonify(error=str(e)), 500

@app.route('/hello')
def hello_world():
    raise Exception("Hello world failed")
    return 'Hello, World!'

@app.route('/ping')
def ping():
    raise Exception("ping failed")
    return 'Ping!'

if __name__ == '__main__':
    app.run()

From https://github.com/airbrake/airbrake-python/pull/76#discussion_r113081044

I think what I'm really trying to say is, if I did something silly/wrong like pass a string to Notice as the exception argument, then I would expect it to either 1) fail or 2) end up as the error message (or _somewhere) in airbrake.

When passed a string, I think the initialiser should use that in the error message. More generally, we could probably clean up __init__ in notice.py so that it's more clear what it's doing.

Right now we don't pick up stacktraces from bare exceptions. I think we should make an effort to do this, even though there is no guarantee that sys.exc_info() will pick up anything. In python 3.5+ we have exception.__traceback__, so we should also use this if it's available. This will also simplify some of the work in integrations.

As mentioned in https://github.com/airbrake/airbrake-python/pull/67#discussion_r105037391. We have this in the ruby notifier, and we should add it here as well.