aiologger 0.5.1

Asynchronous logging for python and asyncio

aiologger

About the Project

The builtin python logger is I/O blocking. This means that using the builtin logging module will interfere with your asynchronouns application performance. aiologger aims to be the standard Asynchronous non blocking logging for python and asyncio.

A word about async, Python and files

Tldr; aiologger is only fully async when logging to stdout/stderr. If you log into files on disk you are not being fully async and will be using Threads.

aiologger was created when we realized that there were no async logging libs to use. At the time, Python's built-in logging infra-structure was fully sync (still is, 3.8 beta is out). That's why we created aiologger.

Despite everything (in Linux) being a file descriptor, a Network file descriptor and the stdout/stderr FDs are treated differently from files on disk FDs. This happens because there's no stable/usable async I/O interface published by the OS to be used by Python (or any other language). That's why logging to files is NOT truly async. aiologger implementation of file logging uses aiofiles, which uses a Thread Pool to write the data. Keep this in mind when using aiologger for file logging.

Other than that, we hope aiologger helps you write fully async apps. :tada: :tada:

Installation

pip install aiologger

Testing

pipenv install --dev
pipenv run test

Implemented interfaces

aiologger implements two different interfaces that you can use to generate your logs. You can generate your logs using the async/await syntax or, if you for any reason can't (or don't want to) change all your codebase to use this syntax you can use aiologger as if it were synchronous, but behind the scenes your logs will be generated asynchronously.

Migrating from standard lib logging

Using aiologger with the standard syntax

If you prefer not to use the async/await all you need to do is to replace you logger instance with an instance of aiologger.Logger. For now on you can call logger.info() the same way you are (probably) already calling. Here is a simple example:

import asyncio
import logging

from logging import getLogger


async def main():
    logger = getLogger(__name__)
    logging.basicConfig(level=logging.DEBUG, format="%(message)s")

    logger.debug("debug")
    logger.info("info")

    logger.warning("warning")
    logger.error("error")
    logger.critical("critical")


loop = asyncio.get_event_loop()
loop.run_until_complete(main())
loop.run_forever()

Which will output the following lines:

debug
info
warning
error
critical

---

If you want to generate all your logs asynchronously, you just have to change the instance of the logger object. To do that, all we need to change those lines from:

from logging import getLogger

logger = getLogger(__name__)

to:

from aiologger import Logger

logger = Logger.with_default_handlers()

and here is the complete example, generating all log lines asynchronously.

import asyncio
from aiologger import Logger


async def main():
    logger = Logger.with_default_handlers(name='my-logger')

    logger.debug("debug")
    logger.info("info")

    logger.warning("warning")
    logger.error("error")
    logger.critical("critical")

    await logger.shutdown()


loop = asyncio.get_event_loop()
loop.run_until_complete(main())
loop.run_forever()

This code will output the following lines:

warning
debug
info
error
critical

As you might have noticed, the output order IS NOT GUARANTEED. If some kind of order is important to you, you'll need to use the await syntax. But thinking about an asyncio application, where every I/O operation is asynchronous, this shouldn't really matter.

Using aiologger with the async/await syntax

import asyncio
from aiologger import Logger


async def main():
    logger = Logger.with_default_handlers(name='my-logger')

    await logger.debug("debug at stdout")
    await logger.info("info at stdout")

    await logger.warning("warning at stderr")
    await logger.error("error at stderr")
    await logger.critical("critical at stderr")

    await logger.shutdown()

loop = asyncio.get_event_loop()
loop.run_until_complete(main())
loop.close()

The most basic use case is to log the output into stdout and stderr. Using Logger.with_default_handlers you're able to effortlessly create a new Logger instance with 2 distinct handlers:

• One for handling debug and info methods and writing to stdout;
• The other, for handling warning, critical, exception and error methods and writing to stderr.

Since everything is asynchronous, this means that for the same handler, the output order is guaranteed, but not between distinct handlers. The above code may output the following:

warning at stderr
debug at stdout
error at stderr
info at stdout
critical at stderr

You may notice that the order between the same handler is guaranteed. E.g.:

• debug at stdout was outputted before info at stdout
• warning at stderr was outputted before error at stderr
• between lines of distinct handlers, the order isn't guaranteed. warning at stderr was outputted before debug at stdout

Lazy initialization

Since the actual stream initialization only happens on the first log call, it's possible to initialize aiologger.Logger instances outside a running event loop:

import asyncio
from aiologger import Logger


logger = Logger.with_default_handlers(name='my-logger')


async def main():
    
    await logger.debug("debug at stdout")
    await logger.info("info at stdout")

    await logger.warning("warning at stderr")
    await logger.error("error at stderr")
    await logger.critical("critical at stderr")

    await logger.shutdown()

loop = asyncio.get_event_loop()
loop.run_until_complete(main())
loop.close()

Loggers

JsonLogger

A simple, featureful, drop-in replacement to the default aiologger.Logger that grants to always log valid, single line, JSON output.

It logs everything

import asyncio
from datetime import datetime

from aiologger.loggers.json import JsonLogger


async def main():
    logger = JsonLogger.with_default_handlers()
    await logger.info("Im a string")
    # {"logged_at": "2018-06-14T09:34:56.482817", "line_number": 9, "function": "main", "level": "INFO", "file_path": "/Users/diogo.mmartins/Library/Preferences/PyCharm2018.1/scratches/scratch_47.py", "msg": "Im a string"}

    await logger.info({
        'date_objects': datetime.now(),
        'exceptions': KeyError("Boooom"),
        'types': JsonLogger
    })
    # {"logged_at": "2018-06-14T09:34:56.483000", "line_number": 13, "function": "main", "level": "INFO", "file_path": "/Users/diogo.mmartins/Library/Preferences/PyCharm2018.1/scratches/scratch_47.py", "msg": {"date_objects": "2018-06-14T09:34:56.482953", "exceptions": "Exception: KeyError('Boooom',)", "types": "<JsonLogger aiologger-json (DEBUG)>"}}

    await logger.shutdown()


loop = asyncio.get_event_loop()
loop.run_until_complete(main())
loop.close()

JsonLogger Options

Callable[[], str] log values may also be used to generate dynamic content that are evaluated at serialization time. All you need to do is wrap the callable using CallableWrapper:

import asyncio
import logging
from random import randint

from aiologger.loggers.json import JsonLogger
from aiologger.utils import CallableWrapper


def rand():
    return randint(1, 100)


logger = JsonLogger.with_default_handlers(level=logging.DEBUG)


async def main():

    await logger.info(CallableWrapper(rand))
    # {"logged_at": "2018-06-14T09:37:52.624123", "line_number": 15, "function": "main", "level": "INFO", "file_path": "/Users/diogo.mmartins/Library/Preferences/PyCharm2018.1/scratches/scratch_47.py", "msg": 70}

    await logger.info({"Xablau": CallableWrapper(rand)})
    # {"logged_at": "2018-06-14T09:37:52.624305", "line_number": 18, "function": "main", "level": "INFO", "file_path": "/Users/diogo.mmartins/Library/Preferences/PyCharm2018.1/scratches/scratch_47.py", "msg": {"Xablau": 29}}

    await logger.shutdown()

loop = asyncio.get_event_loop()
loop.run_until_complete(main())
loop.close()

Adding content to root

By default, everything passed to the log methods is inserted inside the msg root attribute, but sometimes we want to add content to the root level.

Flatten

This behavior may be achieved using flatten. Which is available both as a method parameter and instance attribute.

As an instance attribute, every call to a log method would "flat" the dict attributes.

import asyncio
import logging
from aiologger.loggers.json import JsonLogger


async def main():
    logger = JsonLogger.with_default_handlers(level=logging.DEBUG, flatten=True)

    await logger.info({"status_code": 200, "response_time": 0.00534534})
    # {"status_code": 200, "response_time": 0.534534, "logged_at": "2017-08-11T16:18:58.446985", "line_number": 6, "function": "<module>", "level": "INFO", "path": "/Users/diogo/PycharmProjects/aiologger/bla.py"}

    await logger.error({"status_code": 404, "response_time": 0.00134534})
    # {"status_code": 200, "response_time": 0.534534, "logged_at": "2017-08-11T16:18:58.446986", "line_number": 6, "function": "<module>", "level": "INFO", "path": "/Users/diogo/PycharmProjects/aiologger/bla.py"}

    await logger.shutdown()

loop = asyncio.get_event_loop()
loop.run_until_complete(main())
loop.close()

As a method parameter, only the specific call would add the content to the root.

import asyncio
import logging
from aiologger.loggers.json import JsonLogger


async def main():
    logger = await JsonLogger.with_default_handlers(level=logging.DEBUG)

    await logger.info({"status_code": 200, "response_time": 0.00534534}, flatten=True)
    # {"logged_at": "2017-08-11T16:23:16.312441", "line_number": 6, "function": "<module>", "level": "INFO", "path": "/Users/diogo/PycharmProjects/aiologger/bla.py", "status_code": 200, "response_time": 0.00534534}

    await logger.error({"status_code": 404, "response_time": 0.00134534})
    # {"logged_at": "2017-08-11T16:23:16.312618", "line_number": 8, "function": "<module>", "level": "ERROR", "path": "/Users/diogo/PycharmProjects/aiologger/bla.py", "msg": {"status_code": 404, "response_time": 0.00134534}}

    await logger.shutdown()

loop = asyncio.get_event_loop()
loop.run_until_complete(main())
loop.close()

Warning: It is possible to overwrite keys that are already present at root level.

import asyncio
import logging
from aiologger.loggers.json import JsonLogger


async def main():
    logger = JsonLogger.with_default_handlers(level=logging.DEBUG)

    await logger.info({'logged_at': 'Yesterday'}, flatten=True)
    # {"logged_at": "Yesterday", "line_number": 6, "function": "<module>", "level": "INFO", "path": "/Users/diogo/PycharmProjects/aiologger/bla.py"}

    await logger.shutdown()

loop = asyncio.get_event_loop()
loop.run_until_complete(main())
loop.close()

Extra

The extra parameter allow you to add specific content to root:

import asyncio
import logging
from aiologger.loggers.json import JsonLogger


async def main():
    a = 69
    b = 666
    c = [a, b]
    logger = JsonLogger.with_default_handlers(level=logging.DEBUG)

    await logger.info("I'm a simple log")
    # {"msg": "I'm a simple log", "logged_at": "2017-08-11T12:21:05.722216", "line_number": 5, "function": "<module>", "level": "INFO", "path": "/Users/diogo/PycharmProjects/aiologger/bla.py"}

    await logger.info({"dog": "Xablau"}, extra=locals())
    # {"logged_at": "2018-06-14T09:47:29.477705", "line_number": 14, "function": "main", "level": "INFO", "file_path": "/Users/diogo.mmartins/Library/Preferences/PyCharm2018.1/scratches/scratch_47.py", "msg": {"dog": "Xablau"}, "logger": "<JsonLogger aiologger-json (DEBUG)>", "c": [69, 666], "b": 666, "a": 69}

    await logger.shutdown()


loop = asyncio.get_event_loop()
loop.run_until_complete(main())
loop.close()

It also allows you to override the default root content:

import asyncio
import logging
from aiologger.loggers.json import JsonLogger


async def main():
    logger = JsonLogger.with_default_handlers(level=logging.DEBUG)

    await logger.info("I'm a simple log")
    # {"msg": "I'm a simple log", "logged_at": "2017-08-11T12:21:05.722216", "line_number": 6, "function": "<module>", "level": "INFO", "path": "/Users/diogo/PycharmProjects/aiologger/bla.py"}

    await logger.info("I'm a simple log", extra={'logged_at': 'Yesterday'})
    # {"msg": "I'm a simple log", "logged_at": "Yesterday", "line_number": 6, "function": "<module>", "level": "INFO", "path": "/Users/diogo/PycharmProjects/aiologger/bla.py"}

    await logger.shutdown()

loop = asyncio.get_event_loop()
loop.run_until_complete(main())
loop.close()

and it may also be used as an instance attribute:

import asyncio
import logging
from aiologger.loggers.json import JsonLogger


async def main():
    logger = JsonLogger.with_default_handlers(level=logging.DEBUG, extra={'logged_at': 'Yesterday'})

    await logger.info("I'm a simple log")
    # {"msg": "I'm a simple log", "logged_at": "Yesterday", "line_number": 6, "function": "<module>", "level": "INFO", "path": "/Users/diogo/PycharmProjects/aiologger/bla.py"}

    await logger.info("I'm a simple log")
    # {"msg": "I'm a simple log", "logged_at": "Yesterday", "line_number": 6, "function": "<module>", "level": "INFO", "path": "/Users/diogo/PycharmProjects/aiologger/bla.py"}

    await logger.shutdown()

loop = asyncio.get_event_loop()
loop.run_until_complete(main())
loop.close()

Exclude default logger fields

If you think that the default fields are too much, it's also possible to exclude fields from the output message.

import asyncio
import logging
from aiologger.loggers.json import JsonLogger
from aiologger.formatters.json import FUNCTION_NAME_FIELDNAME, LOGGED_AT_FIELDNAME


async def main():
    logger = JsonLogger.with_default_handlers(
        level=logging.DEBUG,
        exclude_fields=[FUNCTION_NAME_FIELDNAME,
                        LOGGED_AT_FIELDNAME,
                        'file_path',
                        'line_number']
    )

    await logger.info("Function, file path and line number wont be printed")
    # {"level": "INFO", "msg": "Function, file path and line number wont be printed"}

    await logger.shutdown()

loop = asyncio.get_event_loop()
loop.run_until_complete(main())
loop.close()

Serializer options

serializer_kwargs is available both as instance attribute and as a log method parameter and may be used to pass keyword arguments to the serializer function. (See more: https://docs.python.org/3/library/json.html)

For pretty printing the output, you may use the indent kwarg. Ex.:

import asyncio
import logging
from aiologger.loggers.json import JsonLogger


async def main():
    logger = JsonLogger.with_default_handlers(
        level=logging.DEBUG,
        serializer_kwargs={'indent': 4}
    )

    await logger.info({
        "artist": "Black Country Communion",
        "song": "Cold"
    })

    await logger.shutdown()

loop = asyncio.get_event_loop()
loop.run_until_complete(main())
loop.close()

Would result in a pretty indented output:

{
    "logged_at": "2017-08-11T21:04:21.559070",
    "line_number": 5,
    "function": "<module>",
    "level": "INFO",
    "file_path": "/Users/diogo/Library/Preferences/PyCharm2017.1/scratches/scratch_32.py",
    "msg": {
        "artist": "Black Country Communion",
        "song": "Cold"
    }
}

The same result can be achieved making a log call with serializer_kwargs as a parameter.

await logger.warning({'artist': 'Black Country Communion', 'song': 'Cold'}, serializer_kwargs={'indent': 4})

Handlers

AsyncStreamHandler

A handler class for writing logs into a stream which may be sys.stdout or sys.stderr. If a stream isn't provided, it defaults to sys.stderr. If level is not specified, logging.NOTSET is used. If formatter is not None, it is used to format the log record before emit() gets called. A filter may be used to filter log records

import sys
from aiologger.handlers.streams import AsyncStreamHandler


handler = AsyncStreamHandler(stream=sys.stdout)

It also accepts a level, formatter and filter at the initialization.

AsyncFileHandler

Important: AsyncFileHandler depends on a optional dependency and you should install aiologger with pip install aiologger[aiofiles]

A handler class that sends logs into files. The specified file is opened and used as the stream for logging. If mode is not specified, 'a' is used. If encoding is not None, it is used to open the file with that encoding. The file opening is delayed until the first call to emit().

from aiologger.handlers.files import AsyncFileHandler
from tempfile import NamedTemporaryFile


temp_file = NamedTemporaryFile()
handler = AsyncFileHandler(filename=temp_file.name)

Options

• AIOLOGGER_HANDLE_ERROR_FALLBACK_ENABLED - An environment variable that tells aiologger whether it should emit a log to stderr in case of a handler emit raises an exceptions. To disable the default behaviour, set this environment variable to a falsy value ("False", "false", "0"). Default: True

Compatibility

Currently tested only on python 3.6 and 3.7

Depencencies

Has none.