Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

https://github.com/adamchainz/flake8-logging

A Flake8 plugin that checks for issues using the standard library logging module.
https://github.com/adamchainz/flake8-logging

Last synced: 3 months ago
JSON representation

A Flake8 plugin that checks for issues using the standard library logging module.

Lists

README

        

==============
flake8-logging
==============

.. image:: https://img.shields.io/github/actions/workflow/status/adamchainz/flake8-logging/main.yml?branch=main&style=for-the-badge
:target: https://github.com/adamchainz/flake8-logging/actions?workflow=CI

.. image:: https://img.shields.io/badge/Coverage-100%25-success?style=for-the-badge
:target: https://github.com/adamchainz/flake8-logging/actions?workflow=CI

.. image:: https://img.shields.io/pypi/v/flake8-logging.svg?style=for-the-badge
:target: https://pypi.org/project/flake8-logging/

.. image:: https://img.shields.io/badge/code%20style-black-000000.svg?style=for-the-badge
:target: https://github.com/psf/black

.. image:: https://img.shields.io/badge/pre--commit-enabled-brightgreen?logo=pre-commit&logoColor=white&style=for-the-badge
:target: https://github.com/pre-commit/pre-commit
:alt: pre-commit

A `Flake8 `_ plugin that checks for issues using the standard library logging module.

For a brief overview and background, see `the introductory blog post `__.

Requirements
============

Python 3.8 to 3.12 supported.

Installation
============

First, install with ``pip``:

.. code-block:: sh

python -m pip install flake8-logging

Second, if you define Flake8’s ``select`` setting, add the ``L`` prefix to it.
Otherwise, the plugin should be active by default.

----

**Linting a Django project?**
Check out my book `Boost Your Django DX `__ which covers Flake8 and many other code quality tools.

----

Rules
=====

LOG001 use ``logging.getLogger()`` to instantiate loggers
---------------------------------------------------------

The `Logger Objects documentation section `__ starts:

Note that Loggers should NEVER be instantiated directly, but always through the module-level function ``logging.getLogger(name)``.

Directly instantiated loggers are not added into the logger tree.
This means that they bypass all configuration and their messages are only sent to the `last resort handler `__.
This can mean their messages are incorrectly filtered, formatted, and sent only to ``stderr``.
Potentially, such messages will not be visible in your logging tooling and you won’t be alerted to issues.

Use |getLogger()|__ to correctly instantiate loggers.

.. |getLogger()| replace:: ``getLogger()``
__ https://docs.python.org/3/library/logging.html#logging.getLogger

This rule detects any module-level calls to ``Logger()``.

Failing example:

.. code-block:: python

import logging

logger = logging.Logger(__name__)

Corrected:

.. code-block:: python

import logging

logger = logging.getLogger(__name__)

LOG002 use ``__name__`` with ``getLogger()``
--------------------------------------------

The `logging documentation `__ recommends this pattern:

.. code-block:: python

logging.getLogger(__name__)

|__name__|__ is the fully qualified module name, such as ``camelot.spam``, which is the intended format for logger names.

.. |__name__| replace:: ``__name__``
__ https://docs.python.org/3/reference/import.html?#name__

This rule detects probably-mistaken usage of similar module-level dunder constants:

* |__cached__|__ - the pathname of the module’s compiled versio˜, such as ``camelot/__pycache__/spam.cpython-311.pyc``.

.. |__cached__| replace:: ``__cached__``
__ https://docs.python.org/3/reference/import.html?#cached__

* |__file__|__ - the pathname of the module, such as ``camelot/spam.py``.

.. |__file__| replace:: ``__file__``
__ https://docs.python.org/3/reference/import.html?#file__

Failing example:

.. code-block:: python

import logging

logger = logging.getLogger(__file__)

Corrected:

.. code-block:: python

import logging

logger = logging.getLogger(__name__)

LOG003 ``extra`` key ``''`` clashes with LogRecord attribute
-----------------------------------------------------------------

The |extra documentation|__ states:

.. |extra documentation| replace:: ``extra`` documentation
__ https://docs.python.org/3/library/logging.html#logging.Logger.debug

The keys in the dictionary passed in ``extra`` should not clash with the keys used by the logging system.

Such clashes crash at runtime with an error like:

.. code-block:: text

KeyError: "Attempt to overwrite 'msg' in LogRecord"

Unfortunately, this error is only raised if the message is not filtered out by level.
Tests may therefore not encounter the check, if they run with a limited logging configuration.

This rule detects such clashes by checking for keys matching the |LogRecord attributes|__.

.. |LogRecord attributes| replace:: ``LogRecord`` attributes
__ https://docs.python.org/3/library/logging.html#logrecord-attributes

Failing example:

.. code-block:: python

import logging

logger = logging.getLogger(__name__)

response = acme_api()
logger.info("ACME Response", extra={"msg": response.msg})

Corrected:

.. code-block:: python

import logging

logger = logging.getLogger(__name__)

response = acme_api()
logger.info("ACME Response", extra={"response_msg": response.msg})

LOG004 avoid ``exception()`` outside of exception handlers
----------------------------------------------------------

The |exception() documentation|__ states:

.. |exception() documentation| replace:: ``exception()`` documentation
__ https://docs.python.org/3/library/logging.html#logging.exception

This function should only be called from an exception handler.

Calling ``exception()`` outside of an exception handler attaches ``None`` exception information, leading to confusing messages:

.. code-block:: pycon

>>> logging.exception("example")
ERROR:root:example
NoneType: None

Use ``error()`` instead.
To log a caught exception, pass it in the ``exc_info`` argument.

This rule detects ``exception()`` calls outside of exception handlers.

Failing example:

.. code-block:: python

import logging

response = acme_api()
if response is None:
logging.exception("ACME failed")

Corrected:

.. code-block:: python

import logging

response = acme_api()
if response is None:
logging.error("ACME failed")

LOG005 use ``exception()`` within an exception handler
------------------------------------------------------

Within an exception handler, the |exception()|__ method is preferable over ``logger.error()``.
The ``exception()`` method captures the exception automatically, whilst ``error()`` needs it to be passed explicitly in the ``exc_info`` argument.
Both methods log with the level ``ERROR``.

.. |exception()| replace:: ``exception()``
__ https://docs.python.org/3/library/logging.html#logging.Logger.exception

This rule detects ``error()`` calls within exception handlers, excluding those with a falsy ``exc_info`` argument.

Failing example:

.. code-block:: python

try:
acme_api()
except AcmeError as exc:
logger.error("ACME API failed", exc_info=exc)

Corrected:

.. code-block:: python

try:
acme_api()
except AcmeError:
logger.exception("ACME API failed")

Or alternatively, if the exception information is truly uninformative:

.. code-block:: python

try:
acme_api()
except DuplicateError:
logger.error("ACME Duplicate Error", exc_info=False)

LOG006 redundant ``exc_info`` argument for ``exception()``
----------------------------------------------------------

The |exception()2|__ method captures the exception automatically, making a truthy ``exc_info`` argument redundant.

.. |exception()2| replace:: ``exception()``
__ https://docs.python.org/3/library/logging.html#logging.Logger.exception

This rule detects ``exception()`` calls within exception handlers with an ``exc_info`` argument that is truthy or the captured exception object.

Failing example:

.. code-block:: python

try:
acme_api()
except AcmeError:
logger.exception("ACME API failed", exc_info=True)

Corrected:

.. code-block:: python

try:
acme_api()
except AcmeError:
logger.exception("ACME API failed")

LOG007 use ``error()`` instead of ``exception()`` with ``exc_info=False``
-------------------------------------------------------------------------

The |exception()3|__ method captures the exception automatically.
Disabling this by setting ``exc_info=False`` is the same as using ``error()``, which is clearer and doesn’t need the ``exc_info`` argument.

.. |exception()3| replace:: ``exception()``
__ https://docs.python.org/3/library/logging.html#logging.Logger.exception

This rule detects ``exception()`` calls with an ``exc_info`` argument that is falsy.

Failing example:

.. code-block:: python

logger.exception("Left phalange missing", exc_info=False)

Corrected:

.. code-block:: python

logger.error("Left phalange missing")

LOG008 ``warn()`` is deprecated, use ``warning()`` instead
----------------------------------------------------------

The ``warn()`` method is a deprecated, undocumented alias for |warning()|__
``warning()`` should always be used instead.
The method was deprecated in Python 2.7, in commit `04d5bc00a2 `__, and removed in Python 3.13, in commit `dcc028d924 `__.

.. |warning()| replace:: ``warning()``
__ https://docs.python.org/3/library/logging.html#logging.Logger.warning

This rule detects calls to ``warn()``.

Failing example:

.. code-block:: python

logger.warn("Cheesy puns incoming")

Corrected:

.. code-block:: python

logger.warning("Cheesy puns incoming")

LOG009 ``WARN`` is undocumented, use ``WARNING`` instead
--------------------------------------------------------

The ``WARN`` constant is an undocumented alias for |WARNING|__.
Whilst it’s not deprecated, it’s not mentioned at all in the documentation, so the documented ``WARNING`` should always be used instead.

.. |WARNING| replace:: ``WARNING``
__ https://docs.python.org/3/library/logging.html#logging-levels

This rule detects any import or access of ``WARN``.

Failing example:

.. code-block:: python

import logging

logging.WARN

Corrected:

.. code-block:: python

import logging

logging.WARNING

LOG010 ``exception()`` does not take an exception
-------------------------------------------------

Like other logger methods, the |exception()4|__ method takes a string as its first argument.
A common misunderstanding is to pass it an exception instead.
Doing so is redundant, as ``exception()`` will already capture the exception object.
It can also lead to unclear log messages, as the logger will call ``str()`` on the exception, which doesn’t always produce a sensible message.

.. |exception()4| replace:: ``exception()``
__ https://docs.python.org/3/library/logging.html#logging.Logger.exception

This rule detects ``exception()`` calls with a first argument that is the current exception handler’s capture variable.

Failing example:

.. code-block:: python

try:
shuffle_deck()
except Exception as exc:
logger.exception(exc)

Corrected:

.. code-block:: python

try:
shuffle_deck()
except Exception:
logger.exception("Failed to shuffle deck")

LOG011 avoid pre-formatting log messages
----------------------------------------

Logger methods support string formatting for `logging variable data `__, such as:

.. code-block:: python

logger.info("Couldn’t chop %s", vegetable)

Log-aggregating tools, such as `Sentry `__ can group messages based on their unformatted message templates.
Using a pre-formatted message, such as from an f-string, prevents this from happening.
Tools have to rely on imperfect heuristics, which can lead to duplicate groups.

Additionally, the logging framework skips formatting messages that won’t be logged.
Using a pre-formatted string, such as from an f-string, has no such optimization.
This overhead can add up when you have a high volume of logs that are normally skipped.

This rule detects logger method calls with a ``msg`` argument that is one of:

* an f-string
* a call to ``str.format()``
* a string used with the modulus operator (``%``)
* a concatenation of strings with non-strings

Failing examples:

.. code-block:: python

logging.error(f"Couldn’t chop {vegetable}")

.. code-block:: python

logging.error("Couldn’t chop {}".format(vegetable))

.. code-block:: python

logging.error("Couldn’t chop %s" % (vegetable,))

.. code-block:: python

logging.error("Couldn’t chop " + vegetable)

Corrected:

.. code-block:: python

logging.error("Couldn’t chop %s", vegetable)

LOG012 formatting error: ```` ```` placeholders but ``<m>`` arguments
-------------------------------------------------------------------------------

Logger methods support several string formatting options for messages.
If there’s a mismatch between the number of parameters in the message and those provided, the call will error:

.. code-block:: pycon

>>> logging.info("Sent %s to %s", letter)
--- Logging error ---
Traceback (most recent call last):
File "/.../logging/__init__.py", line 1110, in emit
msg = self.format(record)
^^^^^^^^^^^^^^^^^^^
...

File "/.../logging/__init__.py", line 377, in getMessage
msg = msg % self.args
~~~~^~~~~~~~~~~
TypeError: not enough arguments for format string
Call stack:
File "<stdin>", line 1, in <module>
Message: ' %s to %s'
Arguments: ('Red Letter',)

This will only happen when the logger is enabled since loggers don’t perform string formatting when disabled.
Thus a configuration change can reveal such errors.

Additionally, if no arguments are provided, parametrized messages are silently unformatted:

.. code-block:: pycon

>>> logging.info("Sent %s to %s")
INFO:root:Sent %s to %s

This rule detects mismatches between the number of message parameters and those provided.
At the moment, it only supports ``%``-style formatting with at least one parameter.

Failing examples:

.. code-block:: python

logging.info("Blending %s")

.. code-block:: python

logging.info("Blending %s", fruit.name, fruit.size)

Corrected:

.. code-block:: python

logging.info("Blending %s of size %r", fruit.name, fruit.size)

LOG013 formatting error: ``<missing/unreferenced>`` keys: ``<keys>``
--------------------------------------------------------------------

When using named ``%``-style formatting, if the message references a missing key, the call will error:

.. code-block:: pycon

>>> logging.error("Hi %(name)s", {"nam": "hacker"})
--- Logging error ---
Traceback (most recent call last):
File "/.../logging/__init__.py", line 1160, in emit
msg = self.format(record)
^^^^^^^^^^^^^^^^^^^
...

File "/.../logging/__init__.py", line 392, in getMessage
msg = msg % self.args
~~~~^~~~~~~~~~~
KeyError: 'name'
Call stack:
File "<stdin>", line 1, in <module>
Message: 'Hi %(name)s'
Arguments: {'nam': 'hacker'}

This will only happen when the logger is enabled since loggers don’t perform string formatting when disabled.
Thus a configuration change can reveal such errors.

This rule detects mismatches between the message parameter names and those provided.
It only works if there’s a dict literal argument for the parameters.

Failing example:

.. code-block:: python

logging.info("Blending %(fruit)s", {"froot": froot})

.. code-block:: python

logging.info("Blending %(fruit)s", {"fruit": fruit, "colour": "yellow"})

Corrected:

.. code-block:: python

logging.info("Blending %(fruit)s", {"fruit": fruit})

LOG014 avoid ``exc_info=True`` outside of exception handlers
------------------------------------------------------------

Using ``exc_info=True`` outside of an exception handler attaches ``None`` as the exception information, leading to confusing messages:

.. code-block:: pycon

>>> logging.warning("Uh oh", exc_info=True)
WARNING:root:Uh oh
NoneType: None

This rule detects logging calls with ``exc_info=True`` outside of exception handlers.

Failing example:

.. code-block:: python

import logging

logging.warning("Uh oh", exc_info=True)

Corrected:

.. code-block:: python

import logging

logging.warning("Uh oh")