This is a pre-production deployment of Warehouse. Changes made here affect the production instance of PyPI (
Help us improve Python packaging - Donate today!

A more flexible warning module.

Project Description


Better warnings.

See the full documentation.

The Python standard warning module is extremely good, and I believe underutilized; Though it missed a few functionality; in particular it allows filtering only on the code that triggered/called a deprecated functions, but have no ability to filter depending on the module that emitted the warning.

This is an attempt to fix that.

Too long didn’t read:

Explicit is better than implicit:

from warn import patch

# use the warning module as usual

Though the warnings.filterwarning function has now gained the emodule keyword parameter to filer by the module that emitted the warning; example:

import warnings
warnings.filter('default', category=DeprecationWarnings, emodule='matplotlib\.pyplot.*')

All warnings from matplotlib.pyplot and its submodule will now be show by default, regardless of whether you trigger them directly, via pandas, seaborn, your own code…

Warning emitter, warning caller.

Python warnings are a beautiful relative simple piece of code which is extremely powerful in the right hands once you learned how to use it.

It allows you to determine a posteriori whether you want a particular piece of code to trigger an exception, display a message to the user or simply do nothing.

It is difficult to show the full power of the waring with a simple piece of code, but in large code base, and once you start having several layer of dependency a parsimonious usage of warning , and in particular DeprecationWarning can make a large difference.

Caller, vs Emitter

Let’s clear up some vocabulary first, to differentiate the warning “Caller” from the warning “Emitter”

# file

def public_api(param1, deprecated_parameter=None)::

    if deprecated_parameter:
        return _deprecated_function(param1, deprecated_parameter):
        return normal_buisnell_logic(param1)

def _deprecated_function(param1, deprecated_parameter):
    import warningsA
    # warning emitted here
    warnings.warn('using `deprecated_parameter` is deprecated ',
# file
from emitter import public_api
public_api(1, True)  # warning triggered here.

you can now do something like

from warn import path

import warnings
warnings.filter('default', category=DeprecationWarnings, emodule='emitter.*')
import emitter # will log the warning !

Change this to “error” in your test-suite, and filter by all your dependencies !

The Python built-in module allow you to filter warning by caller (assuming the emitter have set the stacklevel options right, which is not always obvious to do). This is extremely useful when you are developing the caller; but not that much when you are developing the emitter.

It is common for a caller to actually have many underlying library the can trigger warnings, or for a developer to only care about a subset of the emitter warnings.

Many libraries are going around this limitation by sub-classing Warnings; two example are Matplolib and sympy in order to selectively enable them. Still this only give a coarse way of filtering warnings, and it required to know where the warnings are defined in order to import and filter for them.

Because of Python default choice to filter out deprecation warnings, this also forced either inherit UserWarning (choice of matplotlib), which removes the semantic meaning of DeprecationWarning offered by Python or to inject a custom filter in the warnings filter module on import (choice of Sympy), which can lead to surprising behavior.

Availability on Python 2

I don’t know if if works on Python 2; I don’t really have the time to investigate; I don’t particularly care a lot; but feel free to send a PR that ads support if necessary I would be happy to merge it.


This does not work on packages that either :

  • Got and keep a reference on warnings.warn before patch() have been called; that is to say things of the form: from warnings iport warn
  • Cannot work on C-extensions (aka won’t filter on numpy) ; Both of the above are technically possible with Assembly Patching which I’m not confortable with.

The Ugly

As Warnings filters have to be 5 tuples with specific types this works by shoving dummy instances in the filters list and using this as keys for a proxy to lookup real filter keys. So worse case scenario the filter you insert with this module will just be no-op. But you will incur a performance penalty if you use this, especially if your codebase triggers a lot of warnings.

Get the upstream

I’d love feedback and have a nicer API to deal with warnings at CPython level in order to provide custom filters, and custom filters functions.


Good Deprecation Warnings.

A good warning and in particular Deprecation warning is extremely helpful and can make the difference for the adoption of an API. Take the following fiction example:

>>> import warnings

>>> warnings.simplefilter('default')

>>> from quezetraste import frobulate, constribule

>>> frobulate('HI', 3)
DeprecationWarning: The 'frobulate' function is deprecated.

>>> contribule('Hi', 3)

DeprecationWarning: The 'constribule(message, recipient_id)' function of the
                    'quezetraste' package is deprecated since version 7.3. It
                    haz been replaced by 'Recipient(id).send(message)' which
                    was available since 7.2. See

Turn the DeprecationWarnings into error in your test-suite!

At least make them visible; at best once you fixed a deprecation warning turn this specific one into an error to not reproduce it.

Release History

Release History

This version
History Node


History Node


Download Files

Download Files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

File Name & Checksum SHA256 Checksum Help Version File Type Upload Date
warn-0.1.0-py2.py3-none-any.whl (23.6 kB) Copy SHA256 Checksum SHA256 py2.py3 Wheel Sep 4, 2016

Supported By

WebFaction WebFaction Technical Writing Elastic Elastic Search Pingdom Pingdom Monitoring Dyn Dyn DNS Sentry Sentry Error Logging CloudAMQP CloudAMQP RabbitMQ Heroku Heroku PaaS Kabu Creative Kabu Creative UX & Design Fastly Fastly CDN DigiCert DigiCert EV Certificate Rackspace Rackspace Cloud Servers DreamHost DreamHost Log Hosting