Everett

Everett is a Python configuration library for your app.

Code:https://github.com/willkg/everett
Issues:https://github.com/willkg/everett/issues
License:MPL v2
Documentation:https://everett.readthedocs.io/

Goals

Goals of Everett:

  1. flexible configuration from multiple configured environments
  2. easy testing with configuration
  3. easy documentation of configuration for users

From that, Everett has the following features:

  • is composeable and flexible
  • makes it easier to provide helpful error messages for users trying to configure your software
  • supports auto-documentation of configuration with a Sphinx autocomponent directive
  • has an API for testing configuration variations in your tests
  • can pull configuration from a variety of specified sources (environment, INI files, YAML files, dict, write-your-own)
  • supports parsing values (bool, int, lists of things, classes, write-your-own)
  • supports key namespaces
  • supports component architectures
  • works with whatever you’re writing–command line tools, web sites, system daemons, etc

Everett is inspired by python-decouple and configman.

Quick start

Fast start example

You have an app and want it to look for configuration first in an .env file in the current working directory, then then in the process environment. You can do this:

from everett.manager import ConfigManager

config = ConfigManager.basic_config()

Then you can use it like this:

debug_mode = config('debug', default='False', parser=bool)

When you outgrow that or need different variations of it, you can change that to creating a ConfigManager from scratch.

More complex example

We have an app and want to pull configuration from an INI file stored in a place specified by MYAPP_INI in the environment, ~/.myapp.ini, or /etc/myapp.ini in that order.

We want to pull infrastructure values from the environment.

Values from the environment should override values from the INI file.

First, we need to install the additional requirements for INI file environments:

pip install everett[ini]

Then we set up our ConfigManager:

import os
import sys

from everett.ext.inifile import ConfigIniEnv
from everett.manager import ConfigManager, ConfigOSEnv


def get_config():
    return ConfigManager(
        # Specify one or more configuration environments in
        # the order they should be checked
        environments=[
            # Look in OS process environment first
            ConfigOSEnv(),

            # Look in INI files in order specified
            ConfigIniEnv([
                os.environ.get('MYAPP_INI'),
                '~/.myapp.ini',
                '/etc/myapp.ini'
            ]),
        ],

        # Provide users a link to documentation for when they hit
        # configuration errors
        doc='Check https://example.com/configuration for docs.'
    )

Then we use it:

def is_debug(config):
    return config('debug', default='False', parser=bool,
        doc='Switch debug mode on and off.')


config = get_config()

if is_debug(config):
    print('DEBUG MODE ON!')

Let’s write some tests that verify behavior based on the debug configuration value:

from myapp import get_config, is_debug

from everett.manager import config_override


@config_override(DEBUG='true')
def test_debug_true():
    assert is_debug(get_config()) is True


@config_override(DEBUG='false')
def test_debug_false():
    assert is_debug(get_config()) is False

If the user sets DEBUG with a bad value, they get a helpful error message with the documentation for the configuration option and the ConfigManager:

$ DEBUG=foo python myprogram.py
<traceback>
namespace=None key=debug requires a value parseable by bool
Switch debug mode on and off.
Check https://example.com/configuration for docs.

Configuration classes

Everett supports centralizing your configuration in a class. Instead of having configuration-related bits defined across your codebase, you can define it in a class. Let’s rewrite the above example using a configuration class.

First, create a configuration class:

import os
import sys

from everett.component import RequiredConfigMixin, ConfigOptions
from everett.ext.inifile import ConfigIniEnv
from everett.manager import ConfigManager, ConfigOSEnv


class AppConfig(RequiredConfigMixin):
    required_config = ConfigOptions()
    required_config.add_option(
        'debug',
        parser=bool,
        default='false',
        doc='Switch debug mode on and off.')
    )

Then we set up our ConfigManager:

def get_config():
    manager = ConfigManager(
        # Specify one or more configuration environments in
        # the order they should be checked
        environments=[
            # Look in OS process environment first
            ConfigOSEnv(),

            # Look in INI files in order specified
            ConfigIniEnv([
                os.environ.get('MYAPP_INI'),
                '~/.myapp.ini',
                '/etc/myapp.ini'
            ]),
        ],

        # Provide users a link to documentation for when they hit
        # configuration errors
        doc='Check https://example.com/configuration for docs.'
    )

    # Apply the configuration class to the configuration manager
    # so that it handles option properties like defaults, parsers,
    # documentation, and so on.
    return manager.with_options(AppConfig())

Then use it:

config = get_config()

if config('debug'):
    print('DEBUG MODE ON!')

Further, you can auto-generate configuration documentation by including the everett.sphinxext Sphinx extension and using the autocomponent directive:

.. autocomponent:: path.to.AppConfig

That has some niceties:

  1. your application configuration is centralized in one place instead of spread out across your code base
  2. you can use the autocomponent Sphinx directive to auto-generate configuration documentation for your users

Everett components

Everett supports components that require configuration. Say your app needs to connect to RabbitMQ. With Everett, you can define the component’s configuration needs in the component class:

from everett.component import RequiredConfigMixin, ConfigOptions


class RabbitMQComponent(RequiredConfigMixin):
    required_config = ConfigOptions()
    required_config.add_option(
        'host',
        doc='RabbitMQ host to connect to'
    )
    required_config.add_option(
        'port',
        default='5672',
        doc='Port to use',
        parser=int
    )
    required_config.add_option(
        'queue_name',
        doc='Queue to insert things into'
    )

    def __init__(self, config):
        # Bind the configuration to just the configuration this
        # component requires such that this component is
        # self-contained
        self.config = config.with_options(self)

        self.host = self.config('host')
        self.port = self.config('port')
        self.queue_name = self.config('queue_name')

Then instantiate a RabbitMQComponent that looks for configuration keys in the rmq namespace:

queue = RabbitMQComponent(config.with_namespace('rmq'))

The RabbitMQComponent has a HOST key, so your configuration would need to define RMQ_HOST.

You can auto-generate configuration documentation for this component in your Sphinx docs by including the everett.sphinxext Sphinx extension and using the autocomponent directive:

.. autocomponent:: path.to.RabbitMQComponent
   :namespace: rmq

Say your app actually needs to connect to two separate queues–one for regular processing and one for priority processing:

from everett.manager import ConfigManager

config = ConfigManager.basic_config()

# Apply the "rmq" namespace to the configuration so all keys are
# prepended with RMQ_
rmq_config = config.with_namespace('rmq')

# Create a RabbitMQComponent with RMQ_REGULAR_ prepended to keys
regular_queue = RabbitMQComponent(rmq_config.with_namespace('regular'))

# Create a RabbitMQComponent with RMQ_PRIORITY_ prepended to keys
priority_queue = RabbitMQComponent(rmq_config.with_namespace('priority'))

In your environment, you provide the regular queue configuration with RMQ_REGULAR_HOST, etc and the priority queue configuration with RMQ_PRIORITY_HOST, etc.

Same component code. Two different instances pulling configuration from two different namespaces.

Components support subclassing, mixins and all that, too.

Install

Install from PyPI

Run:

$ pip install everett

If you want to use the ConfigIniEnv, you need to install its requirements as well:

$ pip install everett[ini]

If you want to use the ConfigYamlEnv, you need to install its requirements as well:

$ pip install everett[yaml]

Install for hacking

Run:

# Clone the repository
$ git clone https://github.com/willkg/everett

# Create a virtualenvironment
$ mkvirtualenv --python /usr/bin/python3 everett
...

# Install Everett and dev requirements
$ pip install -r requirements-dev.txt

Why not other libs?

Most other libraries I looked at had one or more of the following issues:

  • were tied to a specific web app framework
  • didn’t allow you to specify configuration sources
  • provided poor error messages when users configure things wrong
  • had a global configuration object
  • made it really hard to override specific configuration when writing tests
  • had no facilities for auto-generating configuration documentation

Indices and tables