getconf
.. image:: https://secure.travis-ci.org/Polyconseil/getconf.png?branch=master
:target: http://travis-ci.org/Polyconseil/getconf/
.. image:: https://img.shields.io/pypi/v/getconf.svg
:target: https://getconf.readthedocs.io/en/latest/changelog.html
:alt: Latest Version
.. image:: https://img.shields.io/pypi/pyversions/getconf.svg
:target: https://pypi.python.org/pypi/getconf/
:alt: Supported Python versions
.. image:: https://img.shields.io/pypi/wheel/getconf.svg
:target: https://pypi.python.org/pypi/getconf/
:alt: Wheel status
.. image:: https://img.shields.io/pypi/l/getconf.svg
:target: https://pypi.python.org/pypi/getconf/
:alt: License
The getconf
project provides simple configuration helpers for Python programs.
It provides a simple API to read from various configuration files and environment variables:
.. code-block:: python
import getconf
config = getconf.ConfigGetter('myproj', ['/etc/myproj.conf'])
db_host = config.getstr('db.host', 'localhost')
db_port = config.getint('db.port', 5432)
Beyond this API, getconf aims at unifying configuration setup across development and production systems,
respecting the standard procedures in each system:
- Allow userspace configuration on development systems
- Allow multiple different configurations for continuous integration systems
- Use standard configuration space in
/etc
on traditional production servers - Handle environment-based configuration for cloud-based platforms
getconf
is distributed under the two-clause BSD license, a copy of
which is in the source.
getconf
v1.11 onwards supports Python 3.5, 3.6, 3.7, 3.8, 3.9 and 3.10.
v1.11.x are the last versions to support Python 3.5 & 3.6.
v1.9.x are the last versions to support Python 2.7 and 3.4.
v1.8.x are the last versions to support Python 3.3.
v1.5.x are the last versions to support Python 2.6.
Links
Installation
Install the package from PyPI
_, using pip:
.. code-block:: sh
pip install getconf
Or from GitHub:
.. code-block:: sh
git clone git://github.com/Polyconseil/getconf
getconf
has no external dependency beyond Python.
Introduction
.. note:: Please refer to the full doc for reference and
advanced usage.
All configuration values are accessed through the getconf.ConfigGetter
object:
.. code-block:: python
import getconf
config = getconf.ConfigGetter('myproj', ['/etc/myproj/settings.ini', './local_settings.ini'])
The above line declares:
- Use the
myproj
namespace (explained later; this is mostly used for environment-based configuration, as a prefix for environment variables) - Look, in turn, at
/etc/myproj/settings.ini
(for production) and ./local_settings.ini
(for development); the latter overriding the former.
Once the getconf.ConfigGetter
has been configured, it can be used to retrieve settings:
.. code-block:: python
debug = config.getbool('debug', False)
db_host = config.getstr('db.host', 'localhost')
db_port = config.getint('db.port', 5432)
allowed_hosts = config.getlist('django.allowed_hosts', ['*'])
All settings have a type (default is text), and accept a default value.
They use namespaces (think 'sections') for easier reading.
With the above setup, getconf
will try to provide db.host
by inspecting
the following options in order (it stops at the first defined value):
- From the environment variable
MYPROJ_DB_HOST
, if defined - From the
host
key in the [db]
section of ./local_settings.ini
- From the
host
key in the [db]
section of /etc/myproj/settings.ini
- From the default provided value,
'localhost'
Features
Env-based configuration files
An extra configuration file/directory/glob can be provided through MYPROJ_CONFIG
;
it takes precedence over other files
Default options
An extra dictionary can be provided as ConfigGetter(defaults=some_dict)
;
it is used after configuration files and environment variables.
It should be a dict mapping a section name to a dict of ``key => value``:
.. code-block:: pycon
>>> config = ConfigGetter('myproj', defaults={'db': {'host': 'localhost'}})
>>> config.getstr('db.host')
'localhost'
Typed getters
getconf
can convert options into a few standard types:
.. code-block:: python
config.getbool('db.enabled', False)
config.getint('db.port', 5432)
config.getlist('db.tables') # Expects a comma-separated list
config.getfloat('db.auto_vacuum_scale_factor', 0.2)
config.gettimedelta('account_activation.validity', '2d')
config.getpath('django.static_root', pathlib.Path(BASE_DIR / 'static'))
``getconf`` can also convert options to user-defined standard-type-based types:
.. code-block:: python
class Environment(str, enum.Enum):
DEV = 'dev'
PROD = 'prod'
config.getenum('environment', Environment.PROD)
Concepts
getconf
relies on a few key concepts:
namespace
Each ConfigGetter
works within a specific namespace (its first argument).
Its goal is to avoid mistakes while reading the environment:
with ``ConfigGetter(namespace='myproj')``, only environment variables
beginning with ``MYPROJ_`` will be read.
It is, however, possible to disable namespacing by using
``ConfigGetter(namespace=getconf.NO_NAMESPACE)``.
Sections
The configuration options for a project often grow quite a lot;
to restrict complexity, getconf
splits values into sections,
similar to Python's configparser
module.
Section are handled differently depending on the actual configuration
source:
* ``section.key`` is mapped to ``MYPROJ_SECTION_KEY`` for environment variables
* ``section.key`` is mapped to ``[section] key =`` in configuration files
* ``section.key`` is mapped to ``defaults['section']['key']`` in the defaults dict.
Default section
Some settings are actually "globals" for a projet.
This is handled by unset section names:
* ``key`` is mapped to ``MYPROJ_KEY`` for environment variables
* ``key`` is mapped to ``[DEFAULT] key =`` in configuration files
* ``key`` is mapped to ``defaults['DEFAULT']['key']`` in the defaults dict.
.. _PyPI: http://pypi.python.org/