Environment Variable Parsing for Python
Switch branches/tags
Nothing to show
Clone or download
rconradharris Merge branch 'correct-kwarg-name-for-type' of https://github.com/ivaj…
…loip/envparse into ivajloip-correct-kwarg-name-for-type

Latest commit e67e703 Mar 12, 2016



envparse is a simple utility to parse environment variables.

If you use Heroku and/or subscribe to the tenets of the 12 Factor App you'll be using a lot of environment variable-based configuration in your app. os.environ is a great choice to start off with but over time you'll find yourself duplicating quite a bit of code around handling raw environment variables.

envparse aims to eliminate this duplicated, often inconsistent parsing code and instead provide a single, easy-to-use wrapper.

Ideas, and code portions, have been taken from django-environ project but made framework agnostic.


Through PyPI:

$ pip install envparse


$ pip install git+https://github.com/rconradharris/envparse.git
$ git clone https://github.com/rconradharris/envparse && cd envparse
$ python setup.py install


In your settings or configuration module, first either import the standard parser or one with a schema:

# Standard
from envparse import env

# Schema
from envparse import Env
env = Env(BOOLEAN_VAR=bool, LIST_VAR=dict(cast=list, subcast=int))

env can then be called in two ways:

  • Type explicit: env('ENV_VAR_NAME', cast=TYPE, ...)
  • Type implicit (for Python builtin types only): env.TYPE('ENV_VAR_NAME', ...) If type is not specified, explicitly or implicitly, then the default type is str.

Casting to a specified type:

# Environment variable: MAIL_ENABLED=1

mail_enabled = env('MAIL_ENABLED', cast=bool)
# OR mail_enabled = env.bool('MAIL_ENABLED')
assert mail_enabled is True

Casting nested types:

# Environment variable: FOO=1,2,3
foo = env('FOO'), subcast=int)
# OR: foo = env('FOO', cast=list, subcast=int)
# Note that there is no way to implicitly call subcast.
assert foo == [1, 2, 3]

Specifying defaults:

# Environment variable MAX_ROWS has not been defined

max_rows = env.int('MAX_ROWS', default=100)
assert max_rows == 100

Proxying values, useful in Heroku for wiring up the environment variables they provide to the ones that your app actually uses:

# Environment variables: MAILGUN_SMTP_LOGIN=foo,

smtp_login = env('SMTP_LOGIN')
assert smtp_login == 'foo'

Now if you switch to using Mandrill as an email provider, instead of having to modify your app, you can simply make a configuration change:


There are also a few convenience methods:

  • env.json: parses JSON and returns a dict.
  • env.url: parses a url and returns a urlparse.ParseResult object.

Type specific notes:

  • list: the expected environment variable format is FOO=1,2,3 and may contain spaces between the commas as well as preceding or trailing whitespace.
  • dict: the expected environment variable format is FOO='key1=val1, key2=val2. Spaces are also allowed.
  • json: a regular JSON string such as FOO='{"foo": "bar"}' is expected.


Define a schema so you can only need to provide the cast, subcast, and defaults once:

# Environment variables: MAIL_ENABLED=0, LIST_INT='1,2,3'

# Bind schema to Env object to get schema-based lookups
env = Env(MAIL_ENABLED=bool, SMTP_LOGIN=dict(cast=str, default='foo'),
          LIST_INT=dict(cast=list, subcast=int))
assert env('MAIL_ENABLED') is False
assert env('SMTP_LOGIN') == 'foo' # Not defined so uses default
assert env('LIST_INT') == [1, 2, 3]

The Env constructor takes values in the form of either: VAR_NAME=type or VAR_NAME=dict where dict is a dictionary with either one or more of the following keys specified: cast, subcast, default.

Pre- and Postprocessors

Preprocessors are callables that are run on the environment variable string before any type casting takes place:

# Environment variables: FOO=bar

# Preprocessor to change variable to uppercase
to_upper = lambda v: v.upper()
foo = env('FOO', preprocessor=to_upper)
assert foo == 'BAR'

Postprocessors are callables that are run after the type casting takes place. An example of one might be returning a datastructure expected by a framework:

# Environment variable: REDIS_URL='redis://:redispass@'
def django_redis(url):
  return {'BACKEND': 'django_redis.cache.RedisCache',
      'LOCATION': '{}:{}:{}'.format(url.hostname, url.port, url.path.strip('/')),
      'OPTIONS': {'PASSWORD': url.password}}

redis_config = env('REDIS_URL', postprocessor=django_redis)
assert redis_config == {'BACKEND': 'django_redis.cache.RedisCache',
    'LOCATION': '', 'OPTIONS': {'PASSWORD': 'redispass'}}

Environment File

Read from a .env file (line delimited KEY=VALUE):

# This recurses up the directory tree until a file called '.env' is found.

# Manually specifying a path

# Values can be read as normal



To run the tests install tox:

pip install tox

Then run them with:

make test