tonyseek/python-envcfg

For 12-factor apps - retrieve config from envvars.

python-envcfg

Accessing environment variables with a magic module.

>>> import os
>>> from envcfg.raw.python import CONFIGURE_OPTS
>>>
>>> CONFIGURE_OPTS
'--enable-shared --enable-universalsdk=/ --with-universal-archs=intel'
>>> CONFIGURE_OPTS == os.environ['PYTHON_CONFIGURE_OPTS']
True

It works with many frameworks such as Django and Flask. Then you can store your config in the environment variables instead of framework-specific config files. It is recommended by 12-Factor.

Installation

$ pip install python-envcfg

Supported Formats

• import envcfg.raw.foo as config: Import each FOO_* environment variable as string.
• import envcfg.json.foo as config: Import each FOO_* environment variable as JSON body.
• import envcfg.smart.foo as config: Try to import each FOO_* environment variable as JSON body, if fail then import it as string.

There is an example:

	envcfg.raw.foo.VAL	envcfg.json.foo.VAL	envcfg.smart.foo.VAL
FOO_VAL=foo	'foo'	ImportError	'foo'
FOO_VAL="foo"	'"foo"'	'foo'	'foo'
FOO_VAL=42	'42'	42	42
FOO_VAL="42"	'"42"'	'42'	'42'

Framework Integration

Flask

1. Defines environment variables with a prefix:

   $ cat .env  # should not checked into VCS
   # values are valid JSON expressions
   MYAPP_DEBUG=true
   MYAPP_SECRET_KEY='"7950ad141c7e4b3990631fcdf9a1d909"'
   MYAPP_SQLALCHEMY_DATABASE_URI='"sqlite:///tmp/myapp.sqlite3"'
   

2. Creates Flask app and loads config from python-envcfg:

   $ cat myapp.py
   ...
   app = Flask(__name__)
   app.config.from_object('envcfg.json.myapp')  # MYAPP_ -> .myapp
   ...
   

3. Enters your app with those environment variables:

   $ env $(cat .env | xargs) python myapp.py
   

Django

1. Creates a django project and moves all sensitive config items into the environment variables:

   $ cat djapp/settings.py  # codebase-scope config
   ...
   INSTALLED_APPS = (
       'django.contrib.admin',
   )
   ...
   
   $ cat .env  # environment-scope config, should not checked into VCS
   # values are valid JSON expressions
   DJAPP_SECRET_KEY='"wo9g2o#jws=u"'
   DJAPP_DEBUG=true
   DJAPP_TEMPLATE_DEBUG=true
   

2. Adds importing statements in the end of settings.py module:

   $ tail -n 2 djapp/settings.py
   # importing all config items stored in the environment variables
   from envcfg.json.djapp import *  # noqa
   

3. Runs your Django app with environment variables:

   $ env $(cat .env | xargs) python manage.py runserver
   

Tornado

1. Defines environment variables with a prefix:

   $ cat .env
   export TORAPP_PORT='8888'
   export TORAPP_MYSQL_HOST='"127.0.0.1"'
   export TORAPP_MYSQL_DATABASE='"database"'
   

2. Creates a Tornado project and loads config:

   $ cat torapp/server.py
   
   from tornado.web import Application, RequestHandler
   from tornado.ioloop import IOLoop
   from tornado.options import define, options
   from tordb import Connection
   
   
   def options_from_object(*args, **kwargs):
       module = __import__(*args, **kwargs)
       for name, value in vars(module).items():
           name = name.lower()
           if name in options._options:
               options._options[name].set(value)
   
   
   class IndexHandler(RequestHandler):
       def initialize(self):
           self.db = Connection(options.mysql_host, options.mysql_database)
   
       def get(self):
           pass  # some database operations with ``self.db``
   
   
   application = Application([
       (r'/', IndexHandler),
   ])
   
   define('port', type=int)
   define('mysql_host', type=unicode)
   define('mysql_database', type=unicode)
   options_from_object('envcfg.json.torapp', fromlist=['torapp'])
   
   
   if __name__ == '__main__':
       application.listen(options.port)
       IOLoop.instance().start()
   

3. Runs your Tornado app:

   $ env $(cat .env | xargs) python server.py
   

Works on Projects

In development, we can work with per-project environments but no more typing source foo/bar.

I recommend to put your project-specified environment variables in {PROJECT_ROOT}/.env and mark the .env as ignored in your VCS. For example, you can write /.env in .gitignore if you are using Git, and put a .env.example as a copying template for new-cloned projects.

And then, you can use some utility such as honcho or autoenv to apply the .env automatically.

For honcho:

$ echo 'MYPROJECT_DEBUG=true' >> .env
$ echo 'web: python manage.py runserver' >> Procfile
$ honcho run python manage.py check-debug
True
$ honcho start web
Starting development server at http://127.0.0.1:5000/
...

For autoenv:

$ echo 'MYPROJECT_DEBUG=true' >> myproject/.env
$ cd myproject
$ python manage.py check-debug
True
$ python manage.py runserver
Starting development server at http://127.0.0.1:5000/
...

Issues

If you want to report bugs or request features, please create issues on GitHub Issues.

Alternatives

• https://github.com/henriquebastos/python-decouple
• https://github.com/pallets/click

python-envcfg's Issues

跟直接用配置文件相比的优势

跟直接用配置文件相比的优势是什么？

更加隐蔽？