I use this project with my django projection, deployed by uwsgi(emperor mode) + nginx, it worked well, thanks for your working, but when i touch the uwsgi config file, the env vars will not reload, for now, i can send signal 3 to the process which served my django app,, is there any good ways to make this done?

(sorry for my poor english...

Hi People,

I've encountered an issue while setting the EMAIL_HOST_PASSWORD environment variable.

If you have Dollar sign $ in the beginning of your env variable string django-environ treats it as another environment variable (nested variable): .env file:

EMAIL_HOST_PASSWORD='$my_password_including_$_sign_in_the_beginning'

settings.py

EMAIL_HOST_PASSWORD = env.str('EMAIL_HOST_PASSWORD')

This configuration raises:

django.core.exceptions.ImproperlyConfigured: Set the my_password_including_$_sign_in_the_beginning environment variable

which as you can see omitted the first $ sign.

Now if we try to escape the $ character with a backslash and change our .env file to this:

EMAIL_HOST_PASSWORD='\$my_password_including_$_sign_in_the_beginning'

django-environ escapes the backslash instead of the dollar sign $ and we end up with something like this:

>>> from django.conf import settings
>>> settings.EMAIL_HOST_PASSWORD
'\\$my_password_including_$_sign_in_the_beginning'

Which breaks our configurations and settings.

This is my proposal which if is agreed on I'll make a pull request:

# Current string reader function
def str(self, var, default=NOTSET, multiline=False):
    """
    :rtype: str
    """
    value = self.get_value(var, default=default)
    if multiline:
        return value.replace('\\n', '\n')
    return value


# My proposal for string function, which we can use escape=True while passing env variables that have such characters
def str(self, var, default=NOTSET, multiline=False, escape=False):
    """
    :rtype: str
    """
    value = self.get_value(var, default=default)
    if multiline:
        return value.replace('\\n', '\n')
    if escape:
        return value.replace('\$', '$')
    return value

This prevents django-environ's use of os.environ making changes to the ACTUAL environment. All other behavior kept the same. This fixes issues with UWSGI emperor mode where the environment is persistent between vassal restarts leading to any variable set in a .env file becoming semi-permanent as django-environ won't overwrite existing environment settings.

Instead of the old fix, which only copied the os.environ in read_env, this pull shallow copies os.environ inside env.init, meaning that we always have a copy of the environment whether we read an .env file or not.

All tests passing.

The >=1.11a1,<2.0 specifier will avoid having to adjust the version as the Django 1.11 pre-release cycle continues.

Also stops testing Django master against Python 2, since they dropped support for it.

I have

env = environ.Env(
    DEBUG=(bool, False),
)
ENV_FILE = str(env.path('ENV_FILE', default='/'))
if os.path.isfile(ENV_FILE):
    overrides = {'DATABASE_URL': os.environ.pop('DATABASE_URL', '')}
    env.read_env(env_file=ENV_FILE, **overrides)
else:
    # unset if no file was found
    ENV_FILE = None

Problem is that in my ENV_FILE, I want to set DATABASE_URL that already has a value when reading the ENV_FILE.

But because def read_env(env_file=None, **overrides) uses setdefault(key, str(val)) this value can not be overwritten.

Is it possible to extend def read_env with an extra boolean argument to decide whether or not to override existing environment variables?

There has not been a new release since 2018, yet there is a substantial amount of work done in the develop branch. Will any of this be released soon?

Docker (swarm) and Kubernetes are two widely used platforms that store their Secrets in tmpfs inside containers as individual files.

These are a secure way to be able to share configuration data between containers, and having a way to access these in django-environ is useful.

Here's a quote from the official Postgres docker hub image:

As an alternative to passing sensitive information via environment variables, _FILE may be appended to some of the previously listed environment variables, causing the initialization script to load the values for those variables from files present in the container. In particular, this can be used to load passwords from Docker secrets stored in /run/secrets/<secret_name> files. For example:

Hi there,

I've just updated from 0.4.4 to 0.4.5 and noticed a few unexpected behaviours with default. Presumably they relate to the casting changes that were introduced recently?

# the key is in my env
In [3]: env('GOOGLE_ANALYTICS_KEY')
Out[3]: 'UA-123456-78'

# I should have used a string as default here
# but I would still expect these to return my key instead of False

In [4]: env('GOOGLE_ANALYTICS_KEY', default=False)
Out[4]: False

In [7]: env('GOOGLE_ANALYTICS_KEY', default=True)
Out[7]: False

# these ones work as expected

In [5]: env('GOOGLE_ANALYTICS_KEY', default='')
Out[5]: 'UA-123456-78'

In [6]: env('GOOGLE_ANALYTICS_KEY', default=None)
Out[6]: 'UA-123456-78'

Cheers, Matthieu

Expanding variables automatically on a read is an anti-pattern. Variable expansion by the shell should only be done when the value is inserted into the environment, but the vlaue should be treated as opaque data. Any processing or interpretation of the variable should be done by the appliaction, not by the access method.

• Stop testing on unsupported versions of Django (versions prior to 1.8 are no longer maintained).
• Stop testing Django 1.9+ with Python 3.3 (only test against 1.8, the last version to support it).
• Test the latest point release of each major Django version (rather than the initial release of each).
• Test with Django 1.10b1 (this can be updated when the official version is released).
• Test using the PyPy Python release.
• Fix the AppVeyor test run (since the location of get-pip.py has changed).

See individual commit messages for more details.

This PR addresses issue #103. I've had this issue with the uwsgi touch-reload command. It won't pick up updates to .env files.

In some use cases, the .env file should be the master, so an overwrite=True option will resolve the issue in this case.

• All tests pass
• Should be no performance penalty for current use (other than an additional function call)
• Minor performance in overwrite use due to temp dict.

The temp dict was needed since there is no dict.set(k,v) method. The lambda can't do dict[k] = v.

Bumps actions/checkout from 3.0.2 to 3.3.0.

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.

Bumps actions/setup-python from 4.0.0 to 4.4.0.

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.

hello @joke2k

I would like to offer help, because we heavily use this package and we would be interested in its further development.

It seems like right now you might not have so much time, hence: let me know if I can assist.

cheers

There is case when it is required to read envs from files e.g. kubernetes mounted secret volume:

    spec:
      ...
        volumeMounts:
          - name: webapp-secret
            mountPath: "/mnt/secrets"
            readOnly: true
        imagePullPolicy: Never
      ...

I would be happy to see such enhancement!

I was wondering if there has been discussions/plans to support embedded variables in .env file. For example,

# .env file with embedded variables
LIB_PATH=/home/username/.lib
GDAL_LIBRARY_PATH=${LIB_PATH}/osgeo/gdal304.dll
GEOS_LIBRARY_PATH=${LIB_PATH}/osgeo/geos_c.dll

I got used to the ${VAR_NAME} syntax as it is supported by docker-compose and python-dotenv among others. It would be great to get the same support in django-environ. I'm aware of the interpolate feature which covers a more limited use case.

Would that be useful to anyone else?