In the last Understand Django article, we looked at a storage concept in Django called sessions. Sessions help us answer questions like “How does Django know when a user is logged in?” or “Where can the framework store data for a visitor on your app?”
With this article, you’ll learn about Django settings and how to manage the configuration of your application. We’ll also look at tools to help you to be extra effective with settings.
- From Browser To Django
- URLs Lead The Way
- Views On Views
- Templates For User Interfaces
- User Interaction With Forms
- Store Data With Models
- Administer All The Things
- Anatomy Of An Application
- User Authentication
- Middleware Do You Go?
- Serving Static Files
- Test Your Apps
- Deploy A Site Live
- Per-visitor Data With Sessions
- Making Sense Of Settings
How Is Django Configured?
To run properly, Django needs to be configured. We need to understand where this configuration comes from. Django has the ability to use default configuration values or values set by developers like yourself, but where does it get those from?
Early in the process of starting a Django application, Django will internally import the following:
from django.conf import settings
settings import is a module level object
settings object has attributes added to it
from two primary sources.
The first source is a set of global default settings
that come from the framework.
These global settings are from
and provide a set of initial values
that Django needs to operate.
The second source of configuration settings comes
from user defined values.
Django will accept a Python module
and apply its module level attributes
To find the user module,
Django searches for a
DJANGO_SETTINGS_MODULE environment variable.
Sidebar: Environment Variables
Environment variables are not a Django concept. When any program runs on a computer, the operating system makes certain data available to the running program. This set of data is called the program’s “environment,” and each piece of data in that set is an environment variable.
If you’re starting Django from a terminal,
you can view the environment variables
that Django will receive from the operating system
by running the
env command on macOS or Linux,
set command on Windows.
We can add our own environment variables
to the environment
export command on macOS or Linux,
set command on Windows.
Environment variables are typically named
in all capital letters.
$ export HELLO=world
Now that we have a base understanding
of environment variables,
let’s return to the
The variable’s value should be the location
of a Python module containing any settings
that a developer wants to change
from Django’s default values.
If you create a Django project
project as the name,
then you will find a generated file called
in the output.
When Django runs,
you could explicitly instruct Django with:
$ export DJANGO_SETTINGS_MODULE=project.settings
Instead of supplying the file path,
DJANGO_SETTINGS_MODULE should be
in a Python module dotted notation.
You may not actually need
If you stick with the same settings file
that is created by
you can find a line
that looks like:
Because of this line,
Django will attempt to read
(or whatever you named your project)
without the need to explicitly set
Once Django reads the global settings
and any user defined settings,
we can get any configuration
via attribute access.
This convention of keeping all configuration
is a convenient pattern
that the framework,
third party library ecosystem,
and you can depend on.
$ ./manage.py shell >>> from django.conf import settings >>> settings.SECRET_KEY 'a secret to everybody'
settings object is a shared item
so it is generally thought
to be a Really Bad Idea™
to edit and assign to the object directly.
Keep your settings in your settings module!
That’s the core of Django configuration. We’re ready to focus in on the user defined settings and our responsibilities as Django app developers.
Settings Module Patterns
There are multiple ways to deal with settings modules and how to populate those modules with the appropriate values for different environments. Let’s look at some popular patterns.
Multiple Modules Per Environment
A Django settings module is a Python module. Nothing is stopping us from using the full power of Python to configure that module the way we want.
Minimally, you will probably have at least two environments where your Django app runs:
- On your local machine while developing
- On the internet for your live site
We should know by now
DEBUG = True is a terrible idea
for a live Django site,
so how can we get the benefits of the debug mode
DEBUG set to
in our module?
One technique is to use separate settings modules.
With this strategy,
you can pick which environment your Django app should run for
by switching the
to pick a different environment.
You might have modules like:
These examples would be for a local development environment on your laptop, a staging environment (which is a commonly used pattern for testing a site that is as similar to the live site as possible without being the live site), and a production environment. As a reminder from the deployment article, the software industry like to call the primary site for customers “production.”
This strategy has certain challenges to consider. Should you replicate settings in each file or use some common module between them?
If you decide to replicate the settings across modules, you’ll have the advantage that the settings module shows all of the settings in a single place for that environment. The disadvantage is that keeping the common settings the same could be a challenge if you forget to update one of the modules.
On the other hand, you could use a common module. The advantage to this form is that the common settings can be in a single location. The environment specific files only need to record the differences between the environments. The disadvantage is that it is harder to get a clear picture of all the settings of that environment.
If you decide to use a common module,
this style is often implemented
I can probably count on one hand the number
where I’m ok with a
and this is one of them.
In most cases the Python community prefers explicit over implicit,
and the idea extends to the treatment of imports.
Explicit imports make it clear what a module is actually using.
* import is very implicit,
and it makes it unclear what a module uses.
For the case of a common settings module,
* import is actually positive
because we want to use everything
in the common module.
Let’s make this more concrete.
Assume that you have a
This module would hold your common settings
for your app.
I’d recommend that you try to make your settings safe and secure
DEBUG = False in the base settings
and force other settings modules
to the more unsafe behavior.
For your local development environment
on your laptop,
you could use
This settings module would look like:
# project/settings/dev.py from project.settings.base import * DEBUG = True # Define any other settings that you want to override. ...
By using the
* import in the
all the settings from
base.py are pulled
into the module level scope.
Where you want a setting to be different,
you set the value in
When Django starts using
all the values from
base.py will be used
This scheme gives you control to define common things once, but there is still a big challenge with this. What do we do about settings that need to be kept secret (e.g., API keys)?
Don’t commit secret data to your code repository! Adding secrets to your source control tool like Git is usually not a good idea. This is especially true if you have a public repository on GitHub. Think no one is paying attention to your repo? Think again! There are tools out there that scan every public commit made to GitHub. These tools are specifically looking for secret data to exploit.
If you can’t safely add secrets to your code repo, where can we add them instead? You can use environment variables! Let’s look at another scheme for managing settings with environment variables.
Settings Via Environment Variables
you can access environment variables
The module contains the
which functions like a dictionary.
By using environment variables, your settings module can get configuration settings from the external environment that is running the Django app. This is a solid pattern because it can accomplish two things:
- Secret data can be kept out of your code
- Configuration differences between environments can be managed by changing environment variable values
Here’s an example of secret data management:
# project/settings.py import os SECRET_KEY = os.environ['SECRET_KEY'] ...
Django needs a secret key
for a variety of safe hashing purposes.
There is a warning in the default
# SECURITY WARNING: keep the secret key used in production secret!
By moving the secret key value
to an environment variable
that happens to have a matching name of
we won’t be committing the value
to source control
for some nefarious actor to discover.
This pattern works really well for secrets, but it can also work well for any configuration that we want to vary between environments.
on one of my projects,
I use the excellent Anymail package
to send emails
via an email service provider
(of the ESPs, I happen to use SendGrid).
When I’m working with my development environment,
I don’t want to send real email.
Because of that,
I use an environment variable
to set Django’s
This let’s me switch between the Anymail backend
and Django’s built-in
that prints emails to the terminal instead.
If I did this email configuration with
it would look like:
# project/settings.py import os EMAIL_BACKEND = os.environ.get( 'EMAIL_BACKEND', "anymail.backends.sendgrid.EmailBackend") ...
I prefer to make my default settings closer
to the live site context.
This not only leads to safer behavior
(because I have to explicitly opt-out
of safer settings like switching to
DEBUG = False),
but it also means that my live site has less to configure.
That’s good because there are fewer chances
to make configuration mistakes
on the site that matters most:
the one where my customers are.
We need to be aware of a big gotcha
with using environment variables.
Environment variables are only available as a
This is something to be aware
because there will be times when you want a boolean settings value
or some other type of data.
In a situation
where you need a different type,
you have to coerce a
into the type you need.
In other words,
don’t forget that every string
except the empty string is truthy
>>> not_false = "False" >>> bool(not_false) True
In the next section, we will see tools that help alleviate this typing problem.
Note: As you learn more about settings, you will probably encounter advice that says to avoid using environment variables. This is well intentioned advice that highlights that there is some risk with using environment variables. With this kind of advice, you may read a recommendation for secrets management tools like HashiCorp Vault. These are good tools, but consider them a more advanced topic. In my opinion, using environment variables for secrets management is a reasonably low risk storage mechanism.
Settings Management Tools
We can focus on two categories of tools that can help you manage your settings in Django: built-in tools and third party libraries.
The built-in tool that is available to you
This tool makes it easy
to see the computed settings
of your module.
Since settings can come
from multiple files
or environment variables,
inspecting the settings output
diffsettings is more convenient
than thinking through how a setting is set.
diffsettings will show a comparison
of the settings module
to the default Django settings.
Settings that aren’t in the defaults are marked
### after the value
to indicate that they are different.
I find that the default output is not the most useful mode.
you can instruct
to output in a “unified” format.
This format looks a lot more like a code diff.
Django will colorize that output
so that it’s easier to see.
Here’s an example
of some of the security settings
./manage.py diffsettings --output unified
for one of my projects.
- SECURE_HSTS_INCLUDE_SUBDOMAINS = False + SECURE_HSTS_INCLUDE_SUBDOMAINS = True - SECURE_PROXY_SSL_HEADER = None + SECURE_PROXY_SSL_HEADER = ('HTTP_X_FORWARDED_PROTO', 'https')
Finally, I’ll note that you can actually compare two separate settings modules. Let’s say you wanted to compare settings between your development mode and your live site. Assuming your settings files are names like I described earlier, you could run something like:
$ ./manage.py diffsettings \ --default project.settings.dev \ --settings project.settings.prod \ --output unified
By using the
we instruct Django that
project.settings.dev is the baseline
This version of the command will show where the two settings modules are different.
Django only includes this single tool for working with settings, but I hope you can see that it’s really handy. Now let’s talk about a useful third party library that can help you with settings.
Earlier in the article, I noted that dealing with environment variables has the pitfall of working with string data for everything. Thankfully, there is a package that can help you work with environment variables. The project is called django-environ. django-environ primarily does two important things that I value:
- The package allows you coerce strings into a desired data type.
- The package will read from a file to load environment variables into your environment.
What does type coercion look like?
you start with
# project/settings.py import environ env = environ.Env()
The keyword arguments to
Env describe the different environment variables
that you expect the app to process.
The key is the name of the environment variable.
The value is a two element tuple.
The first tuple element is the type you want,
and the second element is a default value
if the environment variable doesn’t exist.
If you want to be able to control
from an environment variable,
the settings would be:
# project/settings.py import environ env = environ.Env( DEBUG=(bool, False), ) DEBUG = env("DEBUG")
With this setup,
your app will be safe by default
DEBUG set to
but you’ll be able to override
that via the environment.
django-environ works with a handful
of strings that it will accept as
such as “on”, “yes”, “true”, and others
(see the documentation for more details).
Once you start using environment variables,
you’ll want an convenient way to set them
when your app runs.
export for all your variables
before running your app is a totally unsustainable way
to run apps.
Env class comes with a handy class method named
With this method,
your app can read environment variables
from a file.
this file is named
and the file contains a list of key/value pairs
that you want as environment variables.
Following our earlier example,
here’s how we could set our app
to be in debug mode:
# .env DEBUG=on
Back in the settings file, you’d include
# project/settings.py import environ environ.Env.read_env() env = environ.Env( DEBUG=(bool, False), ) DEBUG = env("DEBUG")
If you use a
you will occasionally find a need to put secrets
into this file
Since the file can be a source for secrets,
you should add this to
or ignore it
in whatever version control system you use.
As time goes on,
the list of variables and settings will likely grow,
so it’s also a common pattern
to create a
that you can use as a template
in case you ever need to start
with a fresh clone of your repository.
My Preferred Settings Setup
Now we’ve looked at multiple strategies and tools for managing settings. I’ve used many of these schemes on various Django projects, so what is my preferred setup?
For the majority of uses,
I find that working with
in a single file is the best pattern
in my experience.
When I use this approach, I make sure that all of my settings favor a safe default configuration. This minimizes the configuration that I have to do for a live site.
I like the flexibility of the pattern, and I find that I can quickly set certain configurations when developing. For instance, when I want to do certain kinds of testing like checking email rendering, I’ll call something like:
$ EMAIL_TESTING=on ./manage.py runserver
My settings file has a small amount of configuration to alter the email settings to point emails to a local SMTP server tool called MailHog. Because I set an environment variable directly on my command line call, I can easily switch into a mode that sends email to MailHog for quick review.
Overall, I like the environment variable approach, but I do use more than one settings file for one important scenario: testing.
When I run my unit tests,
I want to guarantee
that certain conditions are always true.
There are things
that a test suite should never do
in the vast majority of cases.
Sending real emails is a good example.
If I happen to configure my
to test real emails for the local environment,
I don’t want my tests
to send out an emails accidentally.
Thus, I create a separate testing settings file and configure my test runner (pytest) to use those settings. This settings file does mostly use the base environment, but I’ll override some settings with explicit values. Here’s how I protect myself from accidental live emails:
# project/testing_settings.py from .settings import * # Make sure that tests are never sending real emails. EMAIL_BACKEND = "django.core.mail.backends.locmem.EmailBackend"
Even though my
Env will look for an
EMAIL_BACKEND environment variable
to configure that setting dynamically,
the testing setting is hardcoded
to make email sending accidents impossible.
The combination of a single file for most settings sprinkled with a testing settings file for safety is the approach that has worked the best for me.
In this article, you learned about Django settings and how to manage the configuration of your application. We covered:
- How Django is configured
- Patterns for working with settings in your projects
- Tools that help you observe and manage settings
In the next article, we will look at how to handle files and media provided by users (e.g., profile pictures). You’ll learn about:
- How Django models maintain references to files
- How the files are managed in Django
- Packages that can store files in various cloud services
If you’d like to follow along with the series, please feel free to sign up for my newsletter where I announce all of my new content. If you have other questions, you can reach me online on Twitter where I am @mblayman.
Learn about Python!
You can join my newsletter along with 1,200+ other developers to help you learn more about Django and Python.