Mastering Multi‑Environment Python Configurations with Simple Packages

Overview: Python Modules and Packages

One of Python's best features is that the file and directory layout of your application mirrors the import statements you write. Placing an empty __init__.py in a directory turns it into a package.

When a utils.py file grows too large, you can split it into a utils/ directory containing many smaller modules.

Putting Code in __init__.py

Because __init__.py is just a regular Python file, any code placed there runs on the first import of the package. However, adding executable code can cause unexpected side effects, so it is generally discouraged.

Example: sys.exit() is often used to terminate a process with a specific status.

Multiple Environments & the Twelve‑Factor App

Applications often run in several environments—local development, CI testing, staging, production, etc. Configuration that rarely changes should live in code, while frequently changing or secret values (API keys, credentials) belong outside version control.

Static content that rarely changes belongs in code.

Dynamic or secret content belongs outside the codebase.

How to Switch Environments

Use an environment variable (commonly ENV) to indicate the current environment. If the variable is unset, default to development. The application can abort startup if the variable is missing.

A Real‑World Example

The project uses a config/ directory with the following layout:

Key files:

common.py : static configuration shared by all environments.

environments/development.py : development‑specific overrides (e.g., local Redis, AWS resources, secret credentials).

environments/production.py : production‑specific settings (e.g., SENTRY_DSN, Heroku Redis URL).

In development.py the site title is overridden to indicate a local copy, and local Redis settings differ from production.

How It Works

The config/__init__.py module uses importlib to load the appropriate environment module at import time:

Import importlib.

Read the ENV variable to determine the current environment.

If ENV is unset, fall back to development.

Optionally abort if the variable is missing.

Load the environment module with importlib.import_module.

Update the globals with the environment‑specific settings.

This mirrors the Rails configuration approach but is implemented in plain Python.

Conclusion

The method works for all Python projects, offering flexible, maintainable, and efficient configuration management with only a few lines of standard Python code.