A thin wrapper over Pydantic's settings management. Allows you to define configuration variables and load them from environment or JSON/YAML file. Also generates initial configuration files and documentation for your defined configuration.
pip install goodconf or pip install goodconf[yaml] if
parsing/generating YAML files is required.
Let's use configurable Django settings as an example.
First, create a conf.py file in your project's directory, next to
settings.py:
import base64
import os
from goodconf import GoodConf, Field
from pydantic import PostgresDsn
class AppConfig(GoodConf):
"Configuration for My App"
DEBUG: bool
DATABASE_URL: PostgresDsn = "postgres://localhost:5432/mydb"
SECRET_KEY: str = Field(
initial=lambda: base64.b64encode(os.urandom(60)).decode(),
description="Used for cryptographic signing. "
"https://docs.djangoproject.com/en/2.0/ref/settings/#secret-key")
class Config:
default_files = ["/etc/myproject/myproject.yaml", "myproject.yaml"]
config = AppConfig()Next, use the config in your settings.py file:
import dj_database_url from .conf import config config.load() DEBUG = config.DEBUG SECRET_KEY = config.SECRET_KEY DATABASES = {"default": dj_database_url.parse(config.DATABASE_URL)}
In your initial developer installation instructions, give some advice such as:
python -c "import myproject; print(myproject.conf.config.generate_yaml(DEBUG=True))" > myproject.yamlBetter yet, make it a function and entry point so you can install
your project and run something like generate-config > myproject.yaml.
Your subclassed GoodConf object can include a Config class with the following
attributes:
file_env_var- The name of an environment variable which can be used for the name of the configuration file to load.
default_files- If no file is passed to the
loadmethod, try to load a configuration from these files in order.
It also has one method:
load- Trigger the load method during instantiation. Defaults to False.
Use plain-text docstring for use as a header when generating a configuration file.
Declare configuration values by subclassing GoodConf and defining class
attributes which are standard Python type definitions or Pydantic FieldInfo
instances generated by the Field function.
Goodconf can use one extra argument provided to the Field to define an function
which can generate an initial value for the field:
initial- Callable to use for initial value when generating a config
A helper is provided which monkey-patches Django's management commands to
accept a --config argument. Replace your manage.py with the following:
# Define your GoodConf in `myproject/conf.py`
from myproject.conf import config
if __name__ == '__main__':
config.django_manage()I took inspiration from logan (used by Sentry) and derpconf (used by Thumbor). Both, however used Python files for configuration. I wanted a safer format and one that was easier to serialize data into from a configuration management system.
I don't like working with environment variables. First, there are potential security issues:
- Accidental leaks via logging or error reporting services.
- Child process inheritance (see ImageTragick for an idea why this could be bad).
Second, in practice on deployment environments, environment variables end up getting written to a number of files (cron, bash profile, service definitions, web server config, etc.). Not only is it cumbersome, but also increases the possibility of leaks via incorrect file permissions.
I prefer a single structured file which is explicitly read by the application. I also want it to be easy to run my applications on services like Heroku where environment variables are the preferred configuration method.
This module let's me do things the way I prefer in environments I control, but still run them with environment variables on environments I don't control with minimal fuss.
Create virtual environment and install package and dependencies.
Run tests