Config Class

Introduction

Config Class is a simple package for managing configuration settings in smaller Python projects.

It's goal is to be as simple as writing a dataclass, with automatic handling of multiple configuration sources.

Features

• Easy Configuration: The Config Class simplifies the process of managing configuration settings in your Python projects.
• File-based Configuration: The Config Class can load configuration settings from a yaml or json file, making it easy to switch between different configurations.
• Support command line: Automatically generates command-line switches for the configuration options.
• Env vars support: Configuration can also be set via environment variables.

Installation

To use the Config Class in your Python project, you can install it via pip:

pip install git+https://github.com/TheTerrasque/python-configclass

Examples

Quick start

settings_test.py

from dataclasses import dataclass, field
from configclass import configclass

@dataclass
class Test:
    test: str = "test"

@configclass()
class Settings:
    foo: bool = False
    url: str = ""
    footoo: bool = True
    my_model: str = "model.pt"
    test: Test = field(default_factory=Test)


setting = Settings.load()

print(setting.foo, setting.footoo, setting.test.test, setting.my_model)

$ python setting_test.py --foo true --footoo false --test-test "Test me" --my_model "model2.pt"

config.yaml:

foo: true
test:
    test: "Hello world"

config.json:

{
    "foo": true, 
    "test" {
        "test": "Hello world"
    }
}

Environment variables:

$ CONFIG_FOO="true" python settings_test.py
$ CONFIG_TEST_TEST="Hey mate" python settings_test.py

Overriding defaults

@configclass(yaml_file="config.yaml", 
    json_file="config.json", 
    argparse_text="Program configuration command line arguments", 
    env_prefix="CONFIG")

Setting any of these to None will disable that part as a config source.

TODO

• Add optional documentation to fields for use in argparse, most likely via field() metadata
• Make config file used settable via command line or env vars

Backstory

Most of my python projects are fairly small scripts, but they still have parts I want to configure. I usually use a dataclass for it, and use yaml or json to load from file and populate it. Later on I tend to add some command line arguments and/or env variables (for docker / kubernetes deployments).

After writing the same kind of code for the nth time, I decided to put together a small package for handling just that. I searched a bit online, but couldn't find a simple solution I liked. OmegaConf is a bit too heavy for my small projects, and I never really liked hydra.