envenom

envenom

Introduction

envenom is an elegant application configurator for the more civilized age.

envenom is written with simplicity and type safety in mind. It allows you to express your application configuration declaratively in a dataclass-like format while providing your application with type information about each entry, its nullability and default values.

envenom is designed for modern usecases, allowing for pulling configuration from environment variables or files for more sophisticated deployments on platforms like Kubernetes - all in the spirit of 12factor.

How it works

An envenom config class looks like a regular Python dataclass - because it is one.

The @envenom.config decorator creates a new dataclass by converting the config fields into their dataclass equivalents providing the relevant default field parameters.

This also means it's 100% compatible with dataclasses. You can:

• use a config class as a property of a regular dataclass
• use a regular dataclass as a property of a config class
• declare static or dynamic fields using standard dataclass syntax
• use the InitVar/__post_init__ method for delayed initialization of fields
• use methods, classmethods, staticmethods, and properties

envenom will automatically fetch the environment variable values to populate dataclass fields, optionally running parsers so that fields are automatically converted to desired types. This works out of the box with all built-in types trivially convertible from str (like StrEnum and UUID) and with any object type that can be instantiated easily from a single string (any function (str,) -> T will work as a parser).

If using a static type checker, the type deduction system will correctly identify most mistakes if you declare fields, parsers or default values with mismatched types. There are certain exceptions (for example T will always satisfy type bounds T | None).

envenom also offers reading variable contents from file by specifying an environment variable with the suffix __FILE which contains the path to a file with the respective secret. This aims to facilitate a common deploy pattern where secrets are mounted as files (especially prevalent with Kubernetes).

What envenom isn't

envenom has a clearly defined scope limited to configuration management from the application's point of view.

This means envenom is only interested in converting the environment into application configuration and does not care about how the environment gets populated in the first place.

Things that are out of scope for envenom include, but are not limited to:

• injecting the environment into the runtime or orchestrator
• retrieving configuration or secrets from the cloud or another storage (AWS Parameter/Secret Store, Azure Key Vault, HashiCorp Vault, etc.)
• retrieving and parsing configuration from structured config files (YAML/JSON/INI etc.)

Getting started

Installation

python -m pip install envenom

Config classes

Config classes are created with the envenom.config class decorator. It behaves exactly like dataclasses.dataclass but allows to replace standard dataclasses.field definitions with one of envenom-specific configuration field types.

To append a prefix to all your environment variable names within a config class, a namespace needs to be created.

from uuid import UUID, uuid4

from envenom import config, defaults, namespace, optional, required


@config(namespace("myapp"))
class AppCfg:
    required_str = required()
    optional_int = optional(int)
    defaults_uuid = defaults(UUID, default_factory=uuid4)

Field types

envenom offers three supported field types:

• required for configuration variables that have to be provided. If the value cannot be found, ConfigurationMissing will be raised.
• optional for configuration variables that don't have to be provided. If the value cannot be found, it will be set to None.
• defaults for configuration variables where a default value can be provided. If the value cannot be found, it will be set to the default, which can be either a static value or created at instantiation by a factory function.

Supplying values

To generate the environment variable name:

• join all namespace segments and the variable name together with __
• replace strings of nonsensical characters ([^0-9a-zA-Z_]+) with _
• transform to uppercase

As an example, a field named dsn in a config class with Namespace("myapp", "db") will be mapped to MYAPP__DB__DSN.

Basic usage - complete example

This example shows how to build a basic config structure for an application using a database service with injectable configuration as an example. It is available in the envenom.examples.quickstart runnable module.

from functools import cached_property
from uuid import UUID, uuid4

from envenom import config, defaults, namespace, optional, required, subconfig
from envenom.examples import print_config_tree
from envenom.parsers import bool_parser

myapp = namespace("myapp")
myapp_db = myapp / "db"


@config(myapp_db)
class DbCfg:
    scheme: str = defaults(default="postgresql+psycopg")
    host: str = required()
    port: int = defaults(int, default=5432)
    database: str = required()
    username: str | None = optional()
    password: str | None = optional()
    connection_timeout: int | None = optional(int)
    sslmode_require: bool = defaults(bool_parser, default=False)

    @cached_property
    def auth(self) -> str:
        if not self.username and not self.password:
            return ""

        auth = ""
        if self.username:
            auth += self.username
        if self.password:
            auth += f":{self.password}"
        if auth:
            auth += "@"

        return auth

    @cached_property
    def query_string(self) -> str:
        query: dict[str, str] = {}
        if self.connection_timeout:
            query["timeout"] = str(self.connection_timeout)
        if self.sslmode_require:
            query["sslmode"] = "require"

        if not query:
            return ""

        query_string = "&".join((f"{key}={value}" for key, value in query.items()))
        return f"?{query_string}"

    @cached_property
    def connection_string(self) -> str:
        return (
            f"{self.scheme}://{self.auth}{self.host}:{self.port}"
            f"/{self.database}{self.query_string}"
        )


@config(myapp)
class AppCfg:
    worker_id: UUID = defaults(UUID, default_factory=uuid4)
    secret_key: str = required()
    db: DbCfg = subconfig(DbCfg)


if __name__ == "__main__":
    cfg = AppCfg()

    print_config_tree(cfg)
    print(f"cfg.db.connection_string: {repr(cfg.db.connection_string)}")

Run the example:

python -m envenom.examples.quickstart

Traceback (most recent call last):
    ...
    raise ConfigurationMissing(self.env_name)
envenom.errors.ConfigurationMissing: 'MYAPP__SECRET_KEY'

As soon as it encounters a required field, the config class returns an error because there's no environment set.

Run the example again with the environment set:

MYAPP__SECRET_KEY='}uZ?uvJdKDM+$2[$dR)).n4q1SX!A$0u{(+D$PVB' \
MYAPP__DB__HOST='postgres.example.com' \
MYAPP__DB__DATABASE='test-database' \
MYAPP__DB__USERNAME='user' \
MYAPP__DB__SSLMODE_REQUIRE='t' \
MYAPP__DB__CONNECTION_TIMEOUT='15' \
python -m envenom.examples.quickstart

-----
cfg:
  worker_id: UUID('2fd334a5-5f08-4815-8107-928c291264c3')
  secret_key: '}uZ?uvJdKDM+$2[$dR)).n4q1SX!A$0u{(+D$PVB'
  db:
    scheme: 'postgresql+psycopg'
    host: 'postgres.example.com'
    port: 5432
    database: 'test-database'
    username: 'user'
    password: None
    connection_timeout: 15
    sslmode_require: True
-----
cfg.db.connection_string: 'postgresql+psycopg://[e-mail]:5432/test-database?timeout=15&sslmode=require'