MyENV: Environment Variable Parsing using Type annotations

Project/Repo:

Code Quality/CI:

MyENV parses you're environment variables using type annotations. This allows you to configure your app/service as layed out by 12factor.net/config, while keeping your code type safe.

• Environment Variables and Configuration
  • Declaration
  • Parsing
• my_service/cfg.py
  • Config Files
• config/prod.env

Environment Variables and Configuration

In order of priority, configuration is parsed from

1. Environment variables
2. Configuration files
3. Defaults defined in source code

Declaration

The myenv module provides a convenient way to do this parsing. As far as your application code is concerned, it is enough to declare a subclass of myenv.BaseEnv. Instances of this subclass are populated from config files and the environment.

import myenv


class Database(myenv.BaseEnv):

    _environ_prefix = "DATABASE_"

    vendor    : str  = "postgres"
    host      : str  = "127.0.0.1"
    port      : int  = 5432
    user      : str  = "myuser"
    password  : str
    name      : str  = "app_db_v1"

    @property
    def url(self) -> str:
        db = self
        return f"{db.vendor}://{db.user}:{db.password}@{db.host}:{db.port}/{db.name}"

For each annoatated member of DBEnv declare the 1. name, 2. type and 3. an optional default variable. Members without a default must be set by an environment variable or configuration file, otherwise a KeyError will be raised.

Parsing

To use the above configuration class, simply instanciate it.

# my_service/cfg.py
import sqlalchemay as sqla

db_cfg = Database()     # populated from os.environ
db_cfg.port             # 12345          (parsed from DATABASE_PORT)
db_cfg.password         # "supersecret"  (parsed from DATABASE_PASSWORD)
db_cfg.url              # "mysql://myuser:supersecret@127.0.0.1:12345/mydb"

engine = sqla.create_engine(db_cfg.url)

In case you're worried about the instantiation of Database(), it will return a singleton instance, so the os.environ and configuration files are parsed only once.

Config Files

When parsing configs, the following paths are parsed if they exist. By default the configs are loaded from ${PWD}/config/, but you can override the location of configuration files by setting the environment variable ENV_CONFIG_DIR.

• ${ENV_CONFIG_DIR}/${ENV}.env
• ${ENV_CONFIG_DIR}/prod.env

Any variables defined in these files will be set, unless it was already set in the environ or a previous config file.

This approach allows you to satisfy typical use cases in which a service is hosted:

1. Development Machine
2. Stage/Production Environment

You can put use case specifc configuration files in your project source tree, such as:

• project/config/dev.env
• project/config/stage.env
• project/config/prod.env

To parse the appropriate config file, all you have to do is set a single environment variable ENV=<envname>. If ENV is not set, it will fall back to ENV=prod.

The *.env config files are flat text files, with one KEY=value mapping per line. The only lines which are parsed, are lines which begin with an all upper case token, followed by an = character:

# config/prod.env
DATABASE_PORT=12345
DATABASE_USER=prod_user
DATABASE_NAME=prod_db
DATABASE_PASSWORD=supersecret

For sensitive configuration parameters, such as passwords and authentication tokens, you may prefer to write config files outside of your source tree, or to provide them always and only via environment variables.

Changelog for https://gitlab.com/mbarkhau/myenv

v201812.0003-beta

• Add config parsing
• Simplify instantiation (no need to import myenv to use subclasses)
• Switch to https://gitlab.com/mbarkhau/myenv
• Use https://gitlab.com/mbarkhau/bootstrapit

v201809.0001-beta

• Initial release