Python configuration utilities

Implementation of key-value pair based configuration for Python applications.

Features:

• support for most common sources of application settings
• support for overriding settings in sequence
• support for nested structures and lists, using attribute notation
• strategy to use environment specific settings
• features to handle secrets and values stored in the user folder, for local development
• features to support validation of configuration items, for example using pydantic, or user defined classes

This library is freely inspired by .NET Core Microsoft.Extensions.Configuration (ref. MSDN documentation, Microsoft Extensions Configuration Deep Dive).

The main class is influenced by Luciano Ramalho`s example of JSON structure explorer using attribute notation, in his book Fluent Python.

Overview

essentials-configuration provides a way to handle configuration roots composed of several layers, such as configuration files and environment variables. Layers are applied in order and can override each others' values, enabling different scenarios like configuration by environment and system instance.

Supported sources:

• toml files
• yaml files
• json files
• ini files
• environment variables
• secrets stored in the user folder, for development purpose
• dictionaries
• keys and values
• Azure Key Vault, using essentials-configuration-keyvault
• custom sources, implementing the ConfigurationSource interface

Installation

pip install essentials-configuration

To install with support for YAML configuration files:

pip install essentials-configuration[yaml]

To install with support for YAML configuration files and the CLI to handle user secrets:

pip install essentials-configuration[full]

Extensions

• Azure Key Vault secrets configuration source: essentials-configuration-keyvault

Examples

Please read the list of examples in the examples folder. Below are reported some of the examples that are tested in this repository.

TOML file

from config.common import ConfigurationBuilder
from config.env import EnvVars
from config.toml import TOMLFile


builder = ConfigurationBuilder(
    TOMLFile("settings.toml"),
    EnvVars(prefix="APP_")
)

config = builder.build()

For example, if the TOML file contains the following contents:

title = "TOML Example"

[owner]
name = "Tom Preston-Werner"

And the environment has a variable such as APP_OWNER__NAME=AAA, the owner name from the TOML file gets overridden by the env variable:

>>> config
<Configuration {'title': '...', 'owner': '...'}>
>>> config.title
'TOML Example'
>>> config.owner.name
'AAA'

JSON file and environment variables

In the following example, configuration values will include the structure inside the file settings.json and environment variables whose name starts with "APP_". Settings are applied in order, so environment variables with matching name override values from the json file.

from config.common import ConfigurationBuilder
from config.json import JSONFile
from config.env import EnvVars

builder = ConfigurationBuilder(
    JSONFile("settings.json"),
    EnvVars(prefix="APP_")
)

config = builder.build()

For example, if the JSON file contains the following contents:

{
    "logging": {
        "level": "INFO"
    },
    "example": "Hello World",
    "foo": "foo"
}

And the environment has a variable named APP_foo=AAA:

>>> config
<Configuration {'logging': '...', 'example': '...', 'foo': '...'}>
>>> config.foo
'AAA'
>>> config.logging.level
'INFO'

YAML file and environment variables

In this example, configuration will include anything inside a file settings.yaml and environment variables. Settings are applied in order, so environment variables with matching name override values from the yaml file (using the yaml source requires also PyYAML package).

from config.common import ConfigurationBuilder
from config.env import EnvVars
from config.yaml import YAMLFile

builder = ConfigurationBuilder()

builder.add_source(YAMLFile("settings.yaml"))
builder.add_source(EnvVars())

config = builder.build()

YAML file, optional file by environment

In this example, if an environment variable with name APP_ENVIRONMENT and value dev exists, and a configuration file with name settings.dev.yaml is present, it is read to override values configured in settings.yaml file.

import os

from config.common import ConfigurationBuilder
from config.env import EnvVars
from config.yaml import YAMLFile

environment_name = os.environ["APP_ENVIRONMENT"]

builder = ConfigurationBuilder(
    YAMLFile("settings.yaml"),
    YAMLFile(f"settings.{environment_name}.yaml", optional=True)
)

config = builder.build()

Filtering environment variables by prefix

from config.common import ConfigurationBuilder
from config.env import EnvVars

builder = ConfigurationBuilder()

builder.add_source(EnvVars(prefix="APP_"))

config = builder.build()

INI files

INI files are parsed using the built-in configparser module, therefore support [DEFAULT] section; all values are kept as strings.

from config.common import ConfigurationBuilder
from config.ini import INIFile

builder = ConfigurationBuilder()

builder.add_source(INIFile("settings.ini"))

config = builder.build()

Dictionaries

from config.common import ConfigurationBuilder

builder = ConfigurationBuilder()

builder.add_map({"host": "localhost", "port": 8080})

builder.add_map({"hello": "world", "example": [{"id": 1}, {"id": 2}]})

config = builder.build()

assert config.host == "localhost"
assert config.port == 8080
assert config.hello == "world"
assert config.example[0].id == 1
assert config.example[1].id == 2

Keys and values

from config.common import ConfigurationBuilder

builder = ConfigurationBuilder()

builder.add_map({"host": "localhost", "port": 8080})

builder.add_value("port", 44555)

config = builder.build()

assert config.host == "localhost"
assert config.port == 44555

User secrets

The library provides a strategy to handle secrets during local development, storing them into the user folder.

The following example shows how secrets can be configured for a project:

config settings init
config settings set "Foo" "Some secret value"

Secrets are organized by project, and the project information is obtained from pyproject.toml files (from the project.name property). If pyproject.toml file does not exist, one is generated automatically with a random name.

---

Then, from a Python app, it's possible to load the secrets from the user folder:

from config.common import ConfigurationBuilder
from config.json import JSONFile
from config.secrets import UserSecrets

builder = ConfigurationBuilder(JSONFile("settings.json"), UserSecrets())

config = builder.build()

print(config)
# config contains both values from `settings.json`, and secrets read from the user
# folder

Secrets are optional and should be used only for local development, they are stored in unencrypted form in the user's folder.

Production apps should use dedicated services to handle secrets, like Azure Key Vault, AWS Secrets Manager, or similar services. For Azure Key Vault, an implementation is provided in essentials-configuration-keyvault.

Handling user settings

User settings (stored in the user's folder) can be handled using the provided config CLI.

These settings can be useful to store secrets and other values during local development, or in general when working with desktop applications.

Overriding nested values

It is possible to override nested values by environment variables or dictionary keys using the following notation for sub properties:

• keys separated by colon ":", such as a:d:e
• keys separated by "__", such as a__d__e

from config.common import ConfigurationBuilder, MapSource


builder = ConfigurationBuilder(
    MapSource(
        {
            "a": {
                "b": 1,
                "c": 2,
                "d": {
                    "e": 3,
                    "f": 4,
                },
            }
        }
    )
)

config = builder.build()

assert config.a.b == 1
assert config.a.d.e == 3
assert config.a.d.f == 4

builder.add_value("a:d:e", 5)

config = builder.build()

assert config.a.d.e == 5
assert config.a.d.f == 4

Overriding nested values using env variables

import os

from config.common import ConfigurationBuilder, MapSource
from config.env import EnvVars

builder = ConfigurationBuilder(
    MapSource(
        {
            "a": {
                "b": 1,
                "c": 2,
                "d": {
                    "e": 3,
                    "f": 4,
                },
            }
        }
    )
)

config = builder.build()

assert config.a.b == 1
assert config.a.d.e == 3
assert config.a.d.f == 4

# NB: if an env variable such as:
# a:d:e=5
# or...
# a__d__e=5
#
# is defined, it overrides the value  from the dictionary

os.environ["a__d__e"] = "5"

builder.sources.append(EnvVars())

config = builder.build()

assert config.a.d.e == "5"

Overriding values in list items using env variables

import os

from config.common import ConfigurationBuilder, MapSource
from config.env import EnvVars

builder = ConfigurationBuilder(
    MapSource(
        {
            "b2c": [
                {"tenant": "1"},
                {"tenant": "2"},
                {"tenant": "3"},
            ]
        }
    ),
    EnvVars(),
)

os.environ["b2c__0__tenant"] = "5"

config = builder.build()

assert config.b2c[0].tenant == "5"
assert config.b2c[1].tenant == "2"
assert config.b2c[2].tenant == "3"

Typed config

To bind configuration sections with types checking, for example to use pydantic to validate application settings, use the config.bind method like in the following example:

# example-01.yaml
foo:
  value: "foo"
  x: 100

# example
from pydantic import BaseModel

from config.common import ConfigurationBuilder
from config.yaml import YAMLFile


class FooSettings(BaseModel):
    value: str
    x: int


builder = ConfigurationBuilder(YAMLFile("example-01.yaml"))

config = builder.build()

# the bind method accepts a variable number of fragments to
# obtain the configuration section that should be used to instantiate the given type
foo_settings = config.bind(FooSettings, "foo")

assert isinstance(foo_settings, FooSettings)
assert foo_settings.value == "foo"
assert foo_settings.x == 100

Goal and non-goals

The goal of this package is to provide a way to handle configuration roots, fetching and composing settings from different sources, usually happening once at application's start.

The library implements only a synchronous API and fetching of application settings atomically (it doesn't support generators), like application settings fetched from INI, JSON, or YAML files that are read once in memory entirely. An asynchronous API is currently out of the scope of this library, since its primary use case is to fetch configuration values once at application's start.