ConfStack

Summary: This document describes a multi-layer configuration system with priority order (lowest to highest): (1) in-code defaults, (2) configuration file, (3) lowercase dotted environment variables, (4) uppercase underscored environment variables, (5) command line arguments.

Layer	Priority	Name	Quick Example
1	Lowest	In-code Defaults	key_00: str = "value"
2		Configuration File	{"key_00": "value"}
3		Lowercase Dotted Env. Vars	app_name.key_00=value
4		Uppercase Underscored Env. Vars	APP_NAME_KEY_00=value
5	Highest	Command Line Arguments	--key_00 value

Quickie: How to use confstack

Config Model:

# source: src/confstack/example00.py
import pydantic as pdt
import typing as tp
import confstack


class ConfStackExample00(confstack.ConfStack):
    app_name: tp.ClassVar[str] = "app_name"
    key_00: str = "layer_01_value_00"
    key_01: str = "layer_01_value_01"

    class Key02(pdt.BaseModel):
        subkey_01: str = "layer_01_value_02_01"
        subkey_02: str = "layer_01_value_02_02"

    key_02: Key02 = pdt.Field(default_factory=Key02)


if __name__ == "__main__":
    parser = ConfStackExample00.get_argparser()
    parser.add_argument("--extra_flag_01", action="store_true")
    parser.add_argument("--extra_args_01", type=str, default="default_value_01")
    parser.add_argument("extra_pos_args_01", nargs="?")
    config = ConfStackExample00.load_config(cli_args=parser.parse_args())
    config.print_json()

Scenario A: One key overriden by many

cat ~/.config/app_name/config.json
# {"key_00": "L2"}  # default config file

env "app_name.key_00=L3" \
APP_NAME_KEY_00="L4" \
  python -m confstack.example00 \
    --key_00 "L5"
# {
#   "key_00": "L5", # ⭐ cli arg wins (highest priority)
#   "key_01": "layer_01_value_01",
#   "key_02": {
#     "subkey_01": "layer_01_value_02_01",
#     "subkey_02": "layer_01_value_02_02"
#   },
#   "extra_flag_01": false,
#   "extra_args_01": "default_value_01",
#   "extra_pos_args_01": null
# }

Scenario B: Many keys overriden by many

cat ~/.config/app_name/config.json
# {"key_02": {"subkey_02": "L2"}}  # default config file

env "app_name.key_02.subkey_01=L3" \
APP_NAME_KEY_00="L4" \
  python -m confstack.example00 \
    --key_01 "L5" \
    --extra_args_01 "custom_value"
# {
#   "key_00": "L4", # ⭐ uppercase underscored env
#   "key_01": "L5", # ⭐ cli arg
#   "key_02": {
#     "subkey_01": "L3", # ⭐ lowercase dotted env
#     "subkey_02": "L2" # ⭐ default config file
#   },
#   "extra_args_01": "custom_value",
#   "extra_pos_args_01": null
# }

Multi-Layer Configuration Architecture

Layer 1 : In-code Defaults

import pydantic as pdt
import typing as tp


class Config(pdt.BaseModel):
    key_00: str = "layer_01_value_00"
    key_01: str = "layer_01_value_01"

    class Key02(pdt.BaseModel):
        subkey_01: str = "layer_01_value_02_01"
        subkey_02: str = "layer_01_value_02_02"
        subkey_03: str = "layer_01_value_02_03"

    key_02: Key02 = pdt.Field(default_factory=Key02)

    class Key03(pdt.BaseModel):
        class Subkey00(pdt.BaseModel):
            subsubkey_00: str = "layer_01_value_03_00_00"

        subkey_00: Subkey00 = pdt.Field(default_factory=Subkey00)

        class Subkey01(pdt.BaseModel):
            subsubkey_00: str = "layer_01_value_03_01_00"
            subsubkey_01: str = "layer_01_value_03_01_01"

        subkey_01: Subkey01 = pdt.Field(default_factory=Subkey01)

    key_03: Key03 = pdt.Field(default_factory=Key03)

Layer 2 : Configuration File

{
  "key_00": "layer_02_value_00",
  "key_01": "layer_02_value_01",
  "key_02": {
    "subkey_01": "layer_02_value_02_01",
    "subkey_02": "layer_02_value_02_02",
    "subkey_03": "layer_02_value_02_03"
  },
  "key_03": {
    "subkey_00": {
      "subsubkey_00": "layer_02_value_03_00_00"
    },
    "subkey_01": {
      "subsubkey_00": "layer_02_value_03_01_00",
      "subsubkey_01": "layer_02_value_03_01_01"
    }
  }
}

Layer 3 : Lowercase Dotted Environment Variables

env \
  app_name.key_00="layer_03_value_00" \
  app_name.key_01="layer_03_value_01" \
  \
  app_name.key_02.subkey_01="layer_03_value_02_01" \
  app_name.key_02.subkey_02="layer_03_value_02_02" \
  app_name.key_02.subkey_03="layer_03_value_02_03" \
  \
  app_name.key_03.subkey_00.subsubkey_00="layer_03_value_03_00_00" \
  app_name.key_03.subkey_01.subsubkey_00="layer_03_value_03_01_00" \
  app_name.key_03.subkey_01.subsubkey_01="layer_03_value_03_01_01" \
  \
  path/to/app_name.exe

# file_name: docker-compose.yaml
services:
  app:
    image: your_app_image
    environment:

      app_name.key_00: layer_03_value_00
      app_name.key_01: layer_03_value_01

      app_name.key_02.subkey_01: layer_03_value_02_01
      app_name.key_02.subkey_02: layer_03_value_02_02
      app_name.key_02.subkey_03: layer_03_value_02_03

      app_name.key_03.subkey_00.subsubkey_00: layer_03_value_03_00_00
      app_name.key_03.subkey_01.subsubkey_00: layer_03_value_03_01_00
      app_name.key_03.subkey_01.subsubkey_01: layer_03_value_03_01_01

    command: path/to/app_name.exe

Layer 4 : Uppercase Underscored Environment Variables

APP_NAME_KEY_00="layer_04_value_00" \
APP_NAME_KEY_01="layer_04_value_01" \
\
APP_NAME_KEY_02_SUBKEY_01="layer_04_value_02_01" \
APP_NAME_KEY_02_SUBKEY_02="layer_04_value_02_02" \
APP_NAME_KEY_02_SUBKEY_03="layer_04_value_02_03" \
\
APP_NAME_KEY_03_SUBKEY_00_SUBSUBKEY_00="layer_04_value_03_00_00" \
APP_NAME_KEY_03_SUBKEY_01_SUBSUBKEY_00="layer_04_value_03_01_00" \
APP_NAME_KEY_03_SUBKEY_01_SUBSUBKEY_01="layer_04_value_03_01_01" \
\
  path/to/app_name.exe

# file_name: docker-compose.yaml
version: '3.8'
services:
  app:
    image: your_app_image
    environment:

      APP_NAME_KEY_00: layer_04_value_00
      APP_NAME_KEY_01: layer_04_value_01

      APP_NAME_KEY_02_SUBKEY_01: layer_04_value_02_01
      APP_NAME_KEY_02_SUBKEY_02: layer_04_value_02_02
      APP_NAME_KEY_02_SUBKEY_03: layer_04_value_02_03

      APP_NAME_KEY_03_SUBKEY_00_SUBSUBKEY_00: layer_04_value_03_00_00
      APP_NAME_KEY_03_SUBKEY_01_SUBSUBKEY_00: layer_04_value_03_01_00
      APP_NAME_KEY_03_SUBKEY_01_SUBSUBKEY_01: layer_04_value_03_01_01

Layer 5 : Command Line Arguments

import argparse
parser = argparse.ArgumentParser()

parser.add_argument("--key_00", dest="key_00", default=None)
parser.add_argument("--key_01", dest="key_01", default=None)

parser.add_argument("--key_02.subkey_01", dest="key_02__subkey_01", default=None)
parser.add_argument("--key_02.subkey_02", dest="key_02__subkey_02", default=None)
parser.add_argument("--key_02.subkey_03", dest="key_02__subkey_03", default=None)

parser.add_argument("--key_03.subkey_00.subsubkey_00", dest="key_03__subkey_00__subsubkey_00", default=None)
parser.add_argument("--key_03.subkey_01.subsubkey_00", dest="key_03__subkey_01__subsubkey_00", default=None)
parser.add_argument("--key_03.subkey_01.subsubkey_01", dest="key_03__subkey_01__subsubkey_01", default=None)

args = parser.parse_args()

path/to/app_name.py \
  --key_00 layer_05_value_00 \
  --key_01 layer_05_value_01 \
  \
  --key_02.subkey_01 layer_05_value_02_01 \
  --key_02.subkey_02 layer_05_value_02_02 \
  --key_02.subkey_03 layer_05_value_02_03 \
  \
  --key_03.subkey_00.subsubkey_00 layer_05_value_03_00_00 \
  --key_03.subkey_01.subsubkey_00 layer_05_value_03_01_00 \
  --key_03.subkey_01.subsubkey_01 layer_05_value_03_01_01

Appendix: Dotted Environment Variables in Shell

# Set with `env`
env "my.var=value" ./script.sh

# Access
printenv "my.var"
python3 -c "import os; print(os.environ['my.var'])"