This config file…

input_dir: data
logging:
    file: output.log
    level: 4

… together with this main.py …

import yaml
import quickargs

with open("config.yaml") as f:
    config = yaml.load(f, Loader=quickargs.YAMLArgsLoader)

… will give you this command line interface

usage: main.py [-h] [--input_dir INPUT_DIR] [--logging.file LOGGING.FILE]
               [--logging.level LOGGING.LEVEL]

optional arguments:
  -h, --help            show this help message and exit
  --input_dir INPUT_DIR
                        default: data
  --logging.file LOGGING.FILE
                        default: output.log
  --logging.level LOGGING.LEVEL
                        default: 4

Override settings using the command line

python main.py --logging.file=other_log.txt

You get your merged yaml + command line parameters in a convenient dictionary

# exact same output format as normal yaml.load would produce
{'input_dir': 'data', 'logging': {'file': 'other_log.txt', 'level': 4}}

The types used in the yaml file are automatically enforced

Setting the log-level to a string instead of an int: python main.py --logging.level=WARNING

usage: main.py [-h] [--input_dir INPUT_DIR] [--logging.file LOGGING.FILE]
               [--logging.level LOGGING.LEVEL]
main.py: error: argument --logging.level: invalid int value: 'WARNING'

Setting the log-level to the correct type: python main.py --logging.level=0

{'input_dir': 'data', 'logging': {'file': 'output.log', 'level': 0}}

Load the yaml config and parse command line arguments

main.py

import yaml
import quickargs

with open("config.yaml") as f:
    config = yaml.load(f, Loader=quickargs.YAMLArgsLoader)

Deeply nested arguments are no problem

config.yaml

key1:
  key2:
    key3:
      key4: value

Override nested argument using dot notation: python main.py --key1.key2.key3.key4=other_value

{'key1': {'key2': {'key3': {'key4': 'other_value'}}}}

Of course it is fine to just call your program without any command line arguments

Happy with the default values in config file: python main.py

{'key1': {'key2': {'key3': {'key4': 'value'}}}}

Most yaml types, including sequences are supported

config.yaml

thresholds: [0.2, 0.4, 0.6, 0.8, 1.0]

Override the thresholds: python main.py --thresholds='[0.0, 0.5, 1.0]'

(take care to use ‘ ‘ around your command line arguments if they include spaces)

{'thresholds': [0.0, 0.5, 1.0]}

However, types within sequences are not enforced

config.yaml

thresholds: [0.2, 0.4, 0.6, 0.8, 1.0]

List of strings instead of list of floats does not give an error: python main.py --thresholds=[a,b,c]

{'thresholds': ['a', 'b', 'c']}

You can even pass references to functions or classes (your own or builtins)

config.yaml

function_to_call: !!python/name:yaml.dump

Override with reference to built-in zip function: python main.py --function_to_call=zip

{'function_to_call': <built-in function zip>}

Types

Following types are not supported at all:

• !!python/dict (because it looks just like the rest of the yaml file)

• !!pairs

Following types are not enforced / objects will not be instantiated:

• !!python/object

• !!python/object/new

• !!python/object/apply

Multi-document loading

If the YAML file contains multiple documents, only the first document will be considered. The yaml.load_all functionality is not supported.