lunaconf 0.4.0

lunaconf is a Python library to support easy-to-understand configuration for evaluations

lunaconf

lunaconf is a Python library to support easy-to-understand configuration for evaluations

Installation

pip install lunaconf

Usage

It provides the following interfaces:

• lunaconf.LunaConf: A base class that every configuration class should inherit from. It is a subclass of pydantic.BaseModel and thus Pydantic features can be used.

  import typing, lunaconf, pydantic
  
  class Config(lunaconf.LunaConf):
      require: int
      opt_int: int = 10
      opt_str: str = "default"
      opt_list: list[int] = pydantic.Field(default_factory=lambda: [1, 2, 3])
  
      @classmethod
      def __lunaconf_default__(cls) -> typing.Self:
          return cls(require=42)
  

  Here __lunaconf_default__ is a class method that should be overloaded if the class has required fields. It should return a default instance of the class to set default values for the fields.

• lunaconf.lunaconf_cli: Construct a configuration from the CLI.

  # example.py
  import lunaconf
  
  config = lunaconf.lunaconf_cli(Config)
  

  Some CLI arguments and their corresponding generated configurations are as follows.

  $ python3 example.py
  # Config(require=42, opt_int=10, opt_str='default', opt_list=[1, 2, 3])
  
  $ python3 example.py -J config.json    # suppose config.json contains {"require": 100}
  # Config(require=100, opt_int=10, opt_str='default', opt_list=[1, 2, 3])
  
  $ python3 example.py opt_int=233
  # Config(require=42, opt_int=233, opt_str='default', opt_list=[1, 2, 3])
  
  $ python3 example.py opt_list.0=4
  # Config(require=42, opt_int=10, opt_str='default', opt_list=[4, 2, 3])
  

  The . in the CLI modifications can be used by nested fields and list indices:

  # config
  class Inner(lunaconf.LunaConf):
      a: int = 1
  class Outer(lunaconf.LunaConf):
      inner: typing.Optional[Inner] = None
      lst: list[int] = lunaconf.Field(default_factory=lambda: [1, 2, 3])
  

  $ python3 example.py
  # Outer(inner=None, lst=[1, 2, 3])
  $ python3 example.py inner.a=10 lst.1=20
  # Outer(inner=Inner(a=10), lst=[1, 20, 3])
  

  Available command-line options:

  • command positional arguments: specify the modifications to the configuration in the form of key1.key2=value1; key3.key4=value2 etc. The . can be used to access nested fields and list indices.
  • -j <json_str> / -J <json_file>: specify the JSON to overload the configuration.
  • -t <toml_str> / -T <toml_file>: specify the TOML to overload the configuration.
  • -d <str> / -D <file>: detect the format of the string/file and parse it accordingly. It will first try to parse it as JSON, if it fails, it will try to parse it as TOML. If both fail, an error will be raised.
  • -C <file>: the extra configuration file. This file contains command line arguments (one group per line) that will be parsed interleaved with the other command line arguments. Lines starting with # are treated as comments and ignored.
  • -a: whether or not output all fields with -p / -P flags, and also affect the application of post_action_with_all or post_action_without_all callables passed to lunaconf_cli.
  • -p: print the final configuration in JSON and exit.
  • -P: print the final configuration in TOML and exit.

Special Values

The following special values can be used in the command line arguments to represent certain Python values, and are output in some cases for unsupported values in JSON/TOML:

• Input <null> leads to None; TOML will output <null> for None.
• Input <del> to delete element in an array, or reset the field to its default.
• Input <inf>, <-inf>, <nan> lead to float('inf'), float('-inf'), float('nan') respectively; JSON will output <inf>, <-inf>, <nan> for these values.

Strings inside the angle brackets are case-insensitive.

Examples

For more examples, please refer to the unit tests in the tests folder.