Dataclass to be CLI and logger compatible

This package does only two things:

1. it takes a dataclass and converts it to a command line str.
2. it takes a dataclass and converts it to a loggable dict for 3rd party logging like MLFlow. Contrary to asdict it also converts nested dataclasses by separting their attributes by prefix.

By inheriting from the class ReverseCli you can use the to_command_string method to get the command line str and the property loggable_dict to get the loggable dict.

Example

@dataclass(kw_only=True)
class Parameters(ReverseCli):
 a: int
 b: str
 c: bool
 d: list[int]
@dataclass(kw_only=True)
class Nested(ReverseCli):
 a: int
 b: Parameters
parameters = Nested(a=1, b=Parameters(a=1, b="2", c=True, d=[3, 4]))
print(parameters.to_command_string())
# --a 1 --b.a 1 --b.b 2 --b.c --b.d 3 --b.d 4
print(parameters.loggable_dict)
# {'a': 1, 'b.a': 1, 'b.b': '2', 'b.c': True, 'b.d': [3, 4]}

Installation

We use UV for installation and RUFF for formating and linting.