with-argparse 1.0.4

A simple but handy Python library to generate a `argparse.ArgumentParser` object from a type-annotated method

with-argparse

with-argparse is a very simple and tiny package using argparse.ArgumentParser objects to derive a CLI that is automatically applied to a function using type annotations. Currently supports Python 3.10-11.

Supported features:

• Argument lists via nargs=+
• Argument choices via typing.Literal[x, y, z]
• Optional values and required arguments
• Boolean flags via presence or abscence of --argument_name
• Custom parse functions via @with_argparse(arg_name=custom_fn)
• Ignored values: @with_argparse(ignore_keys={'arg_name'}
• Nested type annotations such as list[int]
• Disabled mode: with_argparse.no_argparse() context manager

In order for this package to work, functions must receive an explicit type annotation. The following type annotations are currently supported:

• str, int, float, bool, ...
• Types that have constructors that accept a single str as input
• list[type] and set[type]
• Optional[type], type | None, Union[type, None],
• Literal[type_val1, type_val2]
• Custom types via custom parse functions (supplied via kwarg to the @with_argparse decorator.

Example code

from typing import Optional
from with_argparse import with_argparse

def custom_parse_fn(inp: str) -> int:
    return 42 if inp == "yeah" else -1

@with_argparse(
    ignore_keys={"ignored_value"}, 
    complex_input=custom_parse_fn
)
def cli(
    theory_of_everything: int,
    complex_input: int,
    ignored_value: Optional[str] = None,
) -> int:
    print(ignored_value)
    return theory_of_everything * complex_input

cli(ignored_value="abc")

will generate the following CLI output when run:

usage: --theory_of_everything THEORY_OF_EVERYTHING
                    [--complex_input COMPLEX_INPUT]

Custom parse functions

Becomes increasingly useful when the target type T does not have a default constructor with a single str argument or more complex logic is required to parse the desired type from a string input.

As it is not type correct to use functions as type annotations (it would work extracting those functions from there, however type checkers such as mypy will complain when doing so), one can specify custom parse functions directly in the @with_argparse() decorator via a keyword argument named as the target parameter.

Say we have a complex function that does call some other functions before eventually returning the string in reverse:

def func_a(a: str) -> str: ...
def func_b(b: str) -> str: ...

def custom_fn(inp: str):
    inp = func_a(inp)
    inp = func_b(inp)
    return str(reversed(inp))

Our custom function can also accept types differing from str, such as int.

We can then use this function to parse the input the following way in our dummy cli function:

@with_argparse(complex_input=custom_fn)
def cli(complex_input: str) -> str:
    return complex_input

Boolean values

For Boolean values, if the default specified is True, the CLI argument name is renamed to --no_argname, such that the user must specify to disable the given argument. In any other case (None, False), the user must specify --arg_name to set the Boolean argument to True,

The renaming of a parameter can be disabled by specifying its name in the set ignore_mapping, again in the @with_argparse decorator to the function.