typedparser

Typing extension for python argparse using attrs.

Includes typechecking and conversion utilities to parse a dictionary into an attrs instance.

Install

Requires python>=3.8

pip install typedparser

Basic usage

1. Create an attrs class (decorate with @attr.define). Note that optional arguments must also be typed as optional.
2. Define and type the fields with typedparser.add_argument - the syntax extends add_argument from argparse.
3. Parse the args with TypedParser and enjoy args with type hints. Disable typechecking by setting strict=False.

from typing import Optional
from attrs import define
from typedparser import add_argument, TypedParser


@define
class Args:
    # omit the argument name to have it inferred from the field name
    foo: str = add_argument(positional=True)
    bar: int = add_argument(shortcut="-b", type=int, default=0)
    opt: Optional[str] = add_argument()

    # # in case you prefer the regular argparse syntax:
    # foo: str = add_argument("foo")
    # bar: int = add_argument("-b", "--bar", type=int, default=0)
    # opt: Optional[str] = add_argument("--opt")



def main():
    parser = TypedParser.create_parser(Args, strict=True)
    args: Args = parser.parse_args()
    print(args)


if __name__ == "__main__":
    main()

Features

• Create commandline arguments with type hints and checks while staying close to the syntax of the standard library's argparse.
• Utilities for typechecking and converting nested objects:
  • Nested checking and conversion of python standard types
  • Supports old and new style typing (e.g. typing.List and list)
  • Supports positional and keyword arguments in classes
  • Can also typecheck existing attrs instances
  • Allows custom conversions, by default converts source type str to target type Path and int to float
  • Allows to redefine which objects will be recursed into, by default recurses into standard containers (list, dict, etc.)
  • @definenumpy decorator for equality check if the instances contains numpy arrays
• Some object utilities in typedparser.objects required for everything else

Advanced usage

• Use TypedParser.from_parser(parser, Args) to add typing to an existing parser. This is useful to cover usecases like subparsers or argument groups.
• Snippet for argument lists xarg: List[int] = add_argument(shortcut="-x", type=int, action="append", help="Xarg", default=[]), use as -x 1 -x 2 to get [1, 2] in the args instance.

Usage of attr utilities

Define the class hierarchy and parse the input using attrs_from_dict. Use @define(slots=False) to allow multiple inheritance and setting attributes later.

from attrs import define
from typing import Optional
from typedparser import attrs_from_dict

@define
class Cfg:
    foo: int = 12
    bar: Optional[int] = None

print(attrs_from_dict(Cfg, {"foo": 1, "bar": 2}))
# Cfg(foo=1, bar=2)


@define
class CfgNested:
    sub_cfg: Cfg = None

print(attrs_from_dict(CfgNested, {"sub_cfg": {"foo": 1, "bar": 2}}))
# CfgNested(sub_cfg=Cfg(foo=1, bar=2))

Strict mode (default)

• Convert everything to the target type, e.g. if the input is a list and the annotation is a tuple, the output will be a tuple
• Raise errors if types cannot be matched, there are unknown fields in the input or abstract annotation types are used (e.g. Sequence)
• Set _allow_extra_keys = True in the class definition to allow unknown fields in the input

Non-strict mode

Enabled by calling attrs_from_dict with strict=False

• No conversion except for creating the attrs instance from the dict
• Ignore silently if types cannot be matched or abstract annotation types are used
• Unknown fields in the input will be added to the attrs instance if possible (see the hint below about slots)

Skip unknowns

Set skip_unknowns=True to ignore all unknown input fields.

Hints

The following behaviour stems from the attrs package:

• New attributes cannot to be added after class definition to an attrs instance, unless it is created with @define(slots=False) Explanation
• Untyped fields or "ClassVar" typed fields will be ignored by @attrs.define and therefore also by this library.

Install locally and run tests

Clone repository and cd into. Setup python 3.7 or higher. Note: Some tests are skipped for python 3.7.

pip install -e .
pip install pytest pytest-cov pylint
pylint typedparser

# run tests
python -m pytest --cov
pylint tests