Add support for Literal

[pavan-uppari]

Currently when used a field with type Literal, it is failing with below traceback

from pydantic_cli import run_and_exit, Cmd
from typing import Literal


class Options(Cmd):
    input_file: Literal[1, 2]

    def run(self) -> None:
        print(f"Mock example running with {self}")


if __name__ == "__main__":
    run_and_exit(Options, description=__doc__, version="0.1.0")

parser: CustomArgumentParser = to_parser_with_overrides(custom_default_values)
                                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/workspaces/pydantic-cli/pydantic_cli/__init__.py", line 465, in to_p
    parser = pydantic_class_to_parser(
             ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/workspaces/pydantic-cli/pydantic_cli/__init__.py", line 256, in pydantic_class_to_parser
    p = _add_pydantic_class_to_parser(p0, cls, default_value_override)
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/workspaces/pydantic-cli/pydantic_cli/__init__.py", line 236, in _add_pydantic_class_to_parser
    _add_pydantic_field_to_parser(p, ix, field, override_value=override_value)
  File "/workspaces/pydantic-cli/pydantic_cli/__init__.py", line 208, in _add_pydantic_field_to_parser
    type_desc = __to_type_description(
                ^^^^^^^^^^^^^^^^^^^^^^
  File "/workspaces/pydantic-cli/pydantic_cli/__init__.py", line 113, in __to_type_description
    t = "" if field_type is NOT_PROVIDED else __try_to_pretty_type(field_type)
                                              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/workspaces/pydantic-cli/pydantic_cli/__init__.py", line 93, in __try_to_pretty_type
    name = "|".join(map(lambda x: x.__name__, args))
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/workspaces/pydantic-cli/pydantic_cli/__init__.py", line 93, in <lambda>
    name = "|".join(map(lambda x: x.__name__, args))

Literal are like Enums only, when Enums are been supported, Literal can be supported too

[pavan-uppari]

Everything works fine, only issue is with prettifying type at __try_to_pretty_type, if that method is fixed, then Literal can be supported. Need to find a better way to prettify type

[mpkocher]

I believe the fundamental solution to this is to delegate this to Pydantic and call repr on FieldInfo.

https://github.com/pydantic/pydantic/blob/main/pydantic/_internal/_repr.py#L82

There's a few notes:

1. pydantic_cli is a passthrough layer.
2. Fundamentally solves the problem with a principled approach (but it is using an internal Pydantic API)
3. Sacrifices how the info is displayed. Yields a very opinionated output.
4. Might be able to leverage rich (because Pydantic supports it)
5. Improvements to FieldInfo should probably be done within Pydantic.

For example:

from pydantic import PositiveInt, Field
from enum import Enum, auto
from pydantic_cli import run_and_exit, Cmd
from typing import Literal, Final

class State(Enum):
    a = auto()
    b = auto()


class Options(Cmd):
    input_file: Literal[1, 2]
    alpha: Final[int]
    beta: PositiveInt
    state: State = Field(..., description="The State of the Job to filter on.")
    name: None | str

    def run(self) -> None:
        print(f"Mock example running with {self}")


if __name__ == "__main__":
    run_and_exit(Options, description=__doc__, version="0.1.0")

With --help yields:

usage: example.py --input_file INPUT_FILE --alpha ALPHA --beta BETA --state STATE --name NAME [--help] [--version]

options:
  --input_file INPUT_FILE
                        FieldInfo(annotation=Literal[1, 2], required=True)
  --alpha ALPHA         FieldInfo(annotation=int, required=True, frozen=True)
  --beta BETA           FieldInfo(annotation=int, required=True, metadata=[Gt(gt=0)])
  --state STATE         FieldInfo(annotation=State, required=True, description='The State of the Job to filter on.')
  --name NAME           FieldInfo(annotation=Union[NoneType, str], required=True)
  --help                Print Help and Exit
  --version             show program's version number and exit

[mpkocher]

@pavan-uppari Thank you for your time and feedback.