Feature request: Positional arguments from pydantic model

[ukch]

According to the documentation, the way to make positional arguments in TAP is to define a configure method and call self.add_argument("argument_name"). However, that does not work with a pydantic model (using either tapify or to_tap_class) as the TAP class is auto-generated without the option to customise.

Here are a few things I have tried:

1. Add configure method to the pydantic model:

class MyModel(pydantic.BaseModel):
    foo: str

    def configure(self):
        assert False  # this is never called

tapify(MyModel)  # the following arguments are required: foo

2. Call add_argument manually:

class MyModel(pydantic.BaseModel):
    foo: str

parser = to_tap_class(MyModel)()
parser.add_argument("foo")  # Raises "ValueError: add_argument cannot be called after initialization."

3. Subclass workaround

class MyModel(pydantic.BaseModel):
    foo: str

class MyModelTap(to_tap_class(MyModel)):  # type: ignore
    def configure(self):
        self.add_argument("foo")

parser = MyModelTap()
parser.parse_args(argv)  # this does work, but breaks typing

Please can support for this use-case be added to the library?

[kddubey]

thank you for opening this issue, @ukch

(3) is the intended pattern. i'll add something to the docs to make this clearer

indeed, unfortunately, the type inference system will not know that args has foo b/c MyModelTap is a dynamic subclass. to "skip" this problem, does it help to do something like this?

args = MyModelTap().parse_args()
my_model = MyModel(**args.as_dict())

maybe it's worth revisiting this pattern tho. it does feel repetitive