Custom Fields not recognised by PyCharm

[avlahop]

I have the following custom Field:

class NonEmptyString(str):
     @classmethod
    def __get_validators__(cls):
        yield cls.validate

    @classmethod
    def validate(cls, v):
        if v == "":
            raise ValueError("Non-empty string expected")
        return v

and a model using it:

class MyModel(BaseModel):
    field: NonEmptyString

but when I instantiate MyModel:

my_model = MyModel(field="a string")

I get the following warning from pycharm's linter:

Expected type 'NonEmptyString' got 'str' instead

Is this behaviour expected?

Versions
pydantic 1.1.1
pycharm-pydantic-plugin: 0.0.27

[dmontagu]

I think this is expected; the only alternative would be to drop the type-checking for the __init__ entirely.

This has been discussed before, but I think it makes sense to make this a plugin config setting. Alternatively, the __init__ types could all be Any, and the plugin could generate a typed signature for BaseModel.construct, but that won't apply any of the validators, so anywhere you are using __init__ for meaningful logic, you'd lose type hints.

@koxudaxi Maybe we could add a pycharm-plugin-specific model Config setting (wouldn't need to be an official part of pydantic at all) that would let you enable/disable a typed init on a model-by-model basis? Something like pycharm_typed_init = False (or True)? That might make it easier to selectively disable warnings.

It would be very easy to add similar behavior to the mypy plugin...

[koxudaxi]

@avlahop

Is this behaviour expected?

Yes, We expect the behavior.
@dmontagu explained about the behavior and history.

We should add a panel of settings to control the warnings.
Sorry, I have not tried to implement the feature yet.
I will try it next week.

@dmontagu
Your idea looks good. But I try to implement the panel of settings.
If I can not do it next week, then let's discuss the option on `config.

[otsuka]

FYI,
pydantic EmailStr has a bug.

EmailStr type is used in a model,

addr = Address(email='foo@example.com')

the plugin raises a warning on editor.

addr = Address(email=EmailStr('foo@example.com'))

This code is type checking safe, but running it on Linux raises ValidationError.
pydantic/pydantic#1070 (comment)

[koxudaxi]

@avlahop

I write a workaround.
I add a new type-checker in the plugin.
You can set parsable-type and acceptable-type. there are details in this document page

If you want to remove warning then, you should change acceptable-type-highlight to "disable"

pyproject.toml:

[tool.pydantic-pycharm-plugin]
# You can select higlith level  from "warning",  "weak_warning", "disable"
parsable-type-highlight = "warning"
acceptable-type-highlight = "weak_warning"
[tool.pydantic-pycharm-plugin.acceptable-types]
"main.NonEmptyString" = ["str"]

[koxudaxi]

I think the workaround is good.
If someone would discuss the issue then, I will re-open the issue.