NameError, what is the logic behind checking against protected namespaces?

[nss350]

Hi,
I've was writing a class using pydantic which has an attribute named "params_". When I try and build documentation for the class using autodoc pydantic, the following error gets raised in pydantic._internal._fields.collect_model_fields.

NameError: Field "model_fields" conflicts with member {'params_': FieldInfo(annotation=ndarray, required=False)} of protected namespace "model_".

From reading about protected namespaces, I don't get the impression that the attribute name "params_" should flag this error, but putting this here as it may not be a bug.

Any suggestions appreciated.

[guibos]

@nss350 My case: it is produced when I try to inherit from a parent class. Parent class does not raise the problem.

[MSSputnik]

Today I debugged the error I ignored for a while.
It looks like that due to the inheritance the field model_field get set in the parent class. Maybe it is set when sphinx or autodoc processes the parent class. I did not find out.
My work around was to import the problematic module in conftest.py, not the class.
Sphinx is doing "importlib.import_module()" and with this, the module is already loaded and not imported again, not causing the problematic check from pydantic.

[samanthavbarron]

Could you please elaborate on the workaround that you found? I am not sure I understand which module you're referring to.

[swinkels]

I faced the same issue. autodoc raised the NameError when it tried to import a module (I wrote) that referenced a class that 1) inherited from the Pydantic BaseModel and 2) used generics. I worked around the NameError by explicitly importing that module in conf.py, so in the Sphinx configuration file.

I looked into the underlying cause and got to the same point as mentioned in the comment above. By then you're deep into Pydantic internals and in the end I just settled for the workaround.

[FredStober]

I have the same issue - but I'm not sure where to open a bug report about this - sphinx or pydantic....
It started happening when changing to pydantic 2 ... But it's happening in so many places in my code base that it's just not feasible to import everything in conf.py.

[Yalchin403]

Thanks @swinkels. I think importing the inherited schemas are the easiest way to get around this issue. But i think this is something to be resolved on the pydantic end.

[swinkels]

Thank you for the shoutout, but the workaround is not my idea. Credit should go to @MSSputnik for that one.