Fix recursive ORM parsing error

[nuno-andre]

The ORM mode doesn't parse dictionaries if it's not explicitly called.

Change Summary

Prepend the orm_mode checking to the dict's.

Related issue number

Checklist

• Unit tests for the changes exist
• Tests pass on CI and coverage remains at 100%
• Documentation reflects the changes where applicable
• changes/<pull request or issue id>-<github username>.md file added describing change
  (see changes/README.md for details)

Example

from pydantic import BaseModel
from pydantic.utils import GetterDict

class Getter(GetterDict):
    # not relevant, just to show the behavior
    def get(self, key, default):
        key = key + key
        try:
            return self._obj[key]
        except TypeError:
            return getattr(self._obj, key, default)
        except KeyError:
            return default

class Model(BaseModel):
    class Config:
        orm_mode = True
        getter_dict = Getter

class ModelA(Model):
    a: int

class ModelB(Model):
    b: ModelA

from types import SimpleNamespace

x = dict(bb=SimpleNamespace(aa=1))
ModelB.from_orm(x)
#> b=ModelA(a=1)

y = dict(bb=dict(aa=1))
ModelB.from_orm(y)
#> ...
#> b -> a
#>   field required (type=value_error.missing)

[PrettyWood]

@nuno-andre Please update with tests and change file thanks

[nuno-andre]

@PrettyWood Should I create a new test case or would you prefer I adapt test_custom_getter_dict_derived_model_class to check this condition?

https://github.com/samuelcolvin/pydantic/blob/c489445cea40c4962ecfa9e342eb30a0ba819129/tests/test_orm_mode.py#L271-L304

[PrettyWood]

Thank you!

[lsapan]

This change actually prevents models (that are ORM-enabled) from being initialized the normal way from dictionaries. Prior to this change, the following code worked:

from types import SimpleNamespace

from pydantic import BaseModel


class User(BaseModel):
    first_name: str
    last_name: str

    class Config:
        orm_mode = True


class State(BaseModel):
    user: User

    class Config:  # (The issue happens regardless of whether the outer class has orm_mode set)
        orm_mode = True


# Pass an "orm instance"
state = State.from_orm(SimpleNamespace(user=SimpleNamespace(first_name='John', last_name='Appleseed')))

# Pass dictionary data directly
state = State(**{'user': {'first_name': 'John', 'last_name': 'Appleseed'}})

After this change, the final line throws a validation error:

Traceback (most recent call last):
  File "/Users/luke/Developer/Clients/doctorschoice/flow/orm_issue.py", line 26, in <module>
    state = State(**{'user': {'first_name': 'John', 'last_name': 'Appleseed'}})
  File "/Users/luke/.pyenv/versions/flow/lib/python3.9/site-packages/pydantic/main.py", line 330, in __init__
    raise validation_error
pydantic.error_wrappers.ValidationError: 2 validation errors for State
user -> first_name
  field required (type=value_error.missing)
user -> last_name
  field required (type=value_error.missing)

This is actually a bit of an issue given the way FastAPI handles response models. During the processing of a request, it serializes the data, and then passes it through a response model (if one was provided) to make sure the response matches the desired output schema. If you're using an ORM-mode model for convenience (like one generated from Tortoise-ORM's pydantic_model_creator), it now crashes with a validation error.

Edit: I've opened an issue about this for the sake of organization - #3181