Fix issue for extra fields not appearing in the errors

[nofalx]

Solves Issue #835

Reference for impact of this change:
https://docs.pydantic.dev/latest/usage/validators/#model-validators

[nofalx]

This PR actually fixes another issue as well:

when u have Schema as attribute for another schema:

class UserTokenType(Schema):
    meta_data: MetaDataSchema | None = None

and you send this json

{
    "meta_data": "xxx",
}

Django ninja reports that schema is correct, however after the change in this pr we get the correct error that we missed before

{
    "error": "validation_errors",
    "details": [
        {
            "field_name": "meta_data",
            "msg": "Input should be a valid dictionary or object to extract fields from",
            "type": "model_attributes_type"
        }
    ]

[vitalik]

well it's not that easy (as you see from failed tests)

I think there must be some trick so that pydantic thinks DjangoGetter is a dict

[vitalik]

Hi @nofalx

regarding - from_orm ? did you find this somewhere on pydantic code ? I think they deprecated this keyword in favour to "from_attributes"...

[baseplate-admin]

Hi, Please use black on your code

[nofalx]

@vitalik turns out the logic does not depend on the from_attributes but more on if the source of validate if from from model_validate or somewhere else.

I dont understand the inners of pydantic very well but I have tested the update code in a django projects and all scenarios seem to be working fine for root of the json and nested json shema we using for the API. The projects tests are all passing as well

[nofalx]

Added new test cases to cover the issues this pr fixed. Code now passes lint and tests

[vitalik]

There might be also development on pydantic side for this: pydantic/pydantic#5434 will check if that covers this case as well

[vitalik]

Hi @nofalx

I simplified a bit your approach (used class.model_config to find out if default validaion wrapper is needed) - see here for more details - #863 - let me know if that works ?

[nofalx]

Hi @vitalik

Looks like a better approach, hope this makes it to next beta