Hi!
I am using ninja_extra.ModelConfig class in order to generate an API for a Django model which has a models.JSONField field. The converter internally used by ninja_schema to get a Pydantic field from the Django field is:
# from ninja_schema.orm.utils.converter.py
@t.no_type_check
@convert_django_field.register(models.JSONField)
def convert_field_to_json_string(
field: Field, **kwargs: DictStrAny
) -> t.Tuple[t.Type, PydanticField]:
python_type = Json
if field.null:
python_type = t.Optional[Json]
return construct_field_info(python_type, field)Unfortunately, the Json Pydantic field expects serialized data as input. When reading an object with a models.JSONField field from the database, Django retrieves a dict (not string, bytes or bytearray) for this field. This leads to this validation error from Pydantic:
JSON input should be string, bytes or bytearray [type=json_type, input_value={'values': [1], 'numbers': []}, input_type=dict]
I think that the Pydantic Json field is perfect for a create_schema or update_schema as their data comes from the HTTP request. But, in the case of a retrieve_schema, the python_type should be a dict (not Json) as there is no conversion to perform.
NB: I have manually defined a retrieve_schema using the dict field to bypass this problem.
Hi!
I am using
ninja_extra.ModelConfigclass in order to generate an API for a Django model which has amodels.JSONFieldfield. The converter internally used byninja_schemato get a Pydantic field from the Django field is:Unfortunately, the
JsonPydantic field expects serialized data as input. When reading an object with amodels.JSONFieldfield from the database, Django retrieves adict(not string, bytes or bytearray) for this field. This leads to this validation error from Pydantic:I think that the Pydantic
Jsonfield is perfect for acreate_schemaorupdate_schemaas their data comes from the HTTP request. But, in the case of aretrieve_schema, thepython_typeshould be adict(notJson) as there is no conversion to perform.NB: I have manually defined a
retrieve_schemausing thedictfield to bypass this problem.