django==3.0.5
pydantic==1.8.2
djantic=0.3.0

Attempting to create a djantic schema for a model that has a django.db.models.TextChoices subclass assigned to a field's choices will result in a ValueError

For example:

from django.db import models
from djantic import ModelSchema


class FoodChoices(models.TextChoices):
    BANANA = 'ba', 'A delicious yellow Banana'
    APPLE = 'ap', 'A delicious red Apple'


class A(models.Model):
    name = models.CharField(max_length=128)
    preferred_food = models.CharField(max_length=2, choices=FoodChoices.choices)


class ASchema(ModelSchema):
    class Config:
        model = A

the above results in:

...
/usr/local/lib/python3.8/site-packages/djantic/main.py in __new__(mcs, name, bases, namespace)
     95 
     96                     else:
---> 97                         python_type, pydantic_field = ModelSchemaField(field)
     98 
     99                     field_values[field_name] = (python_type, pydantic_field)

/usr/local/lib/python3.8/site-packages/djantic/fields.py in ModelSchemaField(field)
     85         if field.choices:
     86             enum_choices = {v: k for k, v in field.choices}
---> 87             python_type = Enum(  # type: ignore
     88                 f"{field.name.title().replace('_', '')}Enum",
     89                 enum_choices,

/usr/local/lib/python3.8/enum.py in __call__(cls, value, names, module, qualname, type, start)
    339             return cls.__new__(cls, value)
    340         # otherwise, functional API: we're creating a new Enum type
--> 341         return cls._create_(
    342                 value,
    343                 names,

/usr/local/lib/python3.8/enum.py in _create_(cls, class_name, names, module, qualname, type, start)
    461                 member_name, member_value = item, names[item]
    462             else:
--> 463                 member_name, member_value = item
    464             classdict[member_name] = member_value
    465         enum_class = metacls.__new__(metacls, class_name, bases, classdict)

ValueError: too many values to unpack (expected 2)

I had an issue with extraction of reverse ManyToMany relationship, and it seems like there simply was an unhandled case. I didn't really fix anything, I just copied your code from below and thus added another case. I didn't test this code with any automatic tool or unit tests, so beware. I'm rather new to python so feel free to do with it what you want. But please, if you don't pull - take at least the issue into account and do it your way :) I really appreciate your project and it saves our team a lot of time. Thank you.

Hi,

This PR will:

• Add support to include_from_annotations as shown in README and test_include_from_annotations.
• Fix the error shown in test_multiple_level_relations where a custom field (user_cache in this example) is present in the Schema.

Thank you.

This PR, separated in 2 commits, aims to:

• Add support for TextChoices fields with blank=True
• Add support for fields with null=True and default value defined

Error if I add in scheme field avatar

Model:

def get_avatar_upload_path(instance, filename):
    return os.path.join(
        "customer_avatar", str(instance.id), filename)


class Customer(models.Model):
    """
    params: user* avatar shop* balance permissions deleted

    prefetch: -
    """
    id = models.AutoField(primary_key=True)
    user = models.ForeignKey(User, on_delete=models.CASCADE, related_name='customers', verbose_name='Пользователь')
    avatar = models.ImageField(upload_to=get_avatar_upload_path, verbose_name='Аватарка', blank=True, null=True)
    retail = models.ForeignKey('retail.Retail', on_delete=models.CASCADE, related_name='customers', verbose_name='Розница')
    balance = models.DecimalField(max_digits=9, decimal_places=2, verbose_name='Баланс', default=0)
    permissions = models.ManyToManyField('retail.Permission', verbose_name='Права доступа', related_name='customers', blank=True)
    deleted = models.BooleanField(default=False, verbose_name='Удален')
    created = models.DateTimeField(auto_now_add=True, verbose_name='Дата создания')

Scheme

class CustomerSchema(ModelSchema):
    user: UserSchema

    class Config:
        model = Customer
        include = ['id', 'user', 'avatar', 'retail', 'balance', 'permissions']

Error:

fastapi_1  | Process SpawnProcess-36:
fastapi_1  | Traceback (most recent call last):
fastapi_1  |   File "/usr/local/lib/python3.8/multiprocessing/process.py", line 315, in _bootstrap
fastapi_1  |     self.run()
fastapi_1  |   File "/usr/local/lib/python3.8/multiprocessing/process.py", line 108, in run
fastapi_1  |     self._target(*self._args, **self._kwargs)
fastapi_1  |   File "/usr/local/lib/python3.8/site-packages/uvicorn/subprocess.py", line 62, in subprocess_started
fastapi_1  |     target(sockets=sockets)
fastapi_1  |   File "/usr/local/lib/python3.8/site-packages/uvicorn/main.py", line 390, in run
fastapi_1  |     loop.run_until_complete(self.serve(sockets=sockets))
fastapi_1  |   File "uvloop/loop.pyx", line 1494, in uvloop.loop.Loop.run_until_complete
fastapi_1  |   File "/usr/local/lib/python3.8/site-packages/uvicorn/main.py", line 397, in serve
fastapi_1  |     config.load()
fastapi_1  |   File "/usr/local/lib/python3.8/site-packages/uvicorn/config.py", line 278, in load
fastapi_1  |     self.loaded_app = import_from_string(self.app)
fastapi_1  |   File "/usr/local/lib/python3.8/site-packages/uvicorn/importer.py", line 20, in import_from_string
fastapi_1  |     module = importlib.import_module(module_str)
fastapi_1  |   File "/usr/local/lib/python3.8/importlib/__init__.py", line 127, in import_module
fastapi_1  |     return _bootstrap._gcd_import(name[level:], package, level)
fastapi_1  |   File "<frozen importlib._bootstrap>", line 1014, in _gcd_import
fastapi_1  |   File "<frozen importlib._bootstrap>", line 991, in _find_and_load
fastapi_1  |   File "<frozen importlib._bootstrap>", line 975, in _find_and_load_unlocked
fastapi_1  |   File "<frozen importlib._bootstrap>", line 671, in _load_unlocked
fastapi_1  |   File "<frozen importlib._bootstrap_external>", line 783, in exec_module
fastapi_1  |   File "<frozen importlib._bootstrap>", line 219, in _call_with_frames_removed
fastapi_1  |   File "/opt/project/config/asgi.py", line 18, in <module>
fastapi_1  |     from backend.retail.fastapp import app
fastapi_1  |   File "/opt/project/backend/retail/fastapp.py", line 9, in <module>
fastapi_1  |     from backend.retail.routers import router
fastapi_1  |   File "/opt/project/backend/retail/routers.py", line 3, in <module>
fastapi_1  |     from backend.retail.api import auth, postitem, ping, customer
fastapi_1  |   File "/opt/project/backend/retail/api/customer.py", line 8, in <module>
fastapi_1  |     from backend.retail.schemas.customer import CustomerSchema
fastapi_1  |   File "/opt/project/backend/retail/schemas/customer.py", line 7, in <module>
fastapi_1  |     class CustomerSchema(ModelSchema):
fastapi_1  |   File "/usr/local/lib/python3.8/site-packages/djantic/main.py", line 97, in __new__
fastapi_1  |     python_type, pydantic_field = ModelSchemaField(field)
fastapi_1  |   File "/usr/local/lib/python3.8/site-packages/djantic/fields.py", line 135, in ModelSchemaField
fastapi_1  |     python_type,
fastapi_1  | UnboundLocalError: local variable 'python_type' referenced before assignment

Hello,

This is a great library. Looking to contribute if you are needing anything done.

Getting this error but I see it looks fixed in your main branch.

  File "\.venv\lib\site-packages\djantic\main.py", line 102, in __new__
    p_model = create_model(
  File "pydantic\main.py", line 990, in pydantic.main.create_model
  File ".venv\lib\site-packages\djantic\main.py", line 34, in __new__
    cls = super().__new__(mcs, name, bases, namespace)
  File "pydantic\main.py", line 299, in pydantic.main.ModelMetaclass.__new__
  File "pydantic\fields.py", line 403, in pydantic.fields.ModelField.infer
  File "pydantic\fields.py", line 388, in pydantic.fields.ModelField._get_field_info
  File "pydantic\fields.py", line 173, in pydantic.fields.FieldInfo._validate
ValueError: cannot specify both default and default_factory

Django==3.2.3
djantic==0.2.0
pydantic==1.8.2

So just posting this for tracking incase anyone else has the same issue.

Hi,

I open this PR/discussion to:

• Add support for multiple level relationships (user -> user.orders -> user.orders.order_items -> user.orders.order_items.order_items_details ...)
• Add support for double underscore aliases (first_name: str = Field(alias="user__first_name")). We can replace it by dot aliases if you prefer (first_name: str = Field(alias="user.first_name"))
• Simplify from_django logic by relying always that possible in Pydantic, this will increase Djantic performance (reducing serialization time).
• Fix ManyToMany relations serialization: ManyToMany relations should access related field by related_name or by *_set and should return a list of instances (from a QuerySet) instead of a instance.
• Harmonize ID/PK usage: instead of returning pk for some fields and id for other, return id for all.
• Calling XSchema.from_django(xs, many=True) should return a list of XSchema instances instead of a list of dicts.
• Include GenericForeignKey in schemas.

Merging this request will cause some breaking changes:

• ManyToMany relations will return a list of instances instead of a instance. The attribute will be called *_set or the Django's field related_name.
• ForeignObjectRel will return a list of dicts with pk instead of id: [{'id': 1}] instead of [{'pk': 1}].
• XSchema.from_django(xs, many=True) will return a list of XSchema instances instead of a list of dicts.

These changes can be seem in test_relations.py and test_queries.py.

I din't separated this PR in several commits as these changes are linked, but I can separate it if it's hard to understand.

Thank you.

Given:

class LabelSchema(ModelSchema):
    class Config:
        model = OrderItemLabel
        include = ['value']

class OrderItemSchema(ModelSchema):
    label: LabelSchema | None
    quantity: int

    class Config:
        model = OrderItem
        include = ['label', 'quantity']

class CustomerSchema(ModelSchema):
    class Config:
        model = Customer
        include = ['username']

class OrderSchema(ModelSchema):
    items: list[OrderItemSchema]
    customer: CustomerSchema
    
    class Config:
        model = Order
        include = ['items', 'customer']

When calling:

OrderSchema.from_django(order).dict()

Will return:

 {
    'items': [
        {'quantity': 1, 'label': None},
        {'quantity': 1, 'label': None},
        {'quantity': 1, 'label': 1}
    ],
    'customer': {'username': '[email protected]'}
}

Note that label = 1, the ID and not the value (str) expected from LabelSchema.

With this PR change, the new output is:

{
    'items': [
        {'quantity': 1, 'label': None},
        {'quantity': 1, 'label': None},
        {'quantity': 1, 'label': {'value': 'Product 10'}}
    ],
     'customer': {'username': '[email protected]'}}

Gave the code a try, went to open a PR for BigAutoField django 3.2 changes, but noticed they were in the main branch already.

Tried install from pypi 0.2.0 and also looked in the tagged zip and didn't see the changes there.

updated my requirements.txt

from

djantic==0.2.0

to:

-e git+https://github.com/jordaneremieff/[email protected]#egg=djantic

Was there some other stuff being worked on for django 3.2 support?

Thanks

class Level1(models.Mode): name = models.CharField()

class Level2(models.Model): parent = models.ForeignKey(Level1) name2 = models.Charge()

class Level3(models.Model): parent = models.ForeignKey(Level2) name3 = models.Charge()

class Level3Schema(ModelSchema): name3: str = Field()

class Level2Schema(ModelSchema): level3_set:List[Level3Schema] = Field(None) name2: str = Field(None)

class Level1Schema(ModelSchema): level2_set :List[Level2Schema] = Field(None) name1:str = Field(None)

   as following is not work:
 Level1Schema.from_django(level1_instance)

HI,

First of all, this is a great project, getting a lot of value from it, so thank you.

Having an issue generating an openapi specification when using it in combination with FastAPI, specifically when using Django model choices.

Django Model

from django.db import models

class SimpleModelWithChoices(models.Model):
    class ChoiceType(models.TextChoices):
        CHOICE1 = "CHOICE1"
        CHOICE2 = "CHOICE2"

    choice = models.CharField("Type", max_length=30, choices=ChoiceType.choices)

Djantic Spec

from djantic import ModelSchema

class Schema1(ModelSchema):
    class Config:
        model = SimpleModelWithChoices


class Schema2(ModelSchema):
    class Config:
        model = SimpleModelWithChoices

FastAPI

from fastapi import APIRouter, FastAPI

router = APIRouter()
application = FastAPI()

@application.get("path1", response_model=Schema1)
def get1(instance: Schema1):
    return instance

@application.get("path2", response_model=Schema2)
def get2(instance: Schema2):
    return instance

application.include_router(router)

assert application.openapi()

Error

.venv\lib\site-packages\fastapi\openapi\utils.py:387: in get_openapi
    definitions = get_model_definitions(

    def get_model_definitions(
        flat_models: Set[Union[Type[BaseModel], Type[Enum]]],
        model_name_map: Dict[Union[Type[BaseModel], Type[Enum]], str],
    ) -> Dict[str, Any]:
        definitions: Dict[str, Dict[str, Any]] = {}
        for model in flat_models:
            m_schema, m_definitions, m_nested_models = model_process_schema(
                model, model_name_map=model_name_map, ref_prefix=REF_PREFIX
            )
            definitions.update(m_definitions)
>           model_name = model_name_map[model]
E           KeyError: <enum 'ChoiceEnum'>

.venv\lib\site-packages\fastapi\utils.py:28: KeyError

I think it is due to the way FastAPI keeps a mapping of classes and uses them internally. I have a workaround in my project

def __fix_enums(new_type):

    for item in new_type.__dict__["__fields__"].values():
        if "enum.EnumMeta" in str(item.type_.__class__):
            enum_values = [(i.value, i.value) for i in item.type_]

            new_enum_type = Enum(f"{new_type.__name__}_{item.name}_Enum", enum_values, module=__name__)
            setattr(item, "type_", new_enum_type)

    return new_type

I would be happy to produce a PR to fix the issue if needed, Maybe adding some sort of uniqueness in the name of the enum class in djantic/fields.py ?

            python_type = Enum(  # type: ignore
                f"{field.name.title().replace('_', '')}Enum",
                enum_choices,
                module=__name__,
            )

Library Versions

djantic = "^0.3.5"
Django = "^3.2.5"
fastapi = "^0.68.0"

Adding the option to implement custom methods in abstract ModelSchema subclasses that cannot be instantiated and thus they don't have to undergo the strict checks that regular ModelSchemata need to fulfill.

Closes !62

I am using Djantic in my project, and I would like to base a base class between ModelSchema and my individual model schemata, to implement a to_orm method directly on the ModelSchema, as in my complex case simply doing DjangoModel(**schema) doesn't work because of foreign keys and some other data transformations required. I could just implement it on each of them individually, but then mypy will be very unhappy.

I would like to be able to do the following:

class ModelSchemaToORM(ModelSchema):
    # either this, django model style
    class Meta:
        abstract = True

    # or this
    class Config:
        abstract = True

    async def to_orm(self):
        raise NotImplementedError

Currently the two versions above yield

pydantic.errors.ConfigError: 'Config' (Is `Config` class defined?)

and

pydantic.errors.ConfigError: type object 'Config' has no attribute 'model' (Is `Config.model` a valid Django model class?)

respectively.

Does it make sense to folks here to add an additional condition to ModelSchemaMetaclass to skip Config checks on abstract subclasses? It could either be in Meta or Config, I don't really mind as long as there is a way to put intermediate parents between my ModelSchemata and Djantic's ModelSchema.

Is MyPy type checking supported at all? Any model property I try to access on a Djantic model is detected as non-existent by mypy. Is this resolvable?

Maybe I can contribute the solution if someone can point me in the right direction. Thank you.

The approach in https://github.com/tiangolo/sqlmodel has me interested exploring a combined model for Django and Pydantic. I would consider doing a PR to that project, though it seems too specific to SQLAlchemy and FastAPI for a Django-specific implementation to be a good fit (also don't know if I'll pursue this at all) - not sure.

Something like this for the model definition:

from djantic.models import DjanticModel, Field

class Plant(DjanticModel):
    """A plant."""

    common_name: str
    genus_name: str = Field(
        max_length=100,
        help_text="A generic name for a collective group of plants.",
    )
    specific_epithet: str

    class Meta:
        ordering = ["genus_name"]

    class Config:
        include = ["genus_name", "specific_epithet"]

Then attribute access would (ideally) look like this:

from myapp.models import Plant

# Schema definition from class
schema = Plant.schema()

# Export data from instance
plant = Plant.objects.create(genus_name="Grevillea", specific_epithet="buxifolia")
data = plant.dict()

Maybe.

I didn't include any docstrings or comments throughout the project because things were changing quite rapidly, but at this point I should go through and add these. Also a mkdocs gh-page and changelog.