Schemas from Django models

Schemas are very useful to define your validation rules and responses, but sometimes you need to reflect your database models into schemas and keep changes in sync.

ModelSchema

ModelSchema is a special base class that can automatically generate schemas from your models.

All you need is to set model and fields attributes on your schema Meta:

from django.contrib.auth.models import User
from ninja import ModelSchema

class UserSchema(ModelSchema):
    class Meta:
        model = User
        fields = ['id', 'username', 'first_name', 'last_name']

# Will create schema like this:
# 
# class UserSchema(Schema):
#     id: int
#     username: str
#     first_name: str
#     last_name: str

Using ALL model fields

To use all fields from a model - you can pass __all__ to fields:

class UserSchema(ModelSchema):
    class Meta:
        model = User
        fields = "__all__"

Warning

Using all is not recommended.
This can lead to accidental unwanted data exposure (like hashed password, in the above example).
General advice - use fields to explicitly define list of fields that you want to be visible in API.

Excluding model fields

To use all fields except a few, you can use exclude configuration:

class UserSchema(ModelSchema):
    class Meta:
        model = User
        exclude = ['password', 'last_login', 'user_permissions']

# Will create schema like this:
# 
# class UserSchema(Schema):
#     id: int
#     username: str
#     first_name: str
#     last_name: str
#     email: str
#     is_superuser: bool
#     ... and the rest

Overriding fields

To change default annotation for some field, or to add a new field, just use annotated attributes as usual.

class GroupSchema(ModelSchema):
    class Meta:
        model = Group
        fields = ['id', 'name']


class UserSchema(ModelSchema):
    groups: List[GroupSchema] = []

    class Meta:
        model = User
        fields = ['id', 'username', 'first_name', 'last_name']

Making fields optional

Pretty often for PATCH API operations you need to make all fields of your schema optional. To do that, you can use config fields_optional

class PatchGroupSchema(ModelSchema):
    class Meta:
        model = Group
        fields = ['id', 'name', 'description'] # Note: all these fields are required on model level
        fields_optional = '__all__'

Also, you can define a subset of optional fields instead of __all__:

     fields_optional = ['description']

When you process input data, you need to tell Pydantic to avoid setting undefined fields to None:

@api.patch("/patch/{pk}")
def patch(request, pk: int, payload: PatchGroupSchema):

    # Notice that we set exclude_unset=True
    updated_fields = payload.dict(exclude_unset=True)

    obj = MyModel.objects.get(pk=pk)

    for attr, value in updated_fields.items():
        setattr(obj, attr, value)

    obj.save()

Custom fields types

For each Django field it encounters, ModelSchema uses the default Field.get_internal_type method to find the correct representation in Pydantic schema (python type). This process works fine for the built-in field types, but there are cases where the user wants to create or use a custom field, with its own mapping to python type. In this case you should use register_field method to tell django-ninja which type should this django field represent:

# models.py

class MyModel(models.Modle):
    embedding = pgvector.VectorField()

# schemas.py
from ninja.orm import register_field

register_field('VectorField', list[float])

PatchDict

Another way to work with patch request data is a PatchDict container which allows you to make a schema with all optional fields and get a dict with only fields that was provide

from ninja import PatchDict

class GroupSchema(Schema):
    # You do not have to make fields optional it will be converted by PatchDict
    name: str
    description: str
    due_date: date


@api.patch("/patch/{pk}")
def modify_data(request, pk: int, payload: PatchDict[GroupSchema]):
    obj = MyModel.objects.get(pk=pk)

    for attr, value in payload.items():
        setattr(obj, attr, value)

    obj.save()

in this example the payload argument will be a type of dict only fields that were passed in request and validated using GroupSchema