Django + Pydantic = 🖤

Django JSONField with Pydantic models as a Schema.

Now supports both Pydantic v1 and v2! Please join the discussion if you have any thoughts or suggestions!

Usage

Install the package with pip install django-pydantic-field.

import pydantic
from datetime import date
from uuid import UUID

from django.db import models
from django_pydantic_field import SchemaField


class Foo(pydantic.BaseModel):
    count: int
    size: float = 1.0


class Bar(pydantic.BaseModel):
    slug: str = "foo_bar"


class MyModel(models.Model):
    # Infer schema from field annotation
    foo_field: Foo = SchemaField()

    # or explicitly pass schema to the field
    bar_list: typing.Sequence[Bar] = SchemaField(schema=list[Bar])

    # Pydantic exportable types are supported
    raw_date_map: dict[int, date] = SchemaField()
    raw_uids: set[UUID] = SchemaField()

...

model = MyModel(
    foo_field={"count": "5"},
    bar_list=[{}],
    raw_date_map={1: "1970-01-01"},
    raw_uids={"17a25db0-27a4-11ed-904a-5ffb17f92734"}
)
model.save()

assert model.foo_field == Foo(count=5, size=1.0)
assert model.bar_list == [Bar(slug="foo_bar")]
assert model.raw_date_map == {1: date(1970, 1, 1)}
assert model.raw_uids == {UUID("17a25db0-27a4-11ed-904a-5ffb17f92734")}

Practically, schema could be of any type supported by Pydantic. In addition, an external config class can be passed for such schemes.

Forward referencing annotations

It is also possible to use SchemaField with forward references and string literals, e.g the code below is also valid:

class MyModel(models.Model):
    foo_field: "Foo" = SchemaField()
    bar_list: typing.Sequence["Bar"] = SchemaField(schema=typing.ForwardRef("list[Bar]"))


class Foo(pydantic.BaseModel):
    count: int
    size: float = 1.0


class Bar(pydantic.BaseModel):
    slug: str = "foo_bar"

Pydantic v2 specific: this behaviour is achieved by the fact that the exact type resolution will be postponed until the initial access to the field. Usually this happens on the first instantiation of the model.

To reduce the number of runtime errors related to the postponed resolution, the field itself performs a few checks against the passed schema during ./manage.py check command invocation, and consequently, in runserver and makemigrations commands.

Here's the list of currently implemented checks:

pydantic.E001: The passed schema could not be resolved. Most likely it does not exist in the scope of the defined field.
pydantic.E002: default= value could not be serialized to the schema.
pydantic.W003: The default value could not be reconstructed to the schema due to include/exclude configuration.

`typing.Annotated` support

As of v0.3.5, SchemaField also supports typing.Annotated[...] expressions, both through schema= attribute or field annotation syntax; though I find the schema=typing.Annotated[...] variant highly discouraged.

The current limitation is not in the field itself, but in possible Annotated metadata -- practically it can contain anything, and Django migrations serializers could refuse to write it to migrations. For most relevant types in context of Pydantic, I wrote the specific serializers (particularly for pydantic.FieldInfo, pydantic.Representation and raw dataclasses), thus it should cover the majority of Annotated use cases.

Django Forms support

It is possible to create Django forms, which would validate against the given schema:

from django import forms
from django_pydantic_field.forms import SchemaField


class Foo(pydantic.BaseModel):
    slug: str = "foo_bar"


class FooForm(forms.Form):
    field = SchemaField(Foo)  # `typing.ForwardRef("Foo")` is fine too, but only in Django 4+


form = FooForm(data={"field": '{"slug": "asdf"}'})
assert form.is_valid()
assert form.cleaned_data["field"] == Foo(slug="asdf")

django_pydantic_field also supports auto-generated fields for ModelForm and modelform_factory:

class MyModelForm(forms.ModelForm):
    class Meta:
        model = MyModel
        fields = ["foo_field"]

form = MyModelForm(data={"foo_field": '{"count": 5}'})
assert form.is_valid()
assert form.cleaned_data["foo_field"] == Foo(count=5)

...

# ModelForm factory support
AnotherModelForm = modelform_factory(MyModel, fields=["foo_field"])
form = AnotherModelForm(data={"foo_field": '{"count": 5}'})

assert form.is_valid()
assert form.cleaned_data["foo_field"] == Foo(count=5)

Note, that forward references would be resolved until field is being bound to the form instance.

`django-jsonform` widgets

django-jsonform offers a dynamic form construction based on the specified JSONSchema. django_pydantic_field.forms.SchemaField plays nicely with its widgets, but only for Pydantic v2:

from django_pydantic_field.forms import SchemaField
from django_jsonform.widgets import JSONFormWidget

class FooForm(forms.Form):
    field = SchemaField(Foo, widget=JSONFormWidget)

It is also possible to override the default form widget for Django Admin site, without writing custom admin forms:

from django.contrib import admin
from django_jsonform.widgets import JSONFormWidget

# NOTE: Importing direct field class instead of `SchemaField` wrapper.
from django_pydantic_field.v2.fields import PydanticSchemaField

@admin.site.register(MyModel)
class MyModelAdmin(admin.ModelAdmin):
    formfield_overrides = {
        PydanticSchemaField: {"widget": JSONFormWidget},
    }

Django REST Framework support

from rest_framework import generics, serializers
from django_pydantic_field.rest_framework import SchemaField, AutoSchema


class MyModelSerializer(serializers.ModelSerializer):
    foo_field = SchemaField(schema=Foo)

    class Meta:
        model = MyModel
        fields = '__all__'


class SampleView(generics.RetrieveAPIView):
    serializer_class = MyModelSerializer

    # optional support of OpenAPI schema generation for Pydantic fields
    schema = AutoSchema()

Global approach with typed parser and renderer classes

from rest_framework import views
from rest_framework.decorators import api_view, parser_classes, renderer_classes
from django_pydantic_field.rest_framework import SchemaRenderer, SchemaParser, AutoSchema


@api_view(["POST"])
@parser_classes([SchemaParser[Foo]]):
@renderer_classes([SchemaRenderer[list[Foo]]])
def foo_view(request):
    assert isinstance(request.data, Foo)

    count = request.data.count + 1
    return Response([Foo(count=count)])


class FooClassBasedView(views.APIView):
    parser_classes = [SchemaParser[Foo]]
    renderer_classes = [SchemaRenderer[list[Foo]]]

    # optional support of OpenAPI schema generation for Pydantic parsers/renderers
    schema = AutoSchema()

    def get(self, request, *args, **kwargs):
        assert isinstance(request.data, Foo)
        return Response([request.data])

    def put(self, request, *args, **kwargs):
        assert isinstance(request.data, Foo)

        count = request.data.count + 1
        return Response([request.data])