Creating polymorphic models in Django

While there are a few ways to approach inheritance database models in Django, the main advantage of using polymorphic models lies in their simplicity of use and speed of development. At Merixstudio we appreciate this handy solution and use it in a number of projects to facilitate the creation of concise database relationships in an easy-to-read, clean-code manner.

Everyone who had a go at object-oriented programming surely experienced the use of polymorphism mechanisms one way or the other. Typically, you would create a class that implements some level of abstraction and override the functions in classes inheriting from it, changing output based on the object type. What if we could do the same thing with Django models? Today we will learn how to perform such magic using two Django packages.

Interesting use case example for polymorphic models

Before we jump into writing code, let’s go through a use case example. The project’s goal is to create an application for real estate agents to store information about the available purchases - what we want to keep there is the data regarding different types of real estate for sale. While some parameters will be shared between sold properties (title, price, description), some will be exclusive for a given type. A couple of examples include:

General real estate information (e.g. price), which is universally necessary for real estate agents, should be present in all models,
The floor parameter should be available for an apartment but it doesn’t make much sense to have it in a plot,
The plot type parameter would be applicable for a plot but not so useful in apartment parameters.

In Django we have a couple of options to choose from when solving the problem of varied data parameters:

We can store everything in one model with null information when an object doesn’t use the parameters that describe other objects,
We may build an elaborate relationship between models,

Or… you guessed it!

Use polymorphic models!

Initial steps with Django REST Framework

This guide assumes that you have a Django Rest Framework (DRF) project up and running. If you do not know how to work with Django or DRF, see the following guides:

We also recommend these insightful articles for a bigger picture about Django in the broader context of backend development:

We will start by installing the required packages using the pip tool (you can also add this straight to the project requirements file):

pip install django-polymorphic django-rest-polymorphic

Now we need to add the polymorphic app into our Django settings.py file:

INSTALLED_APPS = ( ... “polymorphic”, “django.contrib.contenttypes”, # Add this, if you haven't already ... )

And guess what… we are ready to make the magic happen!

Creating polymorphic models

Now that we have everything we need to create some exceptional code, let’s dig right into it. We will start by creating the base model that will contain the general information about the real estate we want to sell. Then, go to your models.py file and add the following:

from polymorphic.models import PolymorphicModel class RealEstate(PolymorphicModel): title = CharField(max_length=255) price = DecimalField(max_digits=9, decimal_places=2) description = TextField()

Notice how we inherit from a PolymorphicModel instead of the standard Django Model.

Next, we will create models for specific properties that will inherit from the RealEstate polymorphic model we just created.

class Flat(RealEstate): floor = PositiveSmallIntegerField(default=1) class Plot(RealEstate): class PlotType(TextChoices): CONSTRUCTION = "construction" AGRICULTURAL = "agricultural" plot_type = CharField(max_length=12, choices=PlotType.choices)

This went fast - we created three useful models in a matter of seconds.

REST a while and listen

Since we already have a data storage solution, now we can build some functionality around it. For this we will use the other django-rest-polymorphic package with DRF. It will allow us to use polymorphic models in a RESTful way, by providing endpoints to CRUD for model instances.

Django-based polymorphic serializers

To create serializers for each of the child models, we will be using the default ModelSerializer. Note how the fields of the parent model are defined in each of the child models below:

from rest_framework.serializers import ModelSerializer from rest_polymorphic.serializers import PolymorphicSerializer from .models import Flat, Plot, RealEstate class RealEstateSerializer(ModelSerializer): class Meta: model = RealEstate fields = ("title", "price", "description") class FlatSerializer(ModelSerializer): class Meta: model = Flat fields = ("title", "price", "description", "floor") class PlotSerializer(ModelSerializer): class Meta: model = Plot fields = ("title", "price", "description", "plot_type")

Now that we have created serializers for specific models, we will create a PolymorphicSerializer to connect them together:

This is done by providing a dictionary, where the key is our model and the value will be the serializer. This dictionary is then passed into the model_serializer_mapping variable:

class RealEstatePolymorphicSerializer(PolymorphicSerializer): model_serializer_mapping = { RealEstate: RealEstateSerializer, Flat: FlatSerializer, Plot: PlotSerializer }

Views creation in Django

To create a view, we will be using the ModelViewSet that provides all the CRUD capabilities out of the box. As you can see below, the view doesn’t look different from other views you create in Django. We pass the base RealEstate objects to the queryset, along with passing the serializer we just created to the serializer_class.

from rest_framework.viewsets import ModelViewSet from .models import RealEstate from .serializers import RealEstatePolymorphicSerializer class RealEstateViewSet(ModelViewSet): queryset = RealEstate.objects.all() serializer_class = RealEstatePolymorphicSerializer

After adding some objects into the database, a GET request in the view will provide data in the following format (look at how the structure is defined by the resourcetype key):

[ { "id": 1, "title": "Generic Real Estate", "price": 200000, "description": "Lorem ipsum dolor", "resourcetype": "RealEstate" }, { "id": 2, "title": "Flat Real Estate", "price": 100000, "description": "Lorem ipsum dolor", "floor": 3, "resourcetype": "Flat" }, { "id": 3, "title": "Plot Real Estate", "price": 300000, "description": "Lorem ipsum dolor", "plot_type": "construction", "resourcetype": "Plot" } ]

The admin panel

One way to work with the admin panel containing our models would be to simply register them like you normally would:

admin.site.register(RealEstate) admin.site.register(Flat) admin.site.register(Plot)

While this would work just fine, it does not provide all the features that django-polymorphic gives us regarding the admin panel. The package provides a cohesive interface for polymorphic models, that allows us to display different child models on the same page, when connected with the parent model. It also provides an easy way to create new models in one place through the admin panel, with regard to the type of the model.

With that in mind, we will start by adding imports into our admin.py file:

from django.contrib import admin from polymorphic.admin import ( PolymorphicChildModelAdmin, PolymorphicParentModelAdmin, ) from .models import Flat, Plot, RealEstate

Let’s continue with creating the child and parent model admins. The polymorphic_list keyword will allow the admin to place all models under a single page (quite neat!).

class FlatAdmin(PolymorphicChildModelAdmin): base_model = Flat class PlotAdmin(PolymorphicChildModelAdmin): base_model = Plot class RealEstateAdmin(PolymorphicParentModelAdmin): base_model = RealEstate

The last thing we need to do is to register the admins to our site:

admin.site.register(Flat, FlatAdmin) admin.site.register(Plot, PlotAdmin) admin.site.register(RealEstate, RealEstateAdmin)

Creating fixtures

What happens, if you want to use django-polymorphic to create some initial data for your project? As you probably noticed, the package uses Django content types to manage relationships between parent and child models. If you dump your data the standard way, you will get errors any time you change the model structure (ContentType objects are automatically created by Django during the database synchronization process). How can we deal with that problem? It is quite simple - by using natural keys.

python manage.py dumpdata --natural-foreign > project_data.json

Natural keys are a feature provided by Django for convenience. Instead of referring to an object by its ID field, a natural key provides a tuple of values to identify an object instance. This also allows you to load data into an existing database, when you can’t guarantee that the primary key is not already in use by another object. For more information please visit the official documentation.

All caught up in Django? Here you will find out how to boost your Django project with Celery!

Are you ready for an adventure?

Today we created a simple application for real estate agents. While that wasn’t the most innovative product around, we learned how to easily bring polymorphism into the world of Django models. We only scratched the surface of what the Django packages have to offer, so if you are interested in learning more, be sure to click the below links we selected for you. Once you check them out, you will know that this is just the beginning of your adventure with polymorphism!

The official sources with more information about both packages:

Want to use Django-based polymorphic models in practice? Join our smashing backend team!