Skip to content

Django classes to make your models, managers, and querysets serializable, with built-in support for related objects in ~150 LoC

License

Notifications You must be signed in to change notification settings

agilgur5/django-serializable-model

Repository files navigation

django-serializable-model

PyPI version releases commits
dm dw
build status code coverage
Django classes to make your models, managers, and querysets serializable, with built-in support for related objects in ~100 LoC (shorter than this README!)

Table of Contents

I. Installation
II. Usage
III. How it Works
IV. Related Libraries
V. Backstory

Installation

pip install django-serializable-model

It is expected that you already have Django installed

Compatibility

Python versions Django versions

Tested on Django 2.2, 1.11, 1.9, and 1.5 as well as Python 3.5, 3.4, and 2.7

  • Should work with Django 1.4-2.x and Python 2.7-3.x
    • Has several Django backward & forward compatibility fixes built-in
  • Likely works with Django 0.95-1.3 as well
    • Pre 1.3 does not support the on_delete argument on relations. This only affects the usage and examples below; the internals are unaffected.
    • Pre 0.95, the Manager API didn't exist, so some functionality may be limited in those versions, or it may just error on import
  • Have not confirmed if this works with earlier versions of Python.

Please submit a PR or file an issue if you have a compatibility problem or have confirmed compatibility on versions.


Usage

Simplest use case, just implements the .serialize() function on a model:

from django.db import models
from django_serializable_model import SerializableModel


class User(SerializableModel):
    email = models.CharField(max_length=765, blank=True)
    name = models.CharField(max_length=100)


new_user = User.objects.create(
    name='John Doe',
    email='[email protected]',
)

print new_user.serialize()
# {'id': 1, 'email': '[email protected]', 'name': 'John Doe'}

With an override of the default .serialize() function to only include whitelisted fields in the serialized dictionary:

from django.db import models
from django_serializable_model import SerializableModel


class User(SerializableModel):
    email = models.CharField(max_length=765, blank=True)
    name = models.CharField(max_length=100)
    # whitelisted fields that are allowed to be seen
    WHITELISTED_FIELDS = set([
        'name',
    ])


    def serialize(self, *args, **kwargs):
        """Override serialize method to only serialize whitelisted fields"""
        fields = kwargs.pop('fields', self.WHITELISTED_FIELDS)
        return super(User, self).serialize(*args, fields=fields)


new_user = User.objects.create(
    name='John Doe',
    email='[email protected]',
)

print new_user.serialize()
# {'name': 'John Doe'}

With a simple, one-to-one relation:

from django.db import models
from django_serializable_model import SerializableModel


class User(SerializableModel):
    email = models.CharField(max_length=765, blank=True)
    name = models.CharField(max_length=100)


class Settings(SerializableModel):
    user = models.OneToOneField(User, primary_key=True,
        on_delete=models.CASCADE)
    email_notifications = models.BooleanField(default=False)

    def serialize(self, *args):
        """Override serialize method to not serialize the user field"""
        return super(Settings, self).serialize(*args, exclude=['user'])


new_user = User.objects.create(
    name='John Doe',
    email='[email protected]',
)
Settings.objects.create(user=new_user)

new_user_refreshed = User.objects.select_related('settings').get(pk=new_user.pk)

print new_user_refreshed.serialize()
# {'id': 1, 'email': '[email protected]', 'name': 'John Doe'}

# recursively serialize Settings object by passing the join in
print new_user_refreshed.serialize('settings')
# {'id': 1, 'email': '[email protected]', 'settings': {'email_notifications': False}, 'name': 'John Doe'}

With a foreign key relation:

from django.db import models
from django_serializable_model import SerializableModel


class User(SerializableModel):
    email = models.CharField(max_length=765, blank=True)
    name = models.CharField(max_length=100)


class Post(SerializableModel):
    user = models.ForeignKey(User, on_delete=models.CASCADE)
    text = models.TextField()


new_user = User.objects.create(
    name='John Doe',
    email='[email protected]',
)
Post.objects.create(user=new_user, text='wat a nice post')
Post.objects.create(user=new_user, text='another nice post')

# called on QuerySet
print Post.objects.all().serialize()
# [{'id': 1, 'text': 'wat a nice post', 'user_id': 1}, {'id': 2, 'text': 'another nice post', 'user_id': 1}]
# adds an _id to the foreign key name, just like when using `.values()`

# called on Manager
user1 = User.objects.get(pk=new_user.pk)
print user1.post_set.serialize()
# [{'id': 1, 'text': 'wat a nice post', 'user_id': 1}, {'id': 2, 'text': 'another nice post', 'user_id': 1}]

# recursively serialize Post objects by passing the join in
print User.objects.prefetch_related('post_set').get(pk=new_user.pk).serialize('post_set')
"""
{
    'id': 1,
    'email': '[email protected]',
    'name': 'John Doe',
    'post_set': [{'id': 1, 'text': 'wat a nice post', 'user_id': 1}, {'id': 2, 'text': 'another nice post', 'user_id': 1}]
}
"""

.serialize takes in any number of joins as its *args and they can be of any depth, using the same __ syntax as prefetch_related. This means if your Post object also had Comment objects, you could write:

User.objects.prefetch_related('post_set__comment_set').serialize('post_set__comment_set')

and get an array of Comment dictionaries within each Post dictionary. If your Post object also had Like objects:

joins = ['post_set__comment_set', 'post_set__like_set']
User.objects.prefetch_related(*joins).serialize(*joins)

JSON and APIs

Since .serialize outputs a dictionary, one can turn it into JSON simply by using json.dumps on the dictionary.

If you're building an API, you can use JSONResponse on the dictionary as well.


How it works

Implementing a .serialize method on Models, Managers, and QuerySets allows for easily customizable whitelists and blacklists (among other things) on a per Model basis. This type of behavior was not possible a simple recursive version of model_to_dict, but is often necessary for various security measures and overrides. In order to recurse over relations / joins, it accepts the same arguments as the familiar prefetch_related, which, in my use cases, often immediately precedes the .serialize calls. .serialize also uses a custom model_to_dict function that behaves a bit differently than the built-in one in a variety of ways that are more expected when building an API (see the docstring).

I'd encourage you to read the source code, since it's shorter than this README :)

Related Libraries

  • django-api-decorators
    • Tiny decorator functions to make it easier to build an API using Django in ~100 LoC

Backstory

This library was built while I was working on Yorango's ad-hoc API. Writing code to serialize various models was complex and quite tedious, resulting in messy spaghetti code for many of our API methods. The only solutions I could find online were the Django Full Serializers from wadofstuff as well as some recursive model_to_dict snippets online -- none of which gave the option for customizable whitelists and blacklists on a per Model basis. Later on, I found that Django REST Framework's ModelSerializers do offer similar functionality to what I was looking for (and without requiring buy-in to the rest of the framework), albeit with some added complexity and robustness.

I ended up writing my own solution in ~100 LoC that handled basically all of my needs and replaced a ton of messy serialiazation code from all around the codebase. It was used in production with fantastic results, including on queries with quite the complexity and depth, such as:

    joins = ['unit_set', 'unit_set__listing_set',
        'unit_set__listing_set__tenants', 'unit_set__listing_set__bill_set',
        'unit_set__listing_set__payment_set__payer',
        'unit_set__listing_set__contract']
    s_props = (user.property_set.all().prefetch_related(*joins)
        .serialize(*joins))

Had been meaning to extract and open source this as well as other various useful utility libraries I had made at Yorango and finally got the chance!