Django backend for YugabyteDB

Prerequisites

GCC
Python 3.8 and above
Psycopg2-binary
Django 3.2 or above

Need for Django Backend for YugabyteDB

YugabyteDB needs a separate backend for Django. This is because of mainly 2 reasons.

Django tries to create Inet data types as primary keys in Django test suites. Since this is not supported, we map Inet types to varchar(15) and varchar(39) in the YB backend.
We also need it to support type change from int to BigInt and numeric(m,n) to double precision. This is required for Django DB migrations. For now, the YB backend ignores these type changes.

Installing from Pypi

Install the django-yugabytedb package with the command:

$ pip install django-yugabytedb

Installing in Python Virtual Environment From Source

The django-yugabytedb package can also be installed from source:

$ git clone https://github.com/yugabyte/yb-django.git

$ python -m pip install -r <repo_path>/yb-django/requirements.txt

$ python -m pip install -e <repo_path>/yb-django/

Check if it is installed correctly:

$ pip list —local

Use the backend with your Application

Update the DATABASES setting in your Django project's settings to point to YB server using Django YB backend:

DATABASES = {
    'default': {
        'ENGINE': 'django_yugabytedb',
        'NAME': 'yugabyte',
        'HOST': 'localhost',
        'PORT': 5433,
        'USER': 'yugabyte'
    }
}

Known bugs and issues

The creation of indexes in YugabyteDB is a little slow.
Since Inet is mapped to varchar in the backend, comparison between data in the inet column wth Inet type will fail.
For YugabyteDB verions earlier than 2.9, the savepoint feature is not supported.
Dropping of Primary keys are not supported
Alter table Add column Unique is not yet supported.
Backfilling of existing rows when new column is added with default value is not yet implemented in yugabytedb.
Type change from int to BigInt and numeric(m,n) to double precision is not yet supported.
ALTER INDEX not supported yet.