Django Initial Data

Django Dynamic Initial Data is a django>=1.6 and postgresql only app that helps solve the problem of initializing data for apps with dependencies and other conditional data. Rather than having static fixtures for each app, the initial data can be created and updated dynamically. Furthermore, Django Dynamic Initial Data also handles when objects are deleted from initial data, a feature that Django's initial data fixture system lacks.

Table of Contents

1. [Installation] (#installation)
2. [A Brief Overview] (#a-brief-overview)
3. [Example] (#example)
4. Handling Deletions

Installation

To install Django Dynamic Initial Data:

pip install django-dynamic-initial-data

Add Django Dynamic Initial Data to your INSTALLED_APPS to get started:

settings.py

INSTALLED_APPS = (
    'dynamic_initial_data',
)

A Brief Overview

A management command update_initial_data is provided which will try to update all INSTALLED_APPS. This command is intended to be called as part of the deployment process of your app. Any missing dependencies will raise an InitialDataMissingApp exception and any circular dependencies will raise an InitialDataCircularDependency exception.

Any app needing to define initial data needs a file called initial_data.py inside of a fixtures directory. This will look like {app_name}/fixtures/initial_data.py. Don't forget to include the __init__.py file in the fixtures directory. initial_data.py must define a class InitialData that inherits from BaseInitialData.

When apps are being initialized, each InitialData class is instantiated and update_initial_data is called. If update_initial_data is not implemented, then a NotImplementedError will be raised.

Any dependencies should be included in a list called dependencies. Each dependency is a string of the app name as defined in INSTALLED_APPS.

Example:

from dynamic_initial_data.base import BaseInitialData

class InitialData(BaseInitialData):
    dependencies = ['my_first_app', 'my.second.app']

    def update_initial_data(self):
        model_obj, created = TestModel.objects.upsert(int_field=5, defaults={'float_field': 2.0})

        TestModel.objects.bulk_upsert([
            TestModel(float_field=1.0, char_field='1', int_field=1),
            TestModel(float_field=2.0, char_field='2', int_field=2),
            TestModel(float_field=3.0, char_field='3', int_field=3),
        ], ['int_field'], ['char_field'])

In this example, the update_initial_data method will be called for my_first_app (following any dependencies first), and then for my.second.app, before finally calling update_initial_data on this class. Again, this can be executed by calling

python manage.py update_initial_data

Similarly, to only initialize a single app, use

python manage.py update_initial_data --app 'app_path'

Documentation on using upsert and bulk_upsert can be found below:

• https://github.com/ambitioninc/django-manager-utils#upsert
• https://github.com/ambitioninc/django-manager-utils#bulk_upsert

Handling Deletions

One difficulty when specifying initial data in Django apps is the inability to deploy initial data to your project and then subsequently remove any initial data fixtures. If one removes an object in an initial_data.json file, Django does not handle its deletion next time it is deployed, which can cause headaches with lingering objects.

Django Dynamic Initial Data fixes this problem by allowing the user to either:

1. Return all managed initial data objects as an array from the update_initial_data function.
2. Explicitly register objects for deletion with the register_for_deletion(*model_objs) method.

Note that it is up to the user to be responsible for always registering every object every time, regardless if the object was updated or created by the initial data process. Doing this allows Django Dynamic Initial Data to remove any objects that were previosly managed. For example, assume you have an InitialData class that manages two users with the user names "hello" and "world".

from dynamic_initial_data.base import BaseInitialData

class InitialData(BaseInitialData):
    def update_initial_data(self):
        hello = Account.objects.get_or_create(name='hello')
        world = Account.objects.get_or_create(name='world')
        # register the accounts for deletion
        self.register_for_deletion(hello, world)

After this code is created, the initial data process now owns the "hello" and "world" account objects. If these objects are not registered for deletion in subsequent versions of the code, they will be deleted when the initial data process executes. For example, assume the first piece of code executed and then the user executed this piece of code:

from dynamic_initial_data.base import BaseInitialData

class InitialData(BaseInitialData):
    def update_initial_data(self):
        world = Account.objects.get_or_create(name='world')
        # register the accounts for deletion
        self.register_for_deletion(world)

When this piece of code executes, the previous "hello" account would then be deleted since the initial data process no longer owns it. And don't worry, if it was already deleted by another process, the deletion will not throw an error.