Why not use normal fixtures?

They’re not only a little verbose, but they need to be loaded into a non-memory database (slower) and you need lots of logic to work out if you should update or delete existing entries. They’re still a better solution for anything that has a lot of data or which needs JOINs, though.

Installation

First, install the package:

pip install yamdl

Then, add it to INSTALLED_APPS:

INSTALLED_APPS = [
    ...
    'yamdl',
    ...
]

Then, add the in-memory database to DATABASES (note that you must have at least Python 3.4 to have a SQlite module that understands shared memory URIs):

DATABASES = {
    ...
    'yamdl': {
        'ENGINE': 'django.db.backends.sqlite3',
        'NAME': 'file:yamdl-db?mode=memory&cache=shared',
    }
}

Then, add a YAMDL_DIRECTORIES setting which defines where your directories of YAML files can be found (it’s a list):

YAMDL_DIRECTORIES = [
    os.path.join(PROJECT_DIR, "content"),
]

Finally, add the database router:

DATABASE_ROUTERS = [
    "yamdl.router.YamdlRouter",
]

Usage

First, add the __yamdl__ attribute to the models you want to use static content. A model can only be static or dynamic, not both:

class MyModel(models.Model):
    ...
    __yamdl__ = True

Then, start making static files under one of the directories you listed in the YAMDL_DIRECTORIES setting above. Within one of these, make a directory with the format appname.modelname, and then YAML files ending in .yaml:

andrew-site/
    content/
        speaking.Talk/
            2017.yaml
            2016.yaml

Within those YAML files, you can define either a list of model instances, like this:

- title: 'Alabama'
  section: us-states

- title: 'Alaska'
  section: us-states
  done: 2016-11-18
  place_name: Fairbanks

- title: 'Arizona'
  section: us-states
  done: 2016-05-20
  place_name: Flagstaff

Or a single model instance at the top level, like this:

conference: DjangoCon AU
title: Horrors of Distributed Systems
when: 2017-08-04
description: Stepping through some of the myriad ways in which systems can fail that programmers don't expect, and how this hostile environment affects the design of distributed systems.
city: Melbourne
country: AU
slides_url: https://speakerdeck.com/andrewgodwin/horrors-of-distributed-systems
video_url: https://www.youtube.com/watch?v=jx1Hkxe64Xs

When you start up Django, as either runserver or in production, it will read the YAML files and load them into an in-memory database and then let you query them using all the standard ORM stuff.

Todo

Here’s a short list of things I’d like to get done before a 1.0:

• Maybe replace the __yamdl__ attribute with something nicer.

• Support for Python versions before 3.4, either by using a global SQLite :memory: instance with thread locking or by supporting disk databases with a wipe phase.

• Include YAML files in the Django auto-reloader so editing them loads changes in development.

• Potentially load changes to flat files in production using mtime checking.