django-banjo 0.2.4

A simplified abstraction over django for beginners.

Banjo

Banjo is an abstraction over django which provides a simplified subset of django's functionality, meant for beginners. All the django documentation can be used for reference, and many django patterns are replicated here.

Needs

The purpose of Banjo is to introduce databases as a persistence layer behind an API server, while abstracting away details for which students are not yet ready and creating as few misconceptions as possible. Banjo should be thought of as scaffolding; when they are ready, students should be able to seamlessly transition to django.

Specific concepts which we target for simplification include:

• Simplify DB schema: A severely-limited subset of field types is provided. Field names correspond to familiar Python types. All fields have default values. Migrations are handled automatically. Within these constraints, django's full ORM is available.
• Simplify filesystem layout: Only two files are required: models.py and views.py.
• Simplify management commands: There is a single command, banjo, which effectively runs django's makemigrations, migrate, and runserver in sequence. banjo --shell enters the REPL with all user-defined models loaded.
• Simplify request/response lifecycle: View functions receive a dict of params and must return a dict. View-URL binding is handled by decorators, as in flask, and all URLs are static (there are no placeholders and no params are passed to the view). Http errors are provided as exceptions, which simplifies control flow. Models have from_dict (class method) and to_dict (instance method) helpers.

Banjo was designed for use by teachers familiar with django; this is intentionally a leaky abstraction which provides a structured introduction into the power and the complexity of the underlying system.

Creating an app

Banjo can be installed with pip install django-banjo.

To write a Banjo app, create a folder called app, define models in models.py and define views in views.py. Here's a simple example.

Models

First, we define our models. Banjo provides five field types:

• BooleanField (True, False)
• IntegerField (1, -102)
• FloatField (0.045, 11.5)
• StringField ("alligator", "hazelnut")
• ForeignKey (An instance of another model)

Create a Model for each object your app will interact with.

# app/models.py
from banjo.models import Model, StringField

class Animal(Model):
    name = StringField()
    sound = StringField()

Views

Next we define our views. Each view is a function which receives a dict (called params in the example below) and which must return a dict. Use the banjo.urls.route_get and banjo.urls.route_post decorators to route URLs to your view functions.

# app/views.py
from banjo.urls import route_get, route_post
from app.models import Animal

@route_post('newanimal', args={'name': str, 'sound': str})
def add_animal(params):
    animal = Animal.from_dict(params)
    animal.save()
    return animal.to_dict()

@route_get('listen')
def list_animal_sounds(params):
    sounds = []
    for animal in Animal.objects.all():
        sounds.append('{} says {}'.format(animal.name, animal.sound))     
    return {'sounds': sounds}

Some views, such as "newanimal," require arguments. When a view requires arguments, pass an args dict to the decorator to specify the expected names and types of arguments. Argument types must be str, bool, int, or float.

HTTP errors

If something goes wrong and it's the client's fault, you can raise an error. For example, you might add another view to app/views.py:

from banjo.http import Forbidden

@route_get('secrets')
def do_not_show_the_secrets(params):
    raise Forbidden("Nice try.")

Again, from the command line:

$ http GET localhost:5000/secrets
HTTP/1.1 403 Forbidden

{
    "error": "Nice try."
}

The following errors are available in banjo.http:

• BadRequest (400)
• Forbidden (403)
• NotFound (404)
• NotAllowed (405)
• ImATeapot (418)

Running the app

Now you can run banjo from the directory containing the app folder and the server will start. Use the --port command to serve from a custom port; the default is 5000.

Banjo provides a visual API browser at /api.

Here is an example of interacting with this app using the httpie command-line utility:

$ http localhost:5000/newanimal name=elehpant sound=pffffftttttt

{ 
  "id": 1,
  "name": "elephant",
  "sound": "pffffftttttt"
}

$ http localhost:5000/newanimal name=squirrel sound=chcheee

{ 
  "id": 2,
  "name": "squirrel",
  "sound": "chcheee"
}

$ http localhost:5000/listen

{
  "sounds": [
    "elephant says pffffftttttt",
    "squirrel says chcheee"
  ]
}

Shell

You can also interact with your app's models from a Python shell. Just pass the --shell argument to banjo:

$ banjo --shell
> Animal.objects.count()
2

Deploying to Heroku

Heroku is a service which simplifies app deployment. Deploying a banjo app wih Heroku lets anyone on the internet interact with it. Be careful about protecting private information. Banjo is best for learning how to make webapps; if anybody is going to rely on your app in a serious way, there are more details you ought to learn about. This would be the right time to move from banjo to django.

1. Create a heroku account and install the heroku command line tool. Follow these steps.

2. Run banjo --init_heroku. This will add a few files to your app which are required by Heroku. You do not need to make any changes to these files.

3. Deployment to Heroku is done via git, so you need to make sure you have your project in a git repo. Learning git is another whole topic, but the simplest workflow would include:

    $ git init
    $ git add .
    $ git commit -m "Initial commit"
    $ git branch -m master main
   

4. Now you can deploy your app by pushing your code to Heroku.

    $ heroku create
    $ git push heroku main
   

5. You should now be able to interact with your app. Run heroku open to open it in the web browser. If there is an error, heroku logs --tail will show you what's going on. This is a good place to start with debugging. After you update your code, commit it and push again to heroku.