Search code examples
pythonrestflasksqlalchemy

Flask blueprints with SQLAlchemy for APIs -- what can I do to simplify?


I am designing an app that has a database, a CRUD web interface to manage its content, and exposing a REST API to be called by an external UI written in React. I am using native Flask Blueprints, SQLAlchemy, along with Marshmallow for API serialization.

The way I have structured Blueprints, based on what I learned from the Flask docs, seems unnecessarily complicated:

  • app.py -> __init__.py.create_app() registers blueprints
    • "factory" Flask pattern
    • each blueprint imports from a routes module
    • register_blueprint specifies two blueprints one for HTML, another for API as each has a different url_prefix
  • each routes module __init__.py
    • imports Blueprint, then
    • creates the blueprint instances for html and api, then
    • imports the actual routes for the module (e.g. from routes.user import routes)
  • each routes implementation
    • imports the blueprints created in the module
    • imports the SQLAlchemy model and Marshmallow schema
    • creates the code for the routes, and decorates with the correct blueprint depending on HTML or API
  • each model imports required elements for SQLAlchemy

So the SQLAlchemy approach seems just right: in one place I define all the characteristics of the model and thus the associated database tables, as well as the schema for JSON output of my API.

But the Blueprints approach seems convoluted, with dependencies scattered about -- to define one new route, I typically have to modify three files:

  • app module (__init.py__)
  • routes module for the specific route (routes/user/__init.py__)
  • routes implementation for the specific route (routes/user/routes.py)
  • and each of these implicitly depend on the model for the entity (e.g. user)

All of this is functional, but I am assuming there's a simpler way to approach the problem that I have not yet fully grasped. Is there a simpler, cleaner way to implement this?


Solution

  • you can follow approach as in flask-cookiecutter

    in app.py you can create app as

    def create_app(config_object="{{cookiecutter.app_name}}.settings"):
        """Create application factory, as explained here: http://flask.pocoo.org/docs/patterns/appfactories/.
    
        :param config_object: The configuration object to use.
        """
        app = Flask(__name__.split(".")[0])
        app.config.from_object(config_object)
      
        register_blueprints(app)
     
        return app
    

    and you can create blueprints in any routes.py or ``views.py file in project and then register them (add them to app) in the register_blueprint function as

    def register_blueprints(app):
        """Register Flask blueprints."""
        app.register_blueprint(public.views.blueprint)
        app.register_blueprint(user.views.blueprint)
        return None
    

    this will eliminate the need to add routes to routes/__init__.py file each time you create views in different routes for different use case.