/aioflask

Flask running on asyncio!

Primary LanguagePythonMIT LicenseMIT

aioflask

Build status codecov

Flask 2.x running on asyncio!

Is there a purpose for this, now that Flask 2.0 is out with support for async views? Yes! Flask's own support for async handlers is very limited, as the application still runs inside a WSGI web server, which severely limits scalability. With aioflask you get a true ASGI application, running in a 100% async environment.

WARNING: This is an experiment at this point. Not at all production ready!

Quick start

To use async view functions and other handlers, use the aioflask package instead of flask.

The aioflask.Flask class is a subclass of flask.Flask that changes a few minor things to help the application run properly under the asyncio loop. In particular, it overrides the following aspects of the application instance:

  • The route, before_request, before_first_request, after_request, teardown_request, teardown_appcontext, errorhandler and cli.command decorators accept coroutines as well as regular functions. The handlers all run inside an asyncio loop, so when using regular functions, care must be taken to not block.
  • The WSGI callable entry point is replaced with an ASGI equivalent.
  • The run() method uses uvicorn as web server.

There are also changes outside of the Flask class:

  • The flask aiorun command starts an ASGI application using the uvicorn web server.
  • The render_template() and render_template_string() functions are asynchronous and must be awaited.
  • The context managers for the Flask application and request contexts are async.
  • The test client and test CLI runner use coroutines.

Example

import asyncio
from aioflask import Flask, render_template

app = Flask(__name__)

@app.route('/')
async def index():
    await asyncio.sleep(1)
    return await render_template('index.html')