goliath_playground

Random playing around with goliath to build performant API endpoints

5 years after

Goliath for asynchronous processing and streaming API services

Why?

I'm curious and I want to play. Also Heroku Cedar is an efficient pathway to deploying this kind of code, since it works out of the box.

Goal

A goliath router that returns interesting, well-formed JSON from two endpoints: one streaming, one not. Firefox is, I think, the only browser that supports streamed responses at the moment (EDIT: nope, new Chrome does too), but non-browser clients (e.g. cURL, Node.js, other Eventmachine-backed Ruby services) should be able to consume streamed JSON just fine. If you're on ruby, check out yajl-ruby. If you're on node.js, you can either parse it yourself or use yajl-js. Check it out: YAJL, DIY help.

Maybe we'll get websockets going and be able to stream to a websocket someday :).

HOWTO

  • bundle install
  • You can run this puppy with bundle exec ruby router.rb -sv -e prod -p $PORT
  • play around with curl, e.g. curl http://localhost:$PORT
  • or with a modern browser (FF4+, newish Chrome) you can check out the streaming action in the browser!
  • visit the root page to get a quick overview and relevant links.

Heroku's new Cedar stack supports all of these toys, too.

A Tip of the Hat to:

  • Ilya Grigorik, natch. His blog post was the inspiration.
  • Heroku, for the Cedar stack.

Top Contributors

hypomodern