GraphQL APIs: REST in Peace

on YouTube ↗30m

" summary

Yelp replaces a three-endpoint, hundreds-of-fields API with a public GraphQL layer running on Python 3.6 and the graafian library on Pyramid, fronting a couple-hundred-service microservice backend. It solves the N+1 problem with per-request data loaders that batch many IDs (six business lookups become one bulk call), adds a Specter caching proxy that caches ID→value maps for bulk endpoints, enforces request budgets (example: 1000 ms) and node-based rate limiting plus middleware to cap resolver executions.

AI-written summary · about

" chapters (15)

1. 0:00Start
2. 0:42In the beginning
3. 2:37Query The representation of data you want returned
4. 3:38Our Setup
5. 9:00The N+1 Problem
6. 9:53Dataloaders!
7. 12:23Considerations
8. 14:35Service Caching
9. 17:00Caching in Bulk
10. 19:05Request Budgets
11. 22:50GraphQL
12. 23:45Node-based
13. 24:33Field-based
14. 25:33Securing the API
15. 27:50Easy* failure handling and retries

" related talks

• 38mBuilding mobile APIs with services at YelpStephan Jaensch · EuroPython 2015
• 31mGraphQL: how to make clients enjoy using your APITomasz Modrzyński · PyCon AU 2017
• 33mGraphQL Will Do To REST What JSON Did To XMLRoy Derks · NDC London 2019
• 55mGraphQL - APIs the New WayVolodymyr Tsukur · JavaDay Ukraine 2017
• 37mGraphQL — A New Look at the APIVladimir Zaets · JavaScript fwdays'18
• 54mGraphQL — Usable API or Just Another Hype?Vladimir Zaets · JS Fest 2018
• 58mGraphQL + Python today: Implementing a Public API with GraphQLSergey Khaletskiy · PiterPy 2019
• 22mWhy Migrate from REST to GraphQL?Erik Andersen · Commit Your Code 2024