Evolving GraphQL Schemas
summary
A team builds Compass on GraphQL and lays down concrete rules: use bounded inputs, unbounded outputs to avoid enum breakage and prefer nullable-first for fields to reduce future churn. They deploy a custom GraphQL Gateway that enforces lifecycle tags (experimental, beta, production), per-user rate limits, OAuth scopes, and hydration directives, and they use persistent queries to control costly queries. A proposed schema generator "graph clippy" is attempted but abandoned, while directive-driven platform features prove effective.
AI-written summary · about
chapters (9)
- 0:00Introduction to GraphQL schemas
- 2:26Basics and early complexities
- 9:37Nullability pitfalls and solutions
- 12:45Scaling with contributor growth
- 16:36Query cost and data control
- 19:29Deprecation and breaking changes
- 21:58GraphQL as a platform service
- 28:03Key takeaways and recap
- 29:26Q&A session
related talks
21mBuilt to Evolve: 13 Years of GraphQL
23mAn Introduction to GraphQL
32mFrom learning to thinking in GraphQL
29mClosing the Loop: How GraphQL Gives Coding Agents Eyes on What Actually Matters
49mGraphQL Schema Design — Building Schemas Correctly (version 2)
26mIncrementally Adopting GraphQL. The Holy Grail?
47mThe State of GraphQL Federation
50mDistributed GraphQL APIs: Breaking Down Monoliths & Unlocking Agility for Dev Teams