The Benefits of Code First Over Schema First in GraphQL
summary
Argument: code-first yields faster, safer GraphQL development by keeping schema and resolvers in one place and leveraging language types. Example implementation uses Strawberry (Python), MyPy for end-to-end type checking, generics for reusable pagination, and module imports instead of schema stitching. Built an export flow to run GraphOS/Rover checks and Apollo Studio via GitHub Actions — Rover flags a bad type name in the demo. Federation works by generating directives (and a lighter federation API), but avoid runtime dynamic schema or blind ORM-generated types (SQLAlchemy/Django).
AI-written summary · about
chapters (18)
- 0:00Introduction to code first
- 1:45Schema first vs code first
- 3:10Introducing Strawberry library
- 5:37Code first with Python types
- 7:24Single source of truth benefit
- 8:28Type safety with Python typings
- 10:13Modularizing the codebase
- 11:16Generic types for pagination
- 13:23Benefits summary so far
- 14:05Enforcing style guide with linters
- 15:06Exporting schema for tools
- 17:12Apollo Federation support
- 18:55Schema directives in code first
- 21:21Advice on code first pitfalls
- 22:48Design first approach
- 23:28SDL to Python converter demo
- 24:54Code first libraries in other languages
- 25:37Summary and closing remarks
related talks
29mWhy You Should Use Implementation-First to Build Your GraphQL Schema
31mAccelerating GraphQL API Development with Schema Storming and Code-first Implementation
33mSchema First, Code First, or Both? Adopting Schema First Development
32mFrom learning to thinking in GraphQL
08mWhen (Not) To Create A GraphQL Server
16mBuild Your GraphQL APIs Faster with Nexus Schema
27mThe ultimate GraphQL client development experience
16mIntro to GraphQL