The 3 Skills That Helped Me Become a Better Software Engineer, Continued

Reference

• Developing good mental models:
  • Programming language.
  • A program.
  • A compiler.
  • Type systems.
  • Functional programming, imperative programming, logical programming.
  • VMs.
• Front-end oriented mental models:
  • Bundling
  • Tree-shaking
  • Change detection
  • Observable objects
  • Virtual DOM
  • CQRS/event sourcing/Redux.
• Developing mental models
  • Functional programming? Elm
  • Build it yourself.

Taming Dynamic Data in TypeScript

Reference

• JSON.parse type is any.
• Use unknown, a type-safe counterpart to any.
• TS doesn’t do validation at runtime to make sure your assertion is correct.
• How do other languages handle this? Rust can automatically generate decoders for a given struct at build-time. Elm has a library to manually construct these decoders.
• Manual decoding:
  • io-ts and runtypes.
• runtypes

import { Record, String, Number, Static } from 'runtypes'

const UserRuntype = Record({
    name: String,
    age: Number
})

type User = Static<typeof UserRuntype>

Designing the perfect Typescript schema validation library

Reference

• Joi - doesn’t support static type inference.
• Yup - supports type inference, but the typings are wrong.
• io-ts - super functional/obscure, but good library.
• Zod
  • Use TS generic interface to statically infer the types of your schemas.
  • Eliminate the need to keep static types and runtime validators in sync by hand.
  • Tricks
    • Fields are required unless explicitly marked as optional.
    • Schemas are immutable.
    • Parse, don’t validate!
  • Extract TS of any schema with z.TypeOf<>.