We recently adopted some clever uses of TypeScript's infer keyword in our team and gathered plenty of experience. I'm sharing it here as a reference, hoping it helps anyone doing similar work.

Core Concepts

Here is a complete example:

type UnwrapPromise<T> = T extends Promise<infer U> ? U : T

async function fetchUser(id: string) {
  const res = await fetch(`/api/users/${id}`)
  return res.json() as Promise<{ id: string; name: string; email: string }>
}

type User = UnwrapPromise<ReturnType<typeof fetchUser>>

// 类型安全的事件系统
interface EventMap {
  login: { userId: string; timestamp: number }
  logout: { userId: string }
}

class TypedEmitter<T extends Record<string, any>> {
  private handlers = new Map<keyof T, Set<Function>>()
  on<K extends keyof T>(event: K, handler: (payload: T[K]) => void) {
    if (!this.handlers.has(event)) this.handlers.set(event, new Set())
    this.handlers.get(event)!.add(handler)
  }
  emit<K extends keyof T>(event: K, payload: T[K]) {
    this.handlers.get(event)?.forEach(h => h(payload))
  }
}

Pay attention to edge-case handling — that is critical in production.

In-Depth Analysis

The key is to understand the core logic:

type DeepPartial<T> = T extends object ? { [P in keyof T]?: DeepPartial<T[P]> } : T

interface AppConfig {
  api: { baseUrl: string; timeout: number; retries: number }
  ui: { theme: 'light' | 'dark'; language: string; pageSize: number }
}

type PartialConfig = DeepPartial<AppConfig>

function mergeConfig(defaults: AppConfig, overrides: PartialConfig): AppConfig {
  const result = { ...defaults }
  for (const key of Object.keys(overrides) as (keyof AppConfig)[]) {
    if (overrides[key] && typeof overrides[key] === 'object') {
      result[key] = { ...defaults[key], ...overrides[key] } as any
    }
  }
  return result
}

Performance optimization has to be tied to the specific scenario; not every case needs over-optimization.

Implementation Experience

Here are some ways we can improve on this:

type UnwrapPromise<T> = T extends Promise<infer U> ? U : T

async function fetchUser(id: string) {
  const res = await fetch(`/api/users/${id}`)
  return res.json() as Promise<{ id: string; name: string; email: string }>
}

type User = UnwrapPromise<ReturnType<typeof fetchUser>>

// 类型安全的事件系统
interface EventMap {
  login: { userId: string; timestamp: number }
  logout: { userId: string }
}

class TypedEmitter<T extends Record<string, any>> {
  private handlers = new Map<keyof T, Set<Function>>()
  on<K extends keyof T>(event: K, handler: (payload: T[K]) => void) {
    if (!this.handlers.has(event)) this.handlers.set(event, new Set())
    this.handlers.get(event)!.add(handler)
  }
  emit<K extends keyof T>(event: K, payload: T[K]) {
    this.handlers.get(event)?.forEach(h => h(payload))
  }
}

This approach has been running stably in production for over half a year and is proven in practice.

Tuning Strategies

Let's start with a basic implementation:

type DeepPartial<T> = T extends object ? { [P in keyof T]?: DeepPartial<T[P]> } : T

interface AppConfig {
  api: { baseUrl: string; timeout: number; retries: number }
  ui: { theme: 'light' | 'dark'; language: string; pageSize: number }
}

type PartialConfig = DeepPartial<AppConfig>

function mergeConfig(defaults: AppConfig, overrides: PartialConfig): AppConfig {
  const result = { ...defaults }
  for (const key of Object.keys(overrides) as (keyof AppConfig)[]) {
    if (overrides[key] && typeof overrides[key] === 'object') {
      result[key] = { ...defaults[key], ...overrides[key] } as any
    }
  }
  return result
}

This snippet shows the basic usage. In real projects you'll also need to account for error handling and edge cases.

Important Notes

Building on this, we can optimize further:

type UnwrapPromise<T> = T extends Promise<infer U> ? U : T

async function fetchUser(id: string) {
  const res = await fetch(`/api/users/${id}`)
  return res.json() as Promise<{ id: string; name: string; email: string }>
}

type User = UnwrapPromise<ReturnType<typeof fetchUser>>

// 类型安全的事件系统
interface EventMap {
  login: { userId: string; timestamp: number }
  logout: { userId: string }
}

class TypedEmitter<T extends Record<string, any>> {
  private handlers = new Map<keyof T, Set<Function>>()
  on<K extends keyof T>(event: K, handler: (payload: T[K]) => void) {
    if (!this.handlers.has(event)) this.handlers.set(event, new Set())
    this.handlers.get(event)!.add(handler)
  }
  emit<K extends keyof T>(event: K, payload: T[K]) {
    this.handlers.get(event)?.forEach(h => h(payload))
  }
}

This pattern is very practical in large projects and can significantly lower maintenance cost.

Summary

• Before using it in production, make sure to verify compatibility.
• In team collaboration, conventions and documentation matter more than the technology itself.
• Keep an eye on the community; technical solutions need continuous iteration.