Vue3 computed and watch source code analysis

computed

Computed and watch are often asked about their differences in interviews, so let’s take a look at their specific implementation from the source code implementation.

// packages/reactivity/src/computed.ts
export function computed<T>(
  getterOrOptions: ComputedGetter<T> | WritableComputedOptions<T>,
  debugOptions?: DebuggerOptions,
  isSSR = false
) {
  let getter: ComputedGetter<T>
  let setter: ComputedSetter<T>
  const onlyGetter = isFunction(getterOrOptions)
  if (onlyGetter) {
    getter = getterOrOptions
    setter = __DEV__
      ? () => {
          console.warn(&#39;Write operation failed: computed value is readonly&#39;)
        }
      : NOOP
  } else {
    getter = getterOrOptions.get
    setter = getterOrOptions.set
  }
  // new ComputedRefImpl
  const cRef = new ComputedRefImpl(getter, setter, onlyGetter || !setter, isSSR)
  if (__DEV__ && debugOptions && !isSSR) {
    cRef.effect.onTrack = debugOptions.onTrack
    cRef.effect.onTrigger = debugOptions.onTrigger
  }
  // 返回ComputedRefImpl实例
  return cRef as any
}

You can see that computed internally only processes getters and setters first, and then returns a new ComputedRefImpl. If you know the implementation of ref API, you can find that their implementations have many similarities

ComputedRefImpl

// packages/reactivity/src/computed.ts
export class ComputedRefImpl<T> {
  public dep?: Dep = undefined // 存储effect的集合
  private _value!: T
  public readonly effect: ReactiveEffect<T>
  public readonly __v_isRef = true
  public readonly [ReactiveFlags.IS_READONLY]: boolean = false
  public _dirty = true // 是否需要重新更新value
  public _cacheable: boolean
  constructor(
    getter: ComputedGetter<T>,
    private readonly _setter: ComputedSetter<T>,
    isReadonly: boolean,
    isSSR: boolean
  ) {
    // 创建effect
    this.effect = new ReactiveEffect(getter, () => {
      // 调度器执行 重新赋值_dirty为true
      if (!this._dirty) {
        this._dirty = true
        // 触发effect
        triggerRefValue(this)
      }
    })
    // 用于区分effect是否是computed
    this.effect.computed = this
    this.effect.active = this._cacheable = !isSSR
    this[ReactiveFlags.IS_READONLY] = isReadonly
  }
  get value() {
    // the computed ref may get wrapped by other proxies e.g. readonly() #3376
    // computed ref可能被其他代理包装，例如readonly() #3376
    // 通过toRaw()获取原始值
    const self = toRaw(this)
    // 收集effect
    trackRefValue(self)
    // 如果是脏的，重新执行effect.run()，并且将_dirty设置为false
    if (self._dirty || !self._cacheable) {
      self._dirty = false
      // run()方法会执行getter方法 值会被缓存到self._value
      self._value = self.effect.run()!
    }
    return self._value
  }
  set value(newValue: T) {
    this._setter(newValue)
  }
}

You can see that the get implementation of ComputedRefImplget is basically the same as the get implementation of ref (please read the previous chapter if you are not familiar with ref implementation). The only difference is the judgment of _dirty value, which is also what we often call computed The value will be cached, so how does computed know that the value needs to be updated?

You can see that in the computed constructor, a relationship between a getter and its internal responsive data will be established. This is the same as the relationship between our component update function and the responsive data, so the responsiveness related to the getter When the data is modified, the scheduler corresponding to the getter effect will be triggered. Here, _dirty will be set to true and the collected effect will be executed (this is usually the effect updated by the function collected in the get), and then it will go Execute the function update function, which will trigger the computed get again. At this time, dirty has been set to true, and the getter will be re-executed to obtain the new value and return, and the value will be cached in _vlaue.

Summary:

So computed has two layers of responsive processing. One layer is the relationship between computed.value and the effect of the function (similar to the implementation of ref). It is the relationship between computed getter and responsive data.

Note: If you are careful enough, you will find that there may be a sequence problem between the effect triggering of the function update function and the triggering of the computed getter effect. If there is a reactive data a that not only exists in the getter, but is also accessed earlier than the getter in the function render, then the effect of the update function in the dep corresponding to a will be collected earlier than the effect of the getter. If at this time a is Change, the effect of the update function will be executed first, then when the render function accesses computed.value, it will find that _dirty is still false, because the effect of the getter has not been executed, so it will still be the old value at this time. The processing of this in vue3 is that when executing effects, the effect corresponding to computed will be executed first (also mentioned in the previous chapter):

// packages/reactivity/src/effect.ts
export function triggerEffects(
  dep: Dep | ReactiveEffect[],
  debuggerEventExtraInfo?: DebuggerEventExtraInfo
) {
  // spread into array for stabilization
  const effects = isArray(dep) ? dep : [...dep]
  // computed的effect会先执行
  // 防止render获取computed值得时候_dirty还没有置为true
  for (const effect of effects) {
    if (effect.computed) {
      triggerEffect(effect, debuggerEventExtraInfo)
    }
  }
  for (const effect of effects) {
    if (!effect.computed) {
      triggerEffect(effect, debuggerEventExtraInfo)
    }
  }
}

watch

watch is simpler than computed because He only needs to establish the relationship between the getter and the responsive data, call the callback passed by the user when the responsive data changes, and pass in the old and new values

// packages/runtime-core/src/apiWatch.ts
export function watch<T = any, Immediate extends Readonly<boolean> = false>(
  source: T | WatchSource<T>,
  cb: any,
  options?: WatchOptions<Immediate>
): WatchStopHandle {
  if (__DEV__ && !isFunction(cb)) {
    warn(...)
  }
  // watch 具体实现
  return doWatch(source as any, cb, options)
}

function doWatch(
  source: WatchSource | WatchSource[] | WatchEffect | object,
  cb: WatchCallback | null,
  { immediate, deep, flush, onTrack, onTrigger }: WatchOptions = EMPTY_OBJ
): WatchStopHandle {
  if (__DEV__ && !cb) {
    ...
  }
  const warnInvalidSource = (s: unknown) => {
    warn(...)
  }
  const instance =
    getCurrentScope() === currentInstance?.scope ? currentInstance : null
  // const instance = currentInstance
  let getter: () => any
  let forceTrigger = false
  let isMultiSource = false
  // 根据不同source 创建不同的getter函数
  // getter 函数与computed的getter函数作用类似
  if (isRef(source)) {
    getter = () => source.value
    forceTrigger = isShallow(source)
  } else if (isReactive(source)) {
    // source是reactive对象时 自动开启deep=true
    getter = () => source
    deep = true
  } else if (isArray(source)) {
    isMultiSource = true
    forceTrigger = source.some(s => isReactive(s) || isShallow(s))
    getter = () =>
      source.map(s => {
        if (isRef(s)) {
          return s.value
        } else if (isReactive(s)) {
          return traverse(s)
        } else if (isFunction(s)) {
          return callWithErrorHandling(s, instance, ErrorCodes.WATCH_GETTER)
        } else {
          __DEV__ && warnInvalidSource(s)
        }
      })
  } else if (isFunction(source)) {
    if (cb) {
      // getter with cb
      getter = () =>
        callWithErrorHandling(source, instance, ErrorCodes.WATCH_GETTER)
    } else {
      // no cb -> simple effect
      getter = () => {
        if (instance && instance.isUnmounted) {
          return
        }
        if (cleanup) {
          cleanup()
        }
        return callWithAsyncErrorHandling(
          source,
          instance,
          ErrorCodes.WATCH_CALLBACK,
          [onCleanup]
        )
      }
    }
  } else {
    getter = NOOP
    __DEV__ && warnInvalidSource(source)
  }
  // 2.x array mutation watch compat
  // 兼容vue2
  if (__COMPAT__ && cb && !deep) {
    const baseGetter = getter
    getter = () => {
      const val = baseGetter()
      if (
        isArray(val) &&
        checkCompatEnabled(DeprecationTypes.WATCH_ARRAY, instance)
      ) {
        traverse(val)
      }
      return val
    }
  }
  // 深度监听
  if (cb && deep) {
    const baseGetter = getter
    // traverse会递归遍历对象的所有属性 以达到深度监听的目的
    getter = () => traverse(baseGetter())
  }
  let cleanup: () => void
  // watch回调的第三个参数 可以用此注册一个cleanup函数 会在下一次watch cb调用前执行
  // 常用于竞态问题的处理
  let onCleanup: OnCleanup = (fn: () => void) => {
    cleanup = effect.onStop = () => {
      callWithErrorHandling(fn, instance, ErrorCodes.WATCH_CLEANUP)
    }
  }
  // in SSR there is no need to setup an actual effect, and it should be noop
  // unless it's eager or sync flush
  let ssrCleanup: (() => void)[] | undefined
  if (__SSR__ && isInSSRComponentSetup) {
    // ssr处理 ...
  }
  // oldValue 声明 多个source监听则初始化为数组
  let oldValue: any = isMultiSource
    ? new Array((source as []).length).fill(INITIAL_WATCHER_VALUE)
    : INITIAL_WATCHER_VALUE
  // 调度器调用时执行
  const job: SchedulerJob = () => {
    if (!effect.active) {
      return
    }
    if (cb) {
      // watch(source, cb)
      // 获取newValue
      const newValue = effect.run()
      if (
        deep ||
        forceTrigger ||
        (isMultiSource
          ? (newValue as any[]).some((v, i) =>
              hasChanged(v, (oldValue as any[])[i])
            )
          : hasChanged(newValue, oldValue)) ||
        (__COMPAT__ &&
          isArray(newValue) &&
          isCompatEnabled(DeprecationTypes.WATCH_ARRAY, instance))
      ) {
        // cleanup before running cb again
        if (cleanup) {
          // 执行onCleanup传过来的函数
          cleanup()
        }
        // 调用cb 参数为newValue、oldValue、onCleanup
        callWithAsyncErrorHandling(cb, instance, ErrorCodes.WATCH_CALLBACK, [
          newValue,
          // pass undefined as the old value when it's changed for the first time
          oldValue === INITIAL_WATCHER_VALUE
            ? undefined
            : isMultiSource && oldValue[0] === INITIAL_WATCHER_VALUE
            ? []
            : oldValue,
          onCleanup
        ])
        // 更新oldValue
        oldValue = newValue
      }
    } else {
      // watchEffect
      effect.run()
    }
  }
  // important: mark the job as a watcher callback so that scheduler knows
  // it is allowed to self-trigger (#1727)
  job.allowRecurse = !!cb
  let scheduler: EffectScheduler
  if (flush === 'sync') {
    // 同步更新 即每次响应式数据改变都会回调一次cb 通常不使用
    scheduler = job as any // the scheduler function gets called directly
  } else if (flush === 'post') {
    // job放入pendingPostFlushCbs队列中
    // pendingPostFlushCbs队列会在queue队列执行完毕后执行 函数更新effect通常会放在queue队列中
    // 所以pendingPostFlushCbs队列执行时组件已经更新完毕
    scheduler = () => queuePostRenderEffect(job, instance && instance.suspense)
  } else {
    // default: 'pre'
    job.pre = true
    if (instance) job.id = instance.uid
    // 默认异步更新 关于异步更新会和nextTick放在一起详细讲解
    scheduler = () => queueJob(job)
  }
  // 创建effect effect.run的时候建立effect与getter内响应式数据的关系
  const effect = new ReactiveEffect(getter, scheduler)
  if (__DEV__) {
    effect.onTrack = onTrack
    effect.onTrigger = onTrigger
  }
  // initial run
  if (cb) {
    if (immediate) {
      // 立马执行一次job
      job()
    } else {
      // 否则执行effect.run() 会执行getter 获取oldValue
      oldValue = effect.run()
    }
  } else if (flush === 'post') {
    queuePostRenderEffect(
      effect.run.bind(effect),
      instance && instance.suspense
    )
  } else {
    effect.run()
  }
  // 返回一个取消监听的函数
  const unwatch = () => {
    effect.stop()
    if (instance && instance.scope) {
      remove(instance.scope.effects!, effect)
    }
  }
  if (__SSR__ && ssrCleanup) ssrCleanup.push(unwatch)
  return unwatch
}

The above is the detailed content of Vue3 computed and watch source code analysis. For more information, please follow other related articles on the PHP Chinese website!