scheduler的处理

流程：当依赖调用notify的时候会顺序调用watcher中的update，因为渲染函数中的wacher不包含，options所以最终会调用queueWatcher， 而它最终通过各种判断就调用了flushSchedulerQueue

function flushSchedulerQueue () {
  currentFlushTimestamp = getNow()
  flushing = true
  let watcher, id

  // Sort queue before flush.
  // This ensures that:
  // 1. Components are updated from parent to child. (because parent is always
  //    created before the child)
  // 2. A component's user watchers are run before its render watcher (because
  //    user watchers are created before the render watcher)
  // initState在_init之前执行
  // 3. If a component is destroyed during a parent component's watcher run,
  //    its watchers can be skipped.
  // 从小到大，排列watcher
  queue.sort((a, b) => a.id - b.id)

  // do not cache length because more watchers might be pushed
  // as we run existing watchers
  // 不要缓存length， 因为queue在操作过程中可能继续增加
  for (index = 0; index < queue.length; index++) {
    watcher = queue[index]
    if (watcher.before) {
      // 在渲染中调用， beforeUpdate
      watcher.before()
    }
    id = watcher.id
    has[id] = null
    watcher.run()
    // in dev build, check and stop circular updates.
    if (process.env.NODE_ENV !== 'production' && has[id] != null) {
      circular[id] = (circular[id] || 0) + 1
      if (circular[id] > MAX_UPDATE_COUNT) {
        warn(
          'You may have an infinite update loop ' + (
            watcher.user
              ? `in watcher with expression "${watcher.expression}"`
              : `in a component render function.`
          ),
          watcher.vm
        )
        break
      }
    }
  }