Skip to content

GitLab

Switch branch/tag
Find file BlameHistoryPermalink
callbacks.md 9.55 KB

Asynchronous idioms: event handlers and callbacks

An example in the browser

JavaScript was designed as a programming language to run on webpages so that users can interact with the content on the page. Some examples of things webpages often do in JavaScript:

  • Show/hide part of the page when the user clicks on a button
  • When the user enters some text in a form, check that it is a valid email address
  • Update the state of a game when the user presses a key
  • Upload data to a server
  • Periodically request updated data from a server

Here is a simple example of JS code (between the <script> and </script> tags) that displays a message when the user clicks on a button:

<html>
  <body>
    <button id='alert-button'>Click me!</button>
    <script>
      // We use the `id` property of the button to access it from JS
      const button = document.getElementById('alert-button')
      button.onclick = () => {
        // You could put any other code you wanted here
        // and it would run each time the button is clicked
        alert('Button was clicked')
      }
    </script>
  </body>
</html>

If you open this HTML file in the browser, it will show a button on the screen. Each time the button is clicked, it displays a message saying that the button was clicked.

How does this work?

Let's tease apart the code responsible for making button clicks display the message. First, note that () => alert('Button was clicked') is a function in JS. When this function is called, it causes the message Button was clicked to appear. Then button.onclick = ... tells the browser "this is the function I want to attach to the click action on button. When a user interacts with the webpage, the browser figures out whether the button was clicked, and if so, it runs the function stored in button.onclick.

How doesn't this work?

Note that nowhere does our code say something like "wait until button is clicked" or "was button clicked yet?". If you've dealt with user input in another language, you might have expected our JS code to handle button clicks to look more like:

// 1.
while (true) {
  button.waitForClick()
  alert('Button was clicked')
}

// or
// 2.
while (true) {
  if (button.wasClicked()) {
    alert('Button was clicked')
  }
}

So why doesn't JS use an interface like one of these?

  1. We call a function button.waitForClick() that asks the browser to pause our code until the user clicks the button. Although this could work in our simple example, what if the user might click either of two buttons and we want to respond to both? If we call button1.waitForClick() and they click button2 instead, we will fail to detect their click.
  2. This approach solves the issue of not being able to respond to multiple events: we could just add an if (button2.wasClicked()) inside the loop. However, it has a more subtle problem. The loop may run billions of times before the user actually clicks the button. Since our program can't actually do anything until the user finally clicks the button, this is an incredible waste of the processor's time.

We see that the "event handler" approach used by JS has two main benefits: it can easily set up handlers for thousands of events at once, and no JS code runs while we wait for an event to occur.

Combining asynchronous events in parallel

Imagine that we would like to wait until several events have all happened before running some code. In this example, we make a combination lock with 4 digits and want to take an action when all the digits are correct:

// Adds a digit to the lock. When the digit changes,
// `changeCallback()` will be called with the new value.
function makeLockDigit(changeCallback) {
  // ... (the full code is at the link above)

  digit.onchange = () => changeCallback(Number(digit.value))
}

// The super secret combination
const COMBO = [1, 2, 3, 4]
// Whether each digit is correct
const digitCorrect = [false, false, false, false]
for (let i = 0; i < COMBO.length; i++) {
  makeLockDigit(value => {
    // Digit i has changed, so store whether it is correct
    digitCorrect[i] = (value === COMBO[i])
    // If all digits are now correct, show a message
    if (digitCorrect.every(correct => correct)) {
      alert('You got the combination')
    }
  })
}

Since any of the 4 digits might be the last one to become correct, we check whether all of the digits are correct whenever each one changes. This is generally the case when we don't know what order some asynchronous events will finish in: whenever each one finishes, it needs to check if it was the last one to finish. There are other ways to store the state of the digits; for example, we could instead track how many are currently correct. This version of the code implements that approach:

// The number of correct digits
let correctCount = 0
for (let i = 0; i < COMBO.length; i++) {
  // Whether this digit was previously correct
  let wasCorrect = false
  makeLockDigit(value => {
    // Compare the digit's actual value to the correct value
    const isCorrect = (value === COMBO[i])
    // Update the number of correct digits if this one changed
    if (!wasCorrect && isCorrect) correctCount++
    else if (wasCorrect && !isCorrect) correctCount--
    wasCorrect = isCorrect
    // If all digits are now correct, show a message
    if (correctCount === COMBO.length) {
      alert('You got the combination')
    }
  })
}

setTimeout()

Another asynchronous JS API is the function setTimeout(), which is used to wait for time to pass. For example, to display a message after 3 seconds (3000 ms), you would write

setTimeout(() => {
  console.log('3 seconds have passed')
}, 3000)

The fact that setTimeout() is asynchronous means it returns immediately, without witing for the time interval. For example, the following code displays Timeout started immediately and Timeout ended after 1 second:

setTimeout(() => {
  console.log('Timeout ended')
}, 1000)
console.log('Timeout started')

As with event handlers, because this API is asynchronous, you can easily wait for multiple intervals of time to pass at the same time. For example, to create events that will finish 1, 2, ..., 10 seconds from now:

for (let i = 1; i <= 10; i++) {
  setTimeout(() => {
    console.log(`${i} seconds passed`)
  }, i * 1000)
}

Combining asynchronous events in sequence

It is also common to have one asynchronous event that depends on the result of another. For example, we might want to wait for the user to enter some text, then send it to a server and wait for the server to respond. In general, this requires us to create the next asynchronous event inside the handler for the previous one.

As a concrete example, let's simulate a "random walk" on a grid. At each step, we randomly move left, right, up, or down, and we count the total number of times we have visited each grid square. Here is a first attempt that only moves 3 times.

// ... (the full code is at the link above)

// Visit a given grid position and
// display the number of times it has been visited
function visit(position) {
  // Don't do anything if we're off the grid
  const row = grid[position.row]
  if (row === undefined) return
  const cell = row[position.col]
  if (cell === undefined) return

  cell.count++
  cell.element.innerText = String(cell.count)
}
// Compute the next position by randomly going left, right, up, or down
function nextPosition(position) {
  const {row, col} = position
  const random = Math.random()
  if (random < 0.25) return {row: row + 1, col}
  else if (random < 0.5) return {row: row - 1, col}
  else if (random < 0.75) return {row, col: col + 1}
  else return {row, col: col - 1}
}

// Start in the middle of the grid and jump every DELAY ms
let position = {row: Math.floor(SIZE / 2), col: Math.floor(SIZE / 2)}
visit(position)
setTimeout(() => {
  // First movement
  position = nextPosition(position)
  visit(position)
  setTimeout(() => {
    // Second movement
    position = nextPosition(position)
    visit(position)
    setTimeout(() => {
      // Third movement
      position = nextPosition(position)
      visit(position)
    }, DELAY)
  }, DELAY)
}, DELAY)

You can see that each sequential delay is represented by nesting another setTimeout() inside the last one. If we want to keep jumping to new squares forever, we need to rewrite this program using a recursive function:

function walkFrom(position) {
  // To perform the walk from the given starting position,
  // visit it and then start from the next position after a delay
  visit(position)
  setTimeout(() => walkFrom(nextPosition(position)), DELAY)
}

// Start in the middle of the grid
walkFrom({row: Math.floor(SIZE / 2), col: Math.floor(SIZE / 2)})

Conclusion

We have seen two examples of asynchronous standard library functions in JavaScript: handling user input and waiting for time. With both of them, we define a function that we want to get called when something happens. This structure is common to any asynchronous function; the function passed in to run when it finishes is refered to as the "callback function". It is up to the browser (or Node.js) to ensure that the callbacks we provide get called at the right times.

We have also looked at some of the primary ways to combine asynchronous actions. Most programs that interact with a user, read/write files, make web requests, etc. can be broken into actions that run sequentially and in parallel. Asynchronous programming makes it simple to run actions in parallel or in sequence, although running an action sequentially in a "loop" requires recursion.