1. Answers

When should I use Arrow Functions?

Some people in the React community warn against using arrow functions -- when used incorrectly, they can cause performance issues.

The thing is, arrow functions can make writing code so much easier. So how can I use them without making my components crawl?

James K Nelson

James is the editor of React Armory, and has been creating things with JavaScript for over 15 years.

Read more by JamesFollow @james_k_nelson on Twitter

Yep, arrow functions are great. They do look nicer, they do take less keystrokes, and they are more fun. But there is a much better reason to use arrow functions…

Arrows prevent this bugs

Arrow functions don’t redefine the value of this within their function body. This makes it a lot easier to predict their behavior when passed as callbacks, and prevents bugs caused by use of this within callbacks.

Here’s an example of the type of bug that arrow functions prevent. At first glance, you may expect the button’s background color to change to red when you click it. But if you actually click the button, you’ll find that nothing changes…

class BrokenButton extends React.Component {
  render() {
    return (
      <button onClick={this.handleClick} style={this.state}>
        Set background to red
      </button>
    )
  }

  handleClick() {
    this.setState({ backgroundColor: 'red' })
  }
}

ReactDOM.render(
  <BrokenButton />,
  document.getElementById('app')
)

Don’t believe this is null? You can check it by inserting alert(this) as the first line of handleClick().

So what’s gone wrong? The problem is that the handleClick method isn’t a method at all! JavaScript doesn’t actually have methods. It only has functions. And the value of this inside those functions has some funny rules. In this case, those rules cause this to be null.

Let’s test your understanding with a little quiz: can you explain the reason that this is null in the example? Spend a few moments figuring it out, then check your answer below.

In a JavaScript method, the value of this has nothing to do with the class on which it was defined. Instead, it depends on the object that it was called upon.

In terms of . syntax, this means that if you call obj.someMethod(), the value of this within someMethod will be obj – regardless of where someMethod was originally defined! If you just call someMethod() without using a ., this will be null.

There are two important exceptions:

  • You can force a specific value of this with a function’s call, apply and bind methods
  • If the function is an => arrow function, this is hard-wired to refer to the value of this at the location where the function was defined

None of these exceptions apply in the above example, and you don’t know how handleClick is called (because it is called within the React’s button component). As such, you have no idea what this will be!

My explanation of this is a bit like high school chemistry; it’s a nice idea but it may give an expert a heart attack.

If all you want to do is use React, it is probably good enough. But otherwise go read the details at MDN.

How’d you go? Don’t worry too much if you’re a little confused; this is confusing, even for the experts. But there is one important thing that you need to remember:

Once you’ve passed a method as a callback, you have no idea how it will be called, and thus no idea what this will be.

That is, unless you use an arrow function.

In an arrow function, this means the same thing within the function body as it does outside of it. Which means that if you use arrow functions within your component’s render or lifecycle methods, they can use this and this.setState with no surprises.

For example, you could fix the above example by replacing the handleClick callback with an arrow function:

class Button extends React.Component {
  render() {
    return (
      <button
        onClick={() => this.setState({ backgroundColor: 'red' })}
        style={this.state}
      >
        Set background to red
      </button>
    )
  }
}

ReactDOM.render(
  <Button />,
  document.getElementById('app')
)

Simple, right? But if you do need help remembering the rule, just remember that arrow functions can’t touch this!

But this old-school classic doesn’t explain why you wouldn’t want to use arrow functions. So let’s continue.

Browser support

Browser support for arrows is currently around 73%, i.e. everything except IE (pun intended). This means that you wouldn’t want to use arrow functions if you’re writing raw JavaScript for IE. But you already know that. And you’re probably transpiling your code anyway.

Performance implications

The biggest problem you can have with arrows is that they’re too easy to define. You may find yourself defining functions where don’t actually need them. But why would too many functions be a problem?

Function definition is expensive

Each time your browser executes a => statement, it needs to create a new function object. This is an expensive operation. In fact, at the time of writing, calling a function is (very roughly) about 100 times faster than creating it.

Of course, this doesn’t actually become a big problem until you’re building a component which needs super-great performance. The only time I’ve personally found this to be an issue is when rendering very long lists, or very large tables.

So don’t panic about this performance cost. If your component is only rendering once or twice on a page, it won’t be a problem. And if you do start rendering enough elements to feel some lag, you can easily factor out the arrows later.

Two functions are never equal to each other

While arrow functions can slow down a component, the bigger problem is that they can slow down child components. Which isn’t to say that they will slow down child components; they probably won’t. That is, they won’t unless those child components are optimized with PureComponent or shouldComponentUpdate.

So what’s the problem? Two identical arrow functions are not equal. You can see this for yourself in the example below, where the functions a and b are equivalent, but not equal:

const a = (x) => x,
      b = (x) => x

ReactDOM.render(
  <div>
    <h3>Are <code>a</code> and <code>b</code> equal by <code>==</code>?</h3>
    <p>{a == b ? 'Yes!' : 'No :('}</p>
  </div>,
  document.getElementById('app')
)

Optimizations based on PureComponent and shouldComponentUpdate work by skipping work when no props have changed. Learn more at the React docs.

If you use arrow functions within render, each call to render will create new function objects. If you then pass these functions to child elements via props, optimizations based on PureComponent or shouldComponentUpdate will fail (as the arrow function props will change on every render).

You can see this in action in the following example where the <PropChangeCounter /> component prints the number of times that each of its props have changed. Notice how constant doesn’t change, while onChange does. This is because each call to render creates a new arrow function!

import PropChangeCounter from 'react-armory-prop-change-counter'

class App extends React.Component {
  constructor(props) {
    super(props)
    this.state = { email: '' }
  }
  render() {
    return <div>
      <input
        placeholder="Email"
        value={this.state.email}
        onChange={e => this.setState({ email: e.target.value })}
      />
      <PropChangeCounter
        constant={"this doesn't change"}
        value={this.state.email}
        onChange={e => this.setState({ email: e.target.value })}
      />
    </div>
  }
}

ReactDOM.render(<App />, document.getElementById('app'))

To reiterate, arrow functions do not affect performance. The problems only arise when you re-create the same arrow function on each render. Which means there is a simple fix…

Defining arrows once

If performance is important for your component, you’ll want to define any arrow functions just once. And the obvious time to do this is when the component loads.

One way of accomplishing this would be to define all your arrow functions in the constructor. Of course, this would become unwieldy for any reasonably complex component, and wouldn’t really have any benefit over simply using Function.prototype.bind.

However, if you’re using Babel (or create-react-app) to build your source, you have another option: setting arrow functions as class fields (or arrow function methods).

For example, you could fix the bug in the first example by redefining handleClick as an arrow function method:

This example isn't live, as React Armory's live examples don't yet support arrow functions methods. Sorry :-(

class Button extends React.Component {
  render() {
    return (
      <button onClick={this.handleClick} style={this.state}>
        Set background to red
      </button>
    )
  }

  // Note: this syntax is not yet part of JavaScript proper, but is slated
  // for inclusion in the next version. It should already work with Babel.
  handleClick = () => {
    this.setState({ backgroundColor: 'red' })
  }
}

This gives you the best of both worlds. You get the improved performance from only defining functions once, but you still get the simple method-like syntax.

Don’t Panic

A wise man once said that “premature optimization is the root of all evil”. He was probably exaggerating a little bit, but the point he was trying to make is a good one.

You won’t introduce any bugs by using too many arrow functions. You probably will introduce bugs but not using enough.

So to wrap things up, here are the four rules of arrow functions:

  1. If your environment supports arrow function methods, use them for everything.

  2. Don’t use arrow functions for React components; it will make debugging hard.

  3. Do use arrow functions in render if they’re helpful.

  4. Move functions out of render (and any methods it calls) if you need to optimize a component’s performance.