Nikolay Grozev

# Introduction

I recently started learning about the React hooks API and I was amazed by how expressive it is. Hooks allow me to rewrite tens of lines of boilerplate code with just a few lines. Unfortunately, this convenience comes at a cost. I found that some more advanced hooks like useCallback and useMemo are hard to learn and appear counter-intuitive at first.

In this article, I’ll demonstrate with a few simple examples why we need these hooks and when and how to use them. This is not an introduction to hooks, and you must be familiar with the useState hook to follow.

# The Problem

Let’s look at the following simple app. It displays 2 counters and allows the user to increment them with 2 buttons. We’ll create 2 functions increment1 and increment2 and assign them to the buttons’ on-click event handlers. Let’s also keep track of how many such functions are created while the user clicks the buttons:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
import React, { useState } from 'react';

// Keeps track of all created functions during the app's life
const functions: Set<any> = new Set();

const App = () => {
const [c1, setC1] = useState(0);
const [c2, setC2] = useState(0);

const increment1 = () => setC1(c1 + 1);
const increment2 = () => setC2(c2 + 1);

// Register the functions so we can count them

return (<div>
<div> Counter 1 is {c1} </div>
<div> Counter 2 is {c2} </div>
<br/>
<div>
<button onClick={increment1}>Increment Counter 1</button>
<button onClick={increment2}>Increment Counter 2</button>
</div>
<br/>
<div> Newly Created Functions: {functions.size - 2} </div>
</div>)
}


When we run the app and start clicking the buttons we observe something interesting. For every click of a button there are 2 newly created functions!

In other words, at every re-render we’re creating 2 new functions, which is excessive. If we increment c1, why do we need to recreate the increment2 function? This is not just about memory. What if we pass increment2 as a property to a future child component? It will needlessly re-render on evey change of c1, because the function instance of increment2 will also change. This can quickly become a performance issue.

One solution would be to move the two functions outside of the the App functional component. Unfortunately, this wouldn’t work because they use the state variables from App’s scope.

# Naive solution - Why dependencies matter

This is where the useCallback hook comes in. It takes as an arguement a function and returns a cached/memoized version of it. It also takes a second parameter which will cover later. Let’s rewrite with useCallBack:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
const App = () => {
const [c1, setC1] = useState(0);
const [c2, setC2] = useState(0);

// Cache/memoize the functions - do not create new ones on every rerender
// Let's ignore the dependencies for now - i.e. use []
const increment1 = useCallback(() => setC1(c1 + 1), []);
const increment2 = useCallback(() => setC2(c2 + 1), []);

// Register the functions so we can count them

return (<div>
<div> Counter 1 is {c1} </div>
<div> Counter 2 is {c2} </div>
<br/>
<div>
<button onClick={increment1}>Increment Counter 1</button>
<button onClick={increment2}>Increment Counter 2</button>
</div>
<br/>
<div> Newly Created Functions: {functions.size - 2} </div>
</div>)
}


When we re-run the app, we notice that we’ve introduced a bug. We can keep clicking the increment buttons, but the counters’ values never increase beyond 1. There’re no newly created functions, and this is the cause of the issue.

The useCallback hook has created single cached versions of the two functions, which encapsulate the initial values of c1 and c2. When App re-renders with different values for c1 and c2, useCallback returns the previous versions of the functions which keep the old values of c1 and c2 from the first rendering.

We need to tell useCallback to create new cached versions of the functions for every change of values of c1 and c2 that they depend on.

# Dependencies

This is where the second arguement of useCallback comes in. It is an array of values, which represents the dependencies of the cache. On any two subsequent re-renders, useCallback will return the same cached function instance if the values of the dependencies are equal:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
const App = () => {
const [c1, setC1] = useState(0);
const [c2, setC2] = useState(0);

// The useCallback caches depend on the values of c1 and c2
const increment1 = useCallback(() => setC1(c1 + 1), [c1]);
const increment2 = useCallback(() => setC2(c2 + 1), [c2]);

// Register the functions so we can count them

return (<div>
<div> Counter 1 is {c1} </div>
<div> Counter 2 is {c2} </div>
<br/>
<div>
<button onClick={increment1}>Increment Counter 1</button>
<button onClick={increment2}>Increment Counter 2</button>
</div>
<br/>
<div> Newly Created Functions: {functions.size - 2} </div>
</div>)
}


Now we can see that the number of newly created functions is the same as the number of re-renders, which is twice as good as the original example without useCallback. In other words, we only create new callbacks, if the part of the closure they use (i.e. their dependencies) has changed since the previous rendering.

A really useful feature of useCallback is that it returns the same function instance if the depencies don’t change. Hence we can use it in the dependecy lists of other hooks. For example, let’s create a cached/memoized function which increments both counters:

1
2
3
4
5
6
7
8
const increment1 = useCallback(() => setC1(c1 + 1), [c1]);
const increment2 = useCallback(() => setC2(c2 + 1), [c2]);

// Can depend on [c1, c2] instead, but it would be brittle
const incrementBoth = useCallback(() => {
increment1();
increment2();
}, [increment1, increment2]);


The new incrementBoth function transitively depends on c1 and c2. We could write useCallback(... ,[c1, c2]) and that would work. However, this is a very brittle approach! If we changed the dependencies of increment1 or increment2, we would have to remember to change the dependencies of incrementBoth.

Since the references of increment1 and increment2 won’t change unless their dependencies change, we could use them instead. Transitive dependencies can be ignored! This makes for a straightforward rule:

Each function declared within a functional component’s scope must be memoized/cached with useCallback. If it references functions or other variables from the component scope it should list them in its depency list.

This rule can be enforced by a linter which checks that your useCallback cache dependenices are consistent.

# Two similar hooks - useCallback and useMemo

React introduces another similar hook called useMemo. It has similar signature, but works differently. Unlike useCallback, which caches the provided function instance, useMemo invokes the provided function and caches its result.

In other words useMemo caches a computed value. This is usefull when the computation requires significant resources and we don’t want to repeat it on every re-render, as in this example:

1
2
3
4
5
6
const [c1, setC1] = useState(0);
const [c2, setC2] = useState(0);

// This value will not be recomputed between re-renders
// unless the value of c1 changes
const sinOfC1: number = useMemo(() => Math.sin(c1) , [c1])


Just as with useCallback, the values returned by useMemo can be used as other hooks’ dependencies.

As an interesting aside, useMemo can cache a function value too. In other words, it is a generalised version of useCallback and can replace it as in the following example

1
2
3
4
5
6
// Some function ...
const f = () => { ... }

// The following are functionally equivalent
const callbackF = useCallback(f, [])
const callbackF = useMemo(() => f, [])