How to add fuzzy search to your React app using Fuse.js

📆 · ⏳ 7 min read · ·

Introduction

Often while building applications we want to add some sort of searching feature to help our users to find the thing they are looking for. In this article, we will learn how to build one in react ↗️ using Fuse.js ↗️ and make it type-safe with Typescript.

If you are interested only in the code part then skip right here

Dead simple searching method

Before we go ahead, I want to make it clear that in this article we would be using an external library to help us with the search but if you want the simplest approach to have a searching-like feature then you can just go ahead and use includes ↗️ method on the array to have straight forward search.

Of course, it has its limitations, for starters, it does exact string search without fuzzy search and it’s not text case insensitive by default but it’s definitely the simplest solution to have and I wanted to point it out before we go ahead.

Why Fuse.js?

Fuse.js is a powerful, lightweight fuzzy-search library, with zero dependencies.

Two main reasons I went ahead and incorporated Fuse.js search for my blog page as well is that it allows me to do fuzzy-search and that it has zero dependencies and the overall bundle size ↗️ is quite small.

So what does fuzzy search means? Well, fuzzy searching (more formally known as approximate string matching) is the technique of finding strings that are approximately equal to a given pattern.

You can read more in detail about fuzzy search on Wikipedia ↗️

Fuse.js uses the concept of relevance score to rank the results. This score is determined by three factors:

  • Fuzziness score
  • Key weight
  • Field-length norm

You can read more about these concepts here ↗️

So go ahead and install Fuse.js library in your project with package manager of your choice. I’ll be using yarn so here’s the command for that.

Terminal window
yarn add fuse.js

useSearch hook

Now we know which library we are going to use for searching, let’s build a reusable react hook that will help us build this functionality.

Let’s create a file called useSearch.ts and lay out the input props and the output for this hook

import { useState } from 'react'
interface IUseSearchProps {
dataSet: any[]
keys: string[]
}
export default function useSearch(
{ dataSet, keys }: IUseSearchProps
) {
const [searchValue, setSearchValue] = useState('')
const results = [] // we will calculate this soon
return {
results,
searchValue,
setSearchValue
}
}

We are taking the dataSet as the prop which is the list of data in which we want to perform the search, keys as another prop which will consist of the keys on which we want Fuse.js to perform the search.

Next, let’s check how we can import the library and instantiate it with our dataSet

import { useMemo } from 'react';
import Fuse from 'fuse.js';
// ...
export default function useSearch({ dataSet, keys }: IUseSearchProps) {
// ...
const fuse = useMemo(() => {
const options = {
includeScore: true,
keys,
};
return new Fuse(dataSet, options);
}, [dataSet, keys]);
// ...
}

We are importing the Fuse.js library using es6 import and instantiating it with the dataSet and the options. For options, we are adding the keys which we imported as props and setting includeScore to true. We would use this score to determine which results we want to include in the search output. Read more about all the options supported by Fuse.js here ↗️.

As an additional tip, we are using useMemo hook here so that we do not create the fuse instance again if the dataSet and the keys are the same.

Now let’s use this fuse instance to calculate the results for us based on the searchValue.

import { useState, useMemo } from 'react';
import Fuse from 'fuse.js';
// ...
const SCORE_THRESHOLD = 0.4;
export default function useSearch({ dataSet, keys }: IUseSearchProps) {
const [searchValue, setSearchValue] = useState('');
// ...
const results = useMemo(() => {
if (!searchValue) return dataSet;
const searchResults = fuse.search(searchValue);
return searchResults
.filter((fuseResult) => fuseResult.score < SCORE_THRESHOLD)
.map((fuseResult) => fuseResult.item);
}, [fuse, searchValue, dataSet]);
// ...
}

For results, we are again using the useMemo hook to recalculate the results whenever the searchValue changes (and we are adding fuse and dataSet as they are used in this hook and hence are dependencies for it).

In this useMemo hook we will first check if we have anything in searchValue or not, if not then we want to simply return the dataSet (or whatever logic you have for showing initial values from dataSet).

But if we have some searchValue then we will use the fuse.search function with this searchValue. This gives us back the results with the information about the score and the actual item in the dataSet.

Now based on the score value we will filter out the results which do not match our threshold.

In Fuse.js a score closer to 0 means it’s a perfect match and a score closer to 1 means it’s far away from the search value.

Once we have filtered out the results, we would map over and return the item which is the individual item data in the dataSet.

So if we wire it up all together then the file should look like this

import { useState, useMemo } from 'react'
import Fuse from 'fuse.js'
interface IUseSearchProps {
dataSet: any[]
keys: string[]
}
const SCORE_THRESHOLD = 0.4
export default function useSearch({ dataSet, keys }: IUseSearchProps) {
const [searchValue, setSearchValue] = useState('')
const fuse = useMemo(() => {
const options = {
includeScore: true,
keys,
}
return new Fuse(dataSet, options)
}, [dataSet, keys])
const results = useMemo(() => {
if (!searchValue) return dataSet
const searchResults = fuse.search(searchValue)
return searchResults
.filter((fuseResult) => fuseResult.score < SCORE_THRESHOLD)
.map((fuseResult) => fuseResult.item)
}, [fuse, searchValue, dataSet])
return {
searchValue,
setSearchValue,
results,
}
}

Usage

The usage is very easy now, in whichever component we want to add search feature we can import the hook and use it as mentioned below

import useSearch from 'hooks/useSearch';
// ...
const PostList = (posts) => {
const { results, searchValue, setSearchValue } = useSearch({
dataSet: posts,
keys: ['title', 'summary', 'tags'],
});
// Search bar here
// Render the results accordingly
};
export default PostList;

Type-Safe

Our hook is functional and ready to use, just one thing, the hook does not have information about what is the type of the dataSet and hence cannot infer what would be the return type of the hook as well.

We’ll fix this by using Typescript Generics ↗️ to help us with just that.

The type-safe version of this would look like this

import { useState, useMemo } from 'react'
import Fuse from 'fuse.js'
interface IUseSearchProps<T> {
dataSet: T[]
keys: string[]
}
const SCORE_THRESHOLD = 0.4
export default function useSearch<T>({ dataSet, keys }: IUseSearchProps<T>) {
const [searchValue, setSearchValue] = useState('')
const fuse = useMemo(() => {
const options = {
includeScore: true,
keys,
}
return new Fuse(dataSet, options)
}, [dataSet, keys])
const results = useMemo(() => {
if (!searchValue) return dataSet
const searchResults = fuse.search(searchValue)
return searchResults
.filter((fuseResult) => fuseResult.score < SCORE_THRESHOLD)
.map((fuseResult) => fuseResult.item)
}, [fuse, searchValue, dataSet])
return {
searchValue,
setSearchValue,
results,
}
}

Notice how we are using the T to typecast the dataSet value in our hook. Now our hook is type-safe and our usage accordingly will change like this

import useSearch from 'hooks/useSearch'
// ...
interface Post {
title: string
summary: string
tags: string[]
// ... other fields
}
const PostList = (posts) => {
const { results, searchValue, setSearchValue } = useSearch<Post>({
dataSet: posts,
keys: ['title', 'summary', 'tags'],
})
// Search bar here
// Render the results accordingly
}
export default PostList;

With this the value of our results is type-safe and typescript won’t be mad anymore.

Hope you found this helpful. See you in another one 👋🏽

You may also like

  • What is GPG and why you should start using it

    GPG is a tool that allows you to encrypt and sign your data and communications. In this post, I will explain what GPG is and why you should start using it in your workflows if you aren't already.

  • Ansible — Configuration as Code for building up my Homelab

    I have been using Ansible to manage my homelab infrastructure. It's a great tool to automate the deployment and configuration of servers and services. Let me share with you how I use Ansible to manage my homelab.

  • AdGuard Home — Network Wide Ad Blocking in your Homelab

    Let's talk about AdGuardHome, a network-wide ad blocking software that you can run in your homelab. It's a great way to block ads and trackers on your network without having to install ad blockers on every device.