conorhastings / match-sorter

Simple, expected, and deterministic best-match sorting of an array in JavaScript

match-sorter

Simple, expected, and deterministic best-match sorting of an array in JavaScript. Demo

The problem

1. You have a list of dozens, hundreds, or thousands of items
2. You want to filter and sort those items intelligently (maybe you have a filter input for the user)
3. You want simple, expected, and deterministic sorting of the items (no fancy math algorithm that fancily changes the sorting as they type)

This solution

This follows a simple and sensible (user friendly) algorithm that makes it easy for you to filter and sort a list of items based on given input. Items are ranked based on sensible criteria that result in a better user experience.

To explain the ranking system, I'll use countries as an example:

1. EQUALS: Equality trumps all. These will be first. (ex. France would match France)
2. STARTS WITH: If the item starts with the given value (ex. Sou would match South Korea or South Africa)
3. WORD STARTS WITH: If the item has multiple words, then if one of those words starts with the given value (ex. Repub would match Dominican Republic)
4. CONTAINS: If the item contains the given value (ex. ham would match Bahamas)
5. ACRONYM: If the item's acronym is the given value (ex. us would match United States)
6. SIMPLE MATCH: If the item has letters in the same order as the letters of the given value (ex. iw would match Zimbabwe, but not Kuwait because it must be in the same order).

This ranking seems to make sense in people's minds. At least it does in mine. Feedback welcome!

Installation

This module is distributed via npm which is bundled with node and should be installed as one of your project's dependencies:

npm install --save match-sorter

Usage

const matchSorter = require('match-sorter')
// ES6 imports work too
// Also available in global environment via `matchSorter` global
const list = ['hi', 'hey', 'hello', 'sup', 'yo']
matchSorter(list, 'h') // ['hi', 'hey', 'hello']
matchSorter(list, 'y') // ['yo', 'hey']
matchSorter(list, 'z') // []

// You can also pass an options object:
// **keys** (defaults to undefined and just uses the value itself as above)
const objList = [
  {name: 'Janice', color: 'Green'},
  {name: 'Fred', color: 'Orange'},
  {name: 'George', color: 'Blue'},
  {name: 'Jen', color: 'Red'},
]
matchSorter(objList, 'g', {keys: ['name', 'color']}) // [{name: 'George', color: 'Blue'}, {name: 'Janice', color: 'Green'}]
matchSorter(objList, 're', {keys: ['color', 'name']}) // [{name: 'Jen', color: 'Red'}, {name: 'Janice', color: 'Green'}, {name: 'Fred', color: 'Orange'}]

// **threshold** (defaults to MATCH)
const fruit = ['orange', 'apple', 'grape', 'banana']
matchSorter(fruit, 'ap', {threshold: matchSorter.rankings.NO_MATCH}) // ['apple', 'grape', 'orange', 'banana'] (returns all items, just sorted by best match)
const things = ['google', 'airbnb', 'apple', 'apply', 'app'],
matchSorter(things, 'app', {threshold: matchSorter.rankings.EQUAL}) // ['app'] (only items that are equal)
const otherThings = ['fiji apple', 'google', 'app', 'crabapple', 'apple', 'apply']
matchSorter(otherThings, 'app', {threshold: matchSorter.rankings.WORD_STARTS_WITH}) // ['app', 'apple', 'apply', 'fiji apple'] (everything that matches with "word starts with" or better)

/*
 * Available thresholds (from top to bottom) are:
 * - EQUAL
 * - STARTS_WITH
 * - WORD_STARTS_WITH
 * - CONTAINS
 * - ACRONYM
 * - MATCHES
 * - NO_MATCH
 */

In the examples above, we're using CommonJS. If you're using ES6 modules, then you can do:

import matchSorter, {rankings} from 'match-sorter'

Inspiration

Actually, most of this code was extracted from the very first library I ever wrote: genie!

Other Solutions

You might try Fuse.js. It uses advanced math fanciness to get the closest match. Unfortunately what's "closest" doesn't always really make sense. So I extracted this from genie.

Contributors

Thanks goes to these people (emoji key):

Kent C. Dodds 💻 📖 🚇 ⚠️

This project follows the all-contributors specification. Contributions of any kind welcome!

LICENSE

MIT