Skip to content

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

License

Notifications You must be signed in to change notification settings

kentcdodds/match-sorter

Repository files navigation

match-sorter

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


Demo

Build Status Code Coverage version downloads MIT License All Contributors PRs Welcome Code of Conduct Examples

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. CASE SENSITIVE EQUALS: Case-sensitive equality trumps all. These will be first. (ex. France would match France, but not france)
  2. EQUALS: Case-insensitive equality (ex. France would match france)
  3. STARTS WITH: If the item starts with the given value (ex. Sou would match South Korea or South Africa)
  4. 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)
  5. CONTAINS: If the item contains the given value (ex. ham would match Bahamas)
  6. ACRONYM: If the item's acronym is the given value (ex. us would match United States)
  7. 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). Furthermore, if the item is a closer match, it will rank higher (ex. ua matches Uruguay more closely than United States of America, therefore Uruguay will be ordered before United States of America)

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 match-sorter

Usage

import {matchSorter} from 'match-sorter'
// or const {matchSorter} = require('match-sorter')
// or window.matchSorter.matchSorter
const list = ['hi', 'hey', 'hello', 'sup', 'yo']
matchSorter(list, 'h') // ['hello', 'hey', 'hi']
matchSorter(list, 'y') // ['yo', 'hey']
matchSorter(list, 'z') // []

Advanced options

keys: [string]

Default: undefined

By default it just uses the value itself as above. Passing an array tells match-sorter which keys to use for the ranking.

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'}, {name: 'Fred', color: 'Orange'}]

matchSorter(objList, 're', {keys: ['color', 'name']})
// [{name: 'Jen', color: 'Red'}, {name: 'Janice', color: 'Green'}, {name: 'Fred', color: 'Orange'}, {name: 'George', color: 'Blue'}]

Array of values: When the specified key matches an array of values, the best match from the values of in the array is going to be used for the ranking.

const iceCreamYum = [
  {favoriteIceCream: ['mint', 'chocolate']},
  {favoriteIceCream: ['candy cane', 'brownie']},
  {favoriteIceCream: ['birthday cake', 'rocky road', 'strawberry']},
]
matchSorter(iceCreamYum, 'cc', {keys: ['favoriteIceCream']})
// [{favoriteIceCream: ['candy cane', 'brownie']}, {favoriteIceCream: ['mint', 'chocolate']}]

Nested Keys: You can specify nested keys using dot-notation.

const nestedObjList = [
  {name: {first: 'Janice'}},
  {name: {first: 'Fred'}},
  {name: {first: 'George'}},
  {name: {first: 'Jen'}},
]
matchSorter(nestedObjList, 'j', {keys: ['name.first']})
// [{name: {first: 'Janice'}}, {name: {first: 'Jen'}}]

const nestedObjList = [
  {name: [{first: 'Janice'}]},
  {name: [{first: 'Fred'}]},
  {name: [{first: 'George'}]},
  {name: [{first: 'Jen'}]},
]
matchSorter(nestedObjList, 'j', {keys: ['name.0.first']})
// [{name: {first: 'Janice'}}, {name: {first: 'Jen'}}]

// matchSorter(nestedObjList, 'j', {keys: ['name[0].first']}) does not work

This even works with arrays of multiple nested objects: just specify the key using dot-notation with the * wildcard instead of a numeric index.

const nestedObjList = [
  {aliases: [{name: {first: 'Janice'}},{name: {first: 'Jen'}}]},
  {aliases: [{name: {first: 'Fred'}},{name: {first: 'Frederic'}}]},
  {aliases: [{name: {first: 'George'}},{name: {first: 'Georgie'}}]},
]
matchSorter(nestedObjList, 'jen', {keys: ['aliases.*.name.first']})
// [{aliases: [{name: {first: 'Janice'}},{name: {first: 'Jen'}}]}]
matchSorter(nestedObjList, 'jen', {keys: ['aliases.0.name.first']})
// []

Property Callbacks: Alternatively, you may also pass in a callback function that resolves the value of the key(s) you wish to match on. This is especially useful when interfacing with libraries such as Immutable.js

const list = [{name: 'Janice'}, {name: 'Fred'}, {name: 'George'}, {name: 'Jen'}]
matchSorter(list, 'j', {keys: [item => item.name]})
// [{name: 'Janice'}, {name: 'Jen'}]

For more complex structures, expanding on the nestedObjList example above, you can use map:

const nestedObjList = [
  {
    name: [
      {first: 'Janice', last: 'Smith'},
      {first: 'Jon', last: 'Doe'},
    ],
  },
  {
    name: [
      {first: 'Fred', last: 'Astaire'},
      {first: 'Jenny', last: 'Doe'},
      {first: 'Wilma', last: 'Flintstone'},
    ],
  },
]
matchSorter(nestedObjList, 'doe', {
  keys: [
    item => item.name.map(i => i.first),
    item => item.name.map(i => i.last),
  ],
})
// [name: [{ first: 'Janice', last: 'Smith' },{ first: 'Jon', last: 'Doe' }], name: [{ first: 'Fred', last: 'Astaire' },{ first: 'Jenny', last: 'Doe' },{ first: 'Wilma', last: 'Flintstone' }]]

Threshold: You may specify an individual threshold for specific keys. A key will only match if it meets the specified threshold. For more information regarding thresholds see below

const list = [
  {name: 'Fred', color: 'Orange'},
  {name: 'Jen', color: 'Red'},
]
matchSorter(list, 'ed', {
  keys: [{threshold: matchSorter.rankings.STARTS_WITH, key: 'name'}, 'color'],
})
//[{name: 'Jen', color: 'Red'}]

Min and Max Ranking: You may restrict specific keys to a minimum or maximum ranking by passing in an object. A key with a minimum rank will only get promoted if there is at least a simple match.

const tea = [
  {tea: 'Earl Grey', alias: 'A'},
  {tea: 'Assam', alias: 'B'},
  {tea: 'Black', alias: 'C'},
]
matchSorter(tea, 'A', {
  keys: ['tea', {maxRanking: matchSorter.rankings.STARTS_WITH, key: 'alias'}],
})
// without maxRanking, Earl Grey would come first because the alias "A" would be CASE_SENSITIVE_EQUAL
// `tea` key comes before `alias` key, so Assam comes first even though both match as STARTS_WITH
// [{tea: 'Assam', alias: 'B'}, {tea: 'Earl Grey', alias: 'A'},{tea: 'Black', alias: 'C'}]
const tea = [
  {tea: 'Milk', alias: 'moo'},
  {tea: 'Oolong', alias: 'B'},
  {tea: 'Green', alias: 'C'},
]
matchSorter(tea, 'oo', {
  keys: ['tea', {minRanking: matchSorter.rankings.EQUAL, key: 'alias'}],
})
// minRanking bumps Milk up to EQUAL from CONTAINS (alias)
// Oolong matches as STARTS_WITH
// Green is missing due to no match
// [{tea: 'Milk', alias: 'moo'}, {tea: 'Oolong', alias: 'B'}]

threshold: number

Default: MATCHES

Thresholds can be used to specify the criteria used to rank the results. Available thresholds (from top to bottom) are:

  • CASE_SENSITIVE_EQUAL
  • EQUAL
  • STARTS_WITH
  • WORD_STARTS_WITH
  • CONTAINS
  • ACRONYM
  • MATCHES (default value)
  • NO_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)

keepDiacritics: boolean

Default: false

By default, match-sorter will strip diacritics before doing any comparisons. This is the default because it makes the most sense from a UX perspective.

You can disable this behavior by specifying keepDiacritics: true

const thingsWithDiacritics = [
  'jalapeΓ±o',
  'Γ  la carte',
  'cafΓ©',
  'papier-mΓ’chΓ©',
  'Γ  la mode',
]
matchSorter(thingsWithDiacritics, 'aa')
// ['jalapeΓ±o', 'Γ  la carte', 'Γ  la mode', 'papier-mΓ’chΓ©']

matchSorter(thingsWithDiacritics, 'aa', {keepDiacritics: true})
// ['jalapeΓ±o', 'Γ  la carte']

matchSorter(thingsWithDiacritics, 'Γ ', {keepDiacritics: true})
// ['Γ  la carte', 'Γ  la mode']

baseSort: function(itemA, itemB): -1 | 0 | 1

Default: (a, b) => String(a.rankedValue).localeCompare(b.rankedValue)

By default, match-sorter uses the String.localeCompare function to tie-break items that have the same ranking. This results in a stable, alphabetic sort.

const list = ['C apple', 'B apple', 'A apple']
matchSorter(list, 'apple')
// ['A apple', 'B apple', 'C apple']

You can customize this behavior by specifying a custom baseSort function:

const list = ['C apple', 'B apple', 'A apple']
// This baseSort function will use the original index of items as the tie breaker
matchSorter(list, 'apple', {baseSort: (a, b) => (a.index < b.index ? -1 : 1)})
// ['C apple', 'B apple', 'A apple']

sorter: function(rankedItems): rankedItems

Default: matchedItems => matchedItems.sort((a, b) => sortRankedValues(a, b, baseSort))

By default, match-sorter uses an internal sortRankedValues function to sort items after matching them.

You can customize the core sorting behavior by specifying a custom sorter function:

Disable sorting entirely:

const list = ['appl', 'C apple', 'B apple', 'A apple', 'app', 'applebutter']
matchSorter(list, 'apple', {sorter: rankedItems => rankedItems})
// ['C apple', 'B apple', 'A apple', 'applebutter']

Return the unsorted rankedItems, but in reverse order:

const list = ['appl', 'C apple', 'B apple', 'A apple', 'app', 'applebutter']
matchSorter(list, 'apple', {sorter: rankedItems => [...rankedItems].reverse()})
// ['applebutter', 'A apple', 'B apple', 'C apple']

Recipes

Match PascalCase, camelCase, snake_case, or kebab-case as words

By default, match-sorter assumes spaces to be the word separator. However, if your data has a different word separator, you can use a property callback to replace your separator with spaces. For example, for snake_case:

const list = [
  {name: 'Janice_Kurtis'},
  {name: 'Fred_Mertz'},
  {name: 'George_Foreman'},
  {name: 'Jen_Smith'},
]
matchSorter(list, 'js', {keys: [item => item.name.replace(/_/g, ' ')]})
// [{name: 'Jen_Smith'}, {name: 'Janice_Kurtis'}]

Match many words across multiple fields (table filtering)

By default, match-sorter will return matches from objects where one of the properties matches the entire search term. For multi-column data sets it can be beneficial to split words in search string and match each word separately. This can be done by chaining match-sorter calls.

The benefit of this is that a filter string of "two words" will match both "two" and "words", but will return rows where the two words are found in different columns as well as when both words match in the same column. For single-column matches it will also return matches out of order (column = "wordstwo" will match just as well as column="twowords", the latter getting a higher score).

function fuzzySearchMultipleWords(
  rows, // array of data [{a: "a", b: "b"}, {a: "c", b: "d"}]
  keys, // keys to search ["a", "b"]
  filterValue: string, // potentially multi-word search string "two words"
) {
  if (!filterValue || !filterValue.length) {
    return rows
  }

  const terms = filterValue.split(' ')
  if (!terms) {
    return rows
  }

  // reduceRight will mean sorting is done by score for the _first_ entered word.
  return terms.reduceRight(
    (results, term) => matchSorter(results, term, {keys}),
    rows,
  )
}

Multi-column code sandbox

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.

Issues

Looking to contribute? Look for the Good First Issue label.

πŸ› Bugs

Please file an issue for bugs, missing documentation, or unexpected behavior.

See Bugs

πŸ’‘ Feature Requests

Please file an issue to suggest new features. Vote on feature requests by adding a πŸ‘. This helps maintainers prioritize what to work on.

See Feature Requests

Contributors ✨

Thanks goes to these people (emoji key):


Kent C. Dodds

πŸ’» πŸ“– πŸš‡ ⚠️ πŸ‘€

Conor Hastings

πŸ’» πŸ“– ⚠️ πŸ‘€

Rogelio Guzman

πŸ“–

ClaudΓ©ric Demers

πŸ’» πŸ“– ⚠️

Kevin Davis

πŸ’» ⚠️

Denver Chen

πŸ’» πŸ“– ⚠️

Christian Ruigrok

πŸ› πŸ’» πŸ“–

Hozefa

πŸ› πŸ’» ⚠️ πŸ€”

pushpinder107

πŸ’»

Mordy Tikotzky

πŸ’» πŸ“– ⚠️

Steven Brannum

πŸ’» ⚠️

Christer van der Meeren

πŸ›

Samuel Petrosyan

πŸ’» πŸ›

Brandon Kalinowski

πŸ›

Eric Berry

πŸ”

Skubie Doo

πŸ“–

MichaΓ«l De Boey

πŸ’» πŸ‘€

Tanner Linsley

πŸ’» ⚠️

Victor

πŸ“–

Rebecca Stevens

πŸ› πŸ“–

Marco Moretti

πŸ“–

Ricardo Busquet

πŸ€” πŸ‘€ πŸ’»

Weyert de Boer

πŸ€” πŸ‘€

Philipp Garbowsky

πŸ’»

Mart

πŸ’» ⚠️ πŸ“–

Aleksey Levenstein

πŸ’»

Take Weiland

πŸ’»

Amit Abershitz

πŸ“–

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

LICENSE

MIT