fast-sort

Fast easy to use and flexible sorting with TypeScript support. For speed comparison of fast-sort vs other popular sort libraries check benchmark section. For list of all available features check highlights section.

Quick examples

  import { sort } from 'fast-sort';

  // Sort flat arrays
  const ascSorted = sort([1,4,2]).asc(); // => [1, 2, 4]
  const descSorted = sort([1, 4, 2]).desc(); // => [4, 2, 1]

  // Sort users (array of objects) by firstName in descending order
  const sorted = sort(users).desc(u => u.firstName);

  // Sort users in ascending order by firstName and lastName
  const sorted = sort(users).asc([
    u => u.firstName,
    u => u.lastName
  ]);

  // Sort users ascending by firstName and descending by city
  const sorted = sort(users).by([
    { asc: u => u.firstName },
    { desc: u => u.address.city }
  ]);

  // Sort based on computed property
  const sorted = sort(repositories).desc(r => r.openIssues + r.closedIssues);

  // Sort using string for object key
  // Only available for root object properties
  const sorted = sort(users).asc('firstName');

Fore more examples check unit tests.

Highlights

Sort flat arrays
Sort array of objects by one or more properties
Sort in multiple directions
Natural sort support
Support for custom sort instances
Easy to read syntax
Faster than other popular sort alternatives
Undefined and null values are always sorted to bottom (with default comparer)
TypeScript support
Packed with features in small footprint with 0 dependencies (~ 850 bytes gzip)
Compatible with any JS environment as Node, Web, etc..

Migrating from older versions

Documentation for v2 and older versions is available here.

For migrating to v3 you can reference CHANGELOG for what has been changed.

In place sorting

By default sort does not mutate provided array it creates new "sorted" instance of array. inPlaceSort on other hand mutates provided array by sorting it without creating new array instance. Benefits of inPlaceSort is that it's slightly faster and more generous on memory as it's not creating new array instance every time sorting is done. Other than that there is no difference between using one or another.

const { sort, inPlaceSort } = require('fast-sort');

const array = [3, 1, 5];
const sorted = sort(array).asc();

// sorted => [1, 3, 5]
// array => [3, 1, 5]

inPlaceSort(array).asc();

// array => [1, 3, 5]

Natural sorting / Language sensitive sorting

By default fast-sort is not doing language sensitive sorting of strings. e.g 'image-11.jpg' will be sorted before 'image-2.jpg' (in ascending sorting). We can provide custom Intl.Collator comparer to fast-sort for language sensitive sorting of strings. Keep in mind that natural sort is slower then default sorting so recommendation is to use it only when needed.

  import { sort, createNewSortInstance } from 'fast-sort';

  const testArr = ['image-2.jpg', 'image-11.jpg', 'image-3.jpg'];

  // By default fast-sort is not doing natural sort
  sort(testArr).desc(); // => ['image-3.jpg', 'image-2.jpg', 'image-11.jpg']

  // We can use `by` sort to override default comparer
  // with the one that is doing language sensitive comparison
  sort(testArr).by({
    desc: true,
    comparer: new Intl.Collator(undefined, { numeric: true, sensitivity: 'base' }).compare,
  }); // => ['image-11.jpg', 'image-3.jpg', 'image-2.jpg']

  // Or we can create new sort instance with language sensitive comparer.
  // Recommended if used in multiple places
  const naturalSort = createNewSortInstance({
    comparer: new Intl.Collator(undefined, { numeric: true, sensitivity: 'base' }).compare,
  });

  naturalSort(testArr).asc(); // => ['image-2.jpg', 'image-3.jpg', 'image-11.jpg']
  naturalSort(testArr).desc(); // => ['image-11.jpg', 'image-3.jpg', 'image-2.jpg']

NOTE: It's known that Intl.Collator might not sort null values correctly so make sure to cast them to undefine as described in the following issue snovakovic#54 (comment)

Custom sorting

Fast sort can be tailored to fit any sorting need or use case by:

creating custom sorting instances
overriding default comparer in by sorter
custom handling in provided callback function
combination of any from above

For example we will sort tags by "custom" tag importance (e.g vip tag is of greater importance then captain tag).

  import { sort, createNewSortInstance } from 'fast-sort';

  const tags = ['influencer', 'unknown', 'vip', 'captain'];
  const tagsImportance = { // Domain specific tag importance
    vip: 3,
    influencer: 2,
    captain: 1,
  };

  // We can use power of computed prop to sort tags by domain specific importance
  const descTags = sort(tags).desc(tag => tagImportance[tag] || 0);
  // => ['vip', 'influencer', 'captain', 'unknown'];

  // Or we can create specialized tagSorter so we can reuse it in multiple places
  const tagSorter = createNewSortInstance({
    comparer: (a, b) => (tagImportance[a] || 0) - (tagImportance[b] || 0),
    inPlaceSorting: true, // default[false] => Check "In Place Sort" section for more info.
  });

  tagSorter(tags).asc(); // => ['unknown', 'captain', 'influencer', 'vip'];
  tagSorter(tags).desc(); // => ['vip', 'influencer', 'captain', 'unknown'];

  // Default sorter will sort tags by comparing string values not by their domain specific value
  const defaultSort = sort(tags).asc(); // => ['captain', 'influencer', 'unknown' 'vip']

More examples

  // Sorting values that are not sortable will return same value back
  sort(null).asc(); // => null
  sort(33).desc(); // => 33

  // By default fast-sort sorts null and undefined values to the
  // bottom no matter if sorting is in asc or decs order.
  // If this is not intended behaviour you can check "Should create sort instance that sorts nil value to the top in desc order" test on how to override
  const addresses = [{ city: 'Split' }, { city: undefined }, { city: 'Zagreb'}];
  sort(addresses).asc(a => a.city); // => Split, Zagreb, undefined
  sort(addresses).desc(a => a.city); // => Zagreb, Split, undefined

Benchmark

Five different benchmarks have been created to get better insight of how fast-sort perform under different scenarios. Each benchmark is run with different array sizes raging from small 100 items to large 100 000 items.

Every run of benchmark outputs different results but the results are constantly showing better scores compared to similar popular sorting libraries.

Benchmark scores

Benchmark has been run on:

16 GB Ram
Intel® Core™ i5-4570 CPU @ 3.20GHz × 4
Ubuntu 16.04
Node 8.9.1

Independent benchmark results from MacBook Air can be found in following PR: snovakovic#48

Running benchmark

To run benchmark on your PC follow steps from below

git clone https://github.com/snovakovic/fast-sort.git
cd fast-sort/benchmark
npm install
npm start

In case you notice any irregularities in benchmark or you want to add sort library to benchmark score please open issue here

gsphamnguyen/fast-sort