UNB/ CS/ David Bremner/ teaching/ cs2613/ books/ mdn/ Reference/ Global Objects/ Array/ Array.prototype.reduce()

The reduce() method of Array instances executes a user-supplied "reducer" callback function on each element of the array, in order, passing in the return value from the calculation on the preceding element. The final result of running the reducer across all elements of the array is a single value.

The first time that the callback is run there is no "return value of the previous calculation". If supplied, an initial value may be used in its place. Otherwise the array element at index 0 is used as the initial value and iteration starts from the next element (index 1 instead of index 0).

Perhaps the easiest-to-understand case for reduce() is to return the sum of all the elements in an array:

The reducer walks through the array element-by-element, at each step adding the current array value to the result from the previous step (this result is the running sum of all the previous steps) — until there are no more elements to add.

Syntax

reduce(callbackFn)
reduce(callbackFn, initialValue)

Parameters

Return value

The value that results from running the "reducer" callback function to completion over the entire array.

Exceptions

Description

The reduce() method is an iterative method. It runs a "reducer" callback function over all elements in the array, in ascending-index order, and accumulates them into a single value. Every time, the return value of callbackFn is passed into callbackFn again on next invocation as accumulator. The final value of accumulator (which is the value returned from callbackFn on the final iteration of the array) becomes the return value of reduce().

callbackFn is invoked only for array indexes which have assigned values. It is not invoked for empty slots in sparse arrays.

Unlike other iterative methods, reduce() does not accept a thisArg argument. callbackFn is always called with undefined as this, which gets substituted with globalThis if callbackFn is non-strict.

reduce() is a central concept in functional programming, where it's not possible to mutate any value, so in order to accumulate all values in an array, one must return a new accumulator value on every iteration. This convention propagates to JavaScript's reduce(): you should use spreading or other copying methods where possible to create new arrays and objects as the accumulator, rather than mutating the existing one. If you decided to mutate the accumulator instead of copying it, remember to still return the modified object in the callback, or the next iteration will receive undefined.

reduce() does not mutate the array on which it is called, but the function provided as callbackFn can. Note, however, that the length of the array is saved before the first invocation of callbackFn. Therefore:

Warning: Concurrent modifications of the kind described above frequently lead to hard-to-understand code and are generally to be avoided (except in special cases).

The reduce() method is generic. It only expects the this value to have a length property and integer-keyed properties.

Edge cases

If the array only has one element (regardless of position) and no initialValue is provided, or if initialValue is provided but the array is empty, the solo value will be returned without calling callbackFn.

If initialValue is provided and the array is not empty, then the reduce method will always invoke the callback function starting at index 0.

If initialValue is not provided then the reduce method will act differently for arrays with length larger than 1, equal to 1 and 0, as shown in the following example:

const getMax = (a, b) => Math.max(a, b);

// callback is invoked for each element in the array starting at index 0
[1, 100].reduce(getMax, 50); // 100
[50].reduce(getMax, 10); // 50

// callback is invoked once for element at index 1
[1, 100].reduce(getMax); // 100

// callback is not invoked
[50].reduce(getMax); // 50
[].reduce(getMax, 1); // 1

[].reduce(getMax); // TypeError

Examples

How reduce() works without an initial value

The code below shows what happens if we call reduce() with an array and no initial value.

const array = [15, 16, 17, 18, 19];

function reducer(accumulator, currentValue, index) {
  const returns = accumulator + currentValue;
  console.log(
    `accumulator: ${accumulator}, currentValue: ${currentValue}, index: ${index}, returns: ${returns}`,
  );
  return returns;
}

array.reduce(reducer);

The callback would be invoked four times, with the arguments and return values in each call being as follows:

accumulator currentValue index Return value
First call 15 16 1 31
Second call 31 17 2 48
Third call 48 18 3 66
Fourth call 66 19 4 85

The array parameter never changes through the process — it's always [15, 16, 17, 18, 19]. The value returned by reduce() would be that of the last callback invocation (85).

How reduce() works with an initial value

Here we reduce the same array using the same algorithm, but with an initialValue of 10 passed as the second argument to reduce():

[15, 16, 17, 18, 19].reduce(
  (accumulator, currentValue) => accumulator + currentValue,
  10,
);

The callback would be invoked five times, with the arguments and return values in each call being as follows:

accumulator currentValue index Return value
First call 10 15 0 25
Second call 25 16 1 41
Third call 41 17 2 58
Fourth call 58 18 3 76
Fifth call 76 19 4 95

The value returned by reduce() in this case would be 95.

Sum of values in an object array

To sum up the values contained in an array of objects, you must supply an initialValue, so that each item passes through your function.

const objects = [{ x: 1 }, { x: 2 }, { x: 3 }];
const sum = objects.reduce(
  (accumulator, currentValue) => accumulator + currentValue.x,
  0,
);

console.log(sum); // 6

Function sequential piping

The pipe function takes a sequence of functions and returns a new function. When the new function is called with an argument, the sequence of functions are called in order, which each one receiving the return value of the previous function.

const pipe =
  (...functions) =>
  (initialValue) =>
    functions.reduce((acc, fn) => fn(acc), initialValue);

// Building blocks to use for composition
const double = (x) => 2 * x;
const triple = (x) => 3 * x;
const quadruple = (x) => 4 * x;

// Composed functions for multiplication of specific values
const multiply6 = pipe(double, triple);
const multiply9 = pipe(triple, triple);
const multiply16 = pipe(quadruple, quadruple);
const multiply24 = pipe(double, triple, quadruple);

// Usage
multiply6(6); // 36
multiply9(9); // 81
multiply16(16); // 256
multiply24(10); // 240

Running promises in sequence

Promise sequencing is essentially function piping demonstrated in the previous section, except done asynchronously.

// Compare this with pipe: fn(acc) is changed to acc.then(fn),
// and initialValue is ensured to be a promise
const asyncPipe =
  (...functions) =>
  (initialValue) =>
    functions.reduce((acc, fn) => acc.then(fn), Promise.resolve(initialValue));

// Building blocks to use for composition
const p1 = async (a) => a * 5;
const p2 = async (a) => a * 2;
// The composed functions can also return non-promises, because the values are
// all eventually wrapped in promises
const f3 = (a) => a * 3;
const p4 = async (a) => a * 4;

asyncPipe(p1, p2, f3, p4)(10).then(console.log); // 1200

asyncPipe can also be implemented using async/await, which better demonstrates its similarity with pipe:

const asyncPipe =
  (...functions) =>
  (initialValue) =>
    functions.reduce(async (acc, fn) => fn(await acc), initialValue);

Using reduce() with sparse arrays

reduce() skips missing elements in sparse arrays, but it does not skip undefined values.

console.log([1, 2, , 4].reduce((a, b) => a + b)); // 7
console.log([1, 2, undefined, 4].reduce((a, b) => a + b)); // NaN

Calling reduce() on non-array objects

The reduce() method reads the length property of this and then accesses each property whose key is a nonnegative integer less than length.

const arrayLike = {
  length: 3,
  0: 2,
  1: 3,
  2: 4,
  3: 99, // ignored by reduce() since length is 3
};
console.log(Array.prototype.reduce.call(arrayLike, (x, y) => x + y));
// 9

When to not use reduce()

Multipurpose higher-order functions like reduce() can be powerful but sometimes difficult to understand, especially for less-experienced JavaScript developers. If code becomes clearer when using other array methods, developers must weigh the readability tradeoff against the other benefits of using reduce().

Note that reduce() is always equivalent to a for...of loop, except that instead of mutating a variable in the upper scope, we now return the new value for each iteration:

const val = array.reduce((acc, cur) => update(acc, cur), initialValue);

// Is equivalent to:
let val = initialValue;
for (const cur of array) {
  val = update(val, cur);
}

As previously stated, the reason why people may want to use reduce() is to mimic functional programming practices of immutable data. Therefore, developers who uphold the immutability of the accumulator often copy the entire accumulator for each iteration, like this:

const names = ["Alice", "Bob", "Tiff", "Bruce", "Alice"];
const countedNames = names.reduce((allNames, name) => {
  const currCount = Object.hasOwn(allNames, name) ? allNames[name] : 0;
  return {
    ...allNames,
    [name]: currCount + 1,
  };
}, {});

This code is ill-performing, because each iteration has to copy the entire allNames object, which could be big, depending how many unique names there are. This code has worst-case O(N^2) performance, where N is the length of names.

A better alternative is to mutate the allNames object on each iteration. However, if allNames gets mutated anyway, you may want to convert the reduce() to a simple for loop instead, which is much clearer:

const names = ["Alice", "Bob", "Tiff", "Bruce", "Alice"];
const countedNames = names.reduce((allNames, name) => {
  const currCount = allNames[name] ?? 0;
  allNames[name] = currCount + 1;
  // return allNames, otherwise the next iteration receives undefined
  return allNames;
}, Object.create(null));
const names = ["Alice", "Bob", "Tiff", "Bruce", "Alice"];
const countedNames = Object.create(null);
for (const name of names) {
  const currCount = countedNames[name] ?? 0;
  countedNames[name] = currCount + 1;
}

Therefore, if your accumulator is an array or an object and you are copying the array or object on each iteration, you may accidentally introduce quadratic complexity into your code, causing performance to quickly degrade on large data.

Some of the acceptable use cases of reduce() are given above (most notably, summing an array, promise sequencing, and function piping). There are other cases where better alternatives than reduce() exist.

  const flattened = array.reduce((acc, cur) => acc.concat(cur), []);
  const flattened = array.flat();
  const groups = array.reduce((acc, obj) => {
    const key = obj.name;
    const curGroup = acc[key] ?? [];
    return { ...acc, [key]: [...curGroup, obj] };
  }, {});
  const groups = Object.groupBy(array, (obj) => obj.name);
  const friends = [
    { name: "Anna", books: ["Bible", "Harry Potter"] },
    { name: "Bob", books: ["War and peace", "Romeo and Juliet"] },
    { name: "Alice", books: ["The Lord of the Rings", "The Shining"] },
  ];
  const allBooks = friends.reduce((acc, cur) => [...acc, ...cur.books], []);
  const allBooks = friends.flatMap((person) => person.books);
  const uniqArray = array.reduce(
    (acc, cur) => (acc.includes(cur) ? acc : [...acc, cur]),
    [],
  );
  const uniqArray = Array.from(new Set(array));
  // Takes an array of numbers and splits perfect squares into its square roots
  const roots = array.reduce((acc, cur) => {
    if (cur < 0) return acc;
    const root = Math.sqrt(cur);
    if (Number.isInteger(root)) return [...acc, root, root];
    return [...acc, cur];
  }, []);
  const roots = array.flatMap((val) => {
    if (val < 0) return [];
    const root = Math.sqrt(val);
    if (Number.isInteger(root)) return [root, root];
    return [val];
  });

If you are only eliminating elements from an array, you also can use filter().

  const allEven = array.reduce((acc, cur) => acc && cur % 2 === 0, true);
  const allEven = array.every((val) => val % 2 === 0);

In cases where reduce() is the best choice, documentation and semantic variable naming can help mitigate readability drawbacks.

Specifications

Browser compatibility

See also