Post cover

How Three Dots Changed JavaScript

I don't like arguments keyword when I access the arguments of a function call. Its hardcoded name makes difficult to access arguments of an outer function in an inner one (which defines its own arguments).

Even worse arguments is an array-like object. You cannot use array methods like .map() or .forEach() directly on it.

To access arguments from the enclosing function, you have to use workarounds by storing it into a separated variable. And to walk through this array-like object, you have to use duck typing and make indirect invocations. See the following example:


function outerFunction() {
// store arguments into a separated variable
const argsOuter = arguments;
function innerFunction() {
// args is an array-like object
const even = Array.prototype.map.call(argsOuter, function(item) {
// do something with argsOuter
});
}
}

Another situation is the function invocation that accepts a dynamic number of arguments. Filling the arguments from an array is unpleasant.

For instance .push(item1, ..., itemN) inserts elements into an array one by one: you have to enumerate each element as an argument. This is not always convenient: often an entire array of elements needs to be pushed into an existing array, without creating a new instance.

In ES5 it's solved with .apply(): an unfriendly and verbose approach. Let's take a look:


const fruits = ['banana'];
const moreFruits = ['apple', 'orange'];
Array.prototype.push.apply(fruits, moreFruits);
console.log(fruits); // => ['banana', 'apple', 'orange']

Fortunately, the JavaScript world is changing. The three dots operator ... fixes many of these situations. The operator is introduced by ECMAScript 6 and in my opinion, is a noticeable improvement.

This article walks through ... operator use cases and shows how to solve similar problems.

1. Three dots

The JavaScript rest operator collects the arguments list passed to a function on invocation or collects the pieces after destructuring an array. A case when the operator gathers the rest remained after the operation.


function countArguments(...args) {
return args.length;
}
// get the number of arguments
countArguments('welcome', 'to', 'Earth'); // => 3
// destructure an array
let otherSeasons, autumn;
[autumn, ...otherSeasons] = ['autumn', 'winter'];
otherSeasons // => ['winter']

The JavaScript spread operator constructs arrays or sets the function arguments on invocation from an array. A case when the operator spreads the array (or iterable object) elements.


let cold = ['autumn', 'winter'];
let warm = ['spring', 'summer'];
// construct an array
[...cold, ...warm] // => ['autumn', 'winter', 'spring', 'summer']
// function arguments from an array
cold.push(...warm);
cold // => ['autumn', 'winter', 'spring', 'summer']

2. Improved parameters access

2.1 Rest parameter

As presented in the introduction, dealing with arguments object in a function body becomes troublesome in complex scenarios.

For example a JavaScript inner function filterNumbers() wants to access arguments from its outer function sumOnlyNumbers():


function sumOnlyNumbers() {
const args = arguments;
const numbers = filterNumbers();
return numbers.reduce((sum, element) => sum + element);
function filterNumbers() {
return Array.prototype.filter.call(args,
element => typeof element === 'number'
);
}
}
sumOnlyNumbers(1, 'Hello', 5, false); // => 6

To access arguments of sumOnlyNumbers() inside filterNumbers(), you have to create a temporary variable args. It happens because filterNumbers() defines its own arguments object that overwrites the external one.

The approach works, but it's too verbose. const args = arguments can be omitted and Array.prototype.filter.call(args) can be transformed to args.filter() using a rest parameter. Let's optimize this part.

The rest operator solves this elegantly. It allows to define a rest parameter ...args in a function declaration:


function sumOnlyNumbers(...args) {
const numbers = filterNumbers();
return numbers.reduce((sum, element) => sum + element);
function filterNumbers() {
return args.filter(element => typeof element === 'number');
}
}
sumOnlyNumbers(1, 'Hello', 5, false); // => 6

The function declaration function sumOnlyNumbers(...args) indicates that args receives the invocation arguments in an array. Because the names conflict is solved, args can be used inside filterNumbers().

Also forget about array-like objects: args is an array - which is a nice bonus. As result, filterNumbers() can get rid of Array.prototype.filter.call() and make a filter method call directly args.filter().

Notice that the rest parameter must be the last one in the function parameters list.

2.2 Selective rest parameter

When not all values should be included in the rest parameter, you can keep those as comma separated parameters at the beginning. Explicitly defined parameters are not included in the rest parameter.

Let's see an example:


function filter(type, ...items) {
return items.filter(item => typeof item === type);
}
filter('boolean', true, 0, false); // => [true, false]
filter('number', false, 4, 'Welcome', 7); // => [4, 7]

arguments object doesn't have this selective property and always includes all the values.

2.3 Arrow function case

An arrow function does not define arguments object in its body but accesses the one from the enclosing scope.

If you want to get all the arguments in an arrow function, use a rest parameter.

Let's try this in an example:


(function() {
let outerArguments = arguments;
const concat = (...items) => {
console.log(arguments === outerArguments); // => true
return items.reduce((result, item) => result + item, '');
};
concat(1, 5, 'nine'); // => '15nine'
})();

items rest parameter contains all function call arguments in an array. Also arguments object is taken from the enclosing scope and equals to outerArguments variable, so it has no valuable meaning.

3. Improved function call

In the article introduction, the second problem asks a better way to fill the invocation arguments from an array.
ES5 provides .apply() method on the function object to solve this. Unfortunately, this technique has 3 problems:

  • It's necessary to indicate manually the context of the function invocation
  • Is not possible to use in a constructor invocation
  • A shorter solution is more preferable

Let's see an example of .apply() usage:


const countries = ['Moldova', 'Ukraine'];
const otherCountries = ['USA', 'Japan'];
countries.push.apply(countries, otherCountries);
console.log(countries); // => ['Moldova', 'Ukraine', 'USA', 'Japan']

As mentioned, it seems irrelevant to indicate in .apply() the second time the context countries. The property accessor countries.push is enough to determine the method invocation on an object.
And the entire invocation looks verbose.

The JavaScript spread operator fills the function invocation arguments with values from an array (or more strictly from an iterable object, see 5.).
Let's improve the above sample with a spread operator:


const countries = ['Moldova', 'Ukraine'];
const otherCountries = ['USA', 'Japan'];
countries.push(...otherCountries);
console.log(countries); // => ['Moldova', 'Ukraine', 'USA', 'Japan']

As seen, the spread operator is a cleaner and straightforward solution. The only additional characters are 3 dots (...).

The JavaScript spread operator configures the constructor invocation arguments from an array, which is not possible directly when using .apply(). Let's see an example:


class King {
constructor(name, country) {
this.name = name;
this.country = country;
}
getDescription() {
return `${this.name} leads ${this.country}`;
}
}
const details = ['Alexander the Great', 'Greece'];
const Alexander = new King(...details);
Alexander.getDescription(); // => 'Alexander the Great leads Greece'

Moreover you can combine multiple spread operators and regular arguments in the same invocation. The following example is removing from an array existing elements, then adds other array and an element:


const numbers = [1, 2];
const evenNumbers = [4, 8];
const zero = 0;
numbers.splice(0, 2, ...evenNumbers, zero);
console.log(numbers); // => [4, 8, 0]

4. Improved array manipulation

4.1 Array construction

The array literal [item1, item2, .., itemN] does not provide functionality other than enumerating the initial array elements.

The spread operator improves array literals by allowing to insert on the fly other arrays (or any other iterables) into the initialized instance. This improvement makes easier to accomplish common tasks described below.

Create an array with initial elements from another array:


const initial = [0, 1];
const numbers1 = [...initial, 5, 7];
console.log(numbers1); // => [0, 1, 5, 7]
const numbers2 = [4, 8, ...initial];
console.log(numbers2); // => [4, 8, 0, 1]

number1 and number2 arrays are created using an array literal and in the meantime initialized with items from initial.

Concatenate 2 or more arrays:


const odds = [1, 5, 7];
const evens = [4, 6, 8];
const all = [...odds, ...evens];
console.log(all); // => [1, 5, 7, 4, 6, 8]

all array is created from concatentation of odds and evens arrays.

Clone an array instance:


const words = ['Hi', 'Hello', 'Good day'];
const otherWords = [...words];
console.log(otherWords); // => ['Hi', 'Hello', 'Good day']
console.log(otherWords === words); // => false

otherWords is a clone version of words array. Notice that cloning happens only on the array itself, but not on the contained elements (i.e. it's not a deep clone).

4.2 Array destructure

Destructuring assignments, available in ECMAScript 6, are powerful expressions to extract data from arrays and objects.

As a part of the destructuring, the rest operator extracts parts of an array. The extraction result is always an array.

In terms of syntax, the rest operator should be the last one in a destructuring assignment: [extractedItem1, ...restArray] = destructuredArray.

Let's see some applications:


const seasons = ['winter', 'spring', 'summer', 'autumn'];
const head, restArray;
[head, ...restArray] = seasons;
console.log(head); // => 'winter'
console.log(restArray); // => ['spring', 'summer', 'autumn']

[head, ...restArray] extracts the first item 'winter' into head variable and the rest of elements into restArray.

5. Spread operator and iteration protocols

The spread operator uses iteration protocols to navigate over elements of a collection. This makes the spread operator even more valuable because any object can define how the operator will extract data.

An object is Iterable when it conforms to iterable protocol.

The iterable protocol requires the object to contain a special property. The property name must be Symbol.iterator and value as a function that returns an iterator object.


interface Iterable {
[Symbol.iterator]() {
//...
return Iterator;
}
}

The Iterator object must conform to iterator protocol.

It needs to provide a property next, which value is a function that returns an object with properties done (a boolean to indicate the end of iteration) and value (the iteration result).


interface Iterator {
next() {
//...
return {
value: <value>,
done: <boolean>
};
};
}

It seems tough to understand the iteration protocols from verbal description, but the code behind those is quite simple.

The object or primitive must be iterable so that the spread operator to extract data from it.

Many native primitive types and objects are iterable: strings, arrays, typed arrays, sets and maps. They work by default with the spread operator.

For instance, let's see how a string conforms to iteration protocols:


const str = 'hi';
const iterator = str[Symbol.iterator]();
iterator.toString(); // => '[object String Iterator]'
iterator.next(); // => { value: 'h', done: false }
iterator.next(); // => { value: 'i', done: false }
iterator.next(); // => { value: undefined, done: true }
[...str]; // => ['h', 'i']

I like the spread operator for its ability to use the object's custom iteration implementation. You can control how the spread operator consumes your object - an effective coding technique.

The following sample makes an array-like object conformed to iteration protocols, then transforms it to an array using spread operator:


function iterator() {
let index = 0;
return {
next: () => ({ // Conform to Iterator protocol
done : index >= this.length,
value: this[index++]
})
};
}
const arrayLike = {
0: 'Cat',
1: 'Bird',
length: 2
};
// Conform to Iterable Protocol
arrayLike[Symbol.iterator] = iterator;
const array = [...arrayLike];
console.log(array); // => ['Cat', 'Bird']

arrayLike[Symbol.iterator] creates a property on the object that contains an iteration function iterator(), making the object conformed to iterable protocol.

iterator() returns an object (conformed to iteration protocol) with next property as a function that returns the control object {done: <boolean>, value: <item>}.

Since arrayLike is now iterable, spread operator is used to extract its elements into an array: [...arrayLike].

6. Finale

Three dots operator adds a bunch of great features to JavaScript.

The rest parameter makes it a lot easier to collect the arguments. It's a reasonable replacement for the hardcoded array-like object arguments. If the situation permits to choose between the rest parameter and arguments, use the first one.

.apply() method is not convenient for its verbose syntax. The spread operator is a good alternative when invocation arguments should be taken from an array.

The spread operator improves array literals. You can initialize, concatenate and clone arrays a lot simpler.

You can extract parts of an array using destructuring assignments. In combination with iteration protocols, the spread operator can be used in a more configurable manner.

I hope from now on the spread operator will appear more often in your code!

Like the post? Please share!

Dmitri Pavlutin

About Dmitri Pavlutin

Software developer and sometimes writer. My daily routine consists of (but not limited to) drinking coffee, coding, writing, overcoming boredom 😉. Living in the sunny Barcelona. 🇪🇸