28 Objects

In this book, JavaScript’s style of object-oriented programming (OOP) is introduced in four steps. This chapter covers step 1 and 2; the next chapter covers step 3 and 4. The steps are (fig. 8):

1. Single objects (this chapter): How do objects, JavaScript’s basic OOP building blocks, work in isolation?
2. Prototype chains (this chapter): Each object has a chain of zero or more prototype objects. Prototypes are JavaScript’s core inheritance mechanism.
3. Classes (next chapter): JavaScript’s classes are factories for objects. The relationship between a class and its instances is based on prototypal inheritance (step 2).
4. Subclassing (next chapter): The relationship between a subclass and its superclass is also based on prototypal inheritance.

28.1 Cheat sheet: objects

28.1.1 Single objects

Creating an object via an object literal (starts and ends with a curly brace):

const myObject = { // object literal
  myProperty: 1,
  myMethod() {
    return 2;
  }, // comma!
  get myAccessor() {
    return this.myProperty;
  }, // comma!
  set myAccessor(value) {
    this.myProperty = value;
  }, // last comma is optional
};

assert.equal(
  myObject.myProperty, 1
);
assert.equal(
  myObject.myMethod(), 2
);
assert.equal(
  myObject.myAccessor, 1
);
myObject.myAccessor = 3;
assert.equal(
  myObject.myProperty, 3
);

Being able to create objects directly (without classes) is one of the highlights of JavaScript.

Spreading into objects:

const original = {
  a: 1,
  b: {
    c: 3,
  },
};

// Spreading (...) copies one object “into” another one:
const modifiedCopy = {
  ...original, // spreading
  d: 4,
};

assert.deepEqual(
  modifiedCopy,
  {
    a: 1,
    b: {
      c: 3,
    },
    d: 4,
  }
);

// Caveat: spreading copies shallowly (property values are shared)
modifiedCopy.a = 5; // does not affect `original`
modifiedCopy.b.c = 6; // affects `original`
assert.deepEqual(
  original,
  {
    a: 1, // unchanged
    b: {
      c: 6, // changed
    },
  },
);

We can also use spreading to make an unmodified (shallow) copy of an object:

const exactCopy = {...obj};

28.1.2 Prototype chains

Prototypes are JavaScript’s fundamental inheritance mechanism. Even classes are based on it. Each object has null or an object as its prototype. The latter object can also have a prototype, etc. In general, we get chains of prototypes.

Prototypes are managed like this:

// `obj1` has no prototype (its prototype is `null`)
const obj1 = Object.create(null); // (A)
assert.equal(
  Object.getPrototypeOf(obj1), null // (B)
);

// `obj2` has the prototype `proto`
const proto = {
  protoProp: 'protoProp',
};
const obj2 = {
  __proto__: proto, // (C)
  objProp: 'objProp',
}
assert.equal(
  Object.getPrototypeOf(obj2), proto
);

Notes:

• Setting an object’s prototype while creating the object: line A, line C
• Retrieving the prototype of an object: line B

Each object inherits all the properties of its prototype:

// `obj2` inherits .protoProp from `proto`
assert.equal(
  obj2.protoProp, 'protoProp'
);
assert.deepEqual(
  Reflect.ownKeys(obj2),
  ['objProp'] // own properties of `obj2`
);

The non-inherited properties of an object are called its own properties.

The most important use case for prototypes is that several objects can share methods by inheriting them from a common prototype.

28.2 What is an object?

Objects in JavaScript:

• An object is a set of slots (key-value entries).
• Public slots are called properties:
  • A property key can only be a string or a symbol.
• Private slots can only be created via classes and are explained in §29.2.4 “Public slots (properties) vs. private slots”.

28.2.1 The two ways of using objects

There are two ways of using objects in JavaScript:

• Fixed-layout objects: Used this way, objects work like records in databases. They have a fixed number of properties, whose keys are known at development time. Their values generally have different types.

  const fixedLayoutObject = {
    product: 'carrot',
    quantity: 4,
  };

• Dictionary objects: Used this way, objects work like lookup tables or maps. They have a variable number of properties, whose keys are not known at development time. All of their values have the same type.

  const dictionaryObject = {
    ['one']: 1,
    ['two']: 2,
  };

Note that the two ways can also be mixed: Some objects are both fixed-layout objects and dictionary objects.

The ways of using objects influence how they are explained in this chapter:

• First, we’ll explore fixed-layout objects. Even though property keys are strings or symbols under the hood, they will appear as fixed identifiers to us.
• Later, we’ll explore dictionary objects. Note that Maps are usually better dictionaries than objects. However, some of the operations that we’ll encounter are also useful for fixed-layout objects.

28.3 Fixed-layout objects

Let’s first explore fixed-layout objects.

28.3.1 Object literals: properties

Object literals are one way of creating fixed-layout objects. They are a stand-out feature of JavaScript: we can directly create objects – no need for classes! This is an example:

const jane = {
  first: 'Jane',
  last: 'Doe', // optional trailing comma
};

In the example, we created an object via an object literal, which starts and ends with curly braces {}. Inside it, we defined two properties (key-value entries):

• The first property has the key first and the value 'Jane'.
• The second property has the key last and the value 'Doe'.

Since ES5, trailing commas are allowed in object literals.

We will later see other ways of specifying property keys, but with this way of specifying them, they must follow the rules of JavaScript variable names. For example, we can use first_name as a property key, but not first-name). However, reserved words are allowed:

const obj = {
  if: true,
  const: true,
};

In order to check the effects of various operations on objects, we’ll occasionally use Object.keys() in this part of the chapter. It lists property keys:

> Object.keys({a:1, b:2})
[ 'a', 'b' ]

28.3.2 Object literals: property value shorthands

Whenever the value of a property is defined via a variable that has the same name as the key, we can omit the key.

function createPoint(x, y) {
  return {x, y}; // Same as: {x: x, y: y}
}
assert.deepEqual(
  createPoint(9, 2),
  { x: 9, y: 2 }
);

28.3.3 Getting properties

This is how we get (read) a property (line A):

const jane = {
  first: 'Jane',
  last: 'Doe',
};

// Get property .first
assert.equal(jane.first, 'Jane'); // (A)

Getting an unknown property produces undefined:

assert.equal(jane.unknownProperty, undefined);

28.3.4 Setting properties

This is how we set (write to) a property (line A):

const obj = {
  prop: 1,
};
assert.equal(obj.prop, 1);
obj.prop = 2; // (A)
assert.equal(obj.prop, 2);

We just changed an existing property via setting. If we set an unknown property, we create a new entry:

const obj = {}; // empty object
assert.deepEqual(
  Object.keys(obj), []);

obj.unknownProperty = 'abc';
assert.deepEqual(
  Object.keys(obj), ['unknownProperty']);

28.3.5 Object literals: methods

The following code shows how to create the method .says() via an object literal:

const jane = {
  first: 'Jane', // value property
  says(text) {   // method
    return `${this.first} says “${text}”`; // (A)
  }, // comma as separator (optional at end)
};
assert.equal(jane.says('hello'), 'Jane says “hello”');

During the method call jane.says('hello'), jane is called the receiver of the method call and assigned to the special variable this (more on this in §28.5 “Methods and the special variable this”). That enables method .says() to access the sibling property .first in line A.

28.3.6 Object literals: accessors

An accessor is defined via syntax inside an object literal that looks like methods: a getter and/or a setter (i.e., each accessor has one or both of them).

Invoking an accessor looks like accessing a value property:

• Reading the property invokes the getter.
• Writing to the property invokes the setter.

28.3.6.1 Getters

A getter is created by prefixing a method definition with the modifier get:

const jane = {
  first: 'Jane',
  last: 'Doe',
  get full() {
    return `${this.first} ${this.last}`;
  },
};

assert.equal(jane.full, 'Jane Doe');
jane.first = 'John';
assert.equal(jane.full, 'John Doe');

28.3.6.2 Setters

A setter is created by prefixing a method definition with the modifier set:

const jane = {
  first: 'Jane',
  last: 'Doe',
  set full(fullName) {
    const parts = fullName.split(' ');
    this.first = parts[0];
    this.last = parts[1];
  },
};

jane.full = 'Richard Roe';
assert.equal(jane.first, 'Richard');
assert.equal(jane.last, 'Roe');

28.4 Spreading into object literals (...) [ES2018]

Inside an object literal, a spread property adds the properties of another object to the current one:

> const obj = {one: 1, two: 2};
> {...obj, three: 3}
{ one: 1, two: 2, three: 3 }

const obj1 = {one: 1, two: 2};
const obj2 = {three: 3};
assert.deepEqual(
  {...obj1, ...obj2, four: 4},
  {one: 1, two: 2, three: 3,  four: 4}
);

If property keys clash, the property that is mentioned last “wins”:

> const obj = {one: 1, two: 2, three: 3};
> {...obj, one: true}
{ one: true, two: 2, three: 3 }
> {one: true, ...obj}
{ one: 1, two: 2, three: 3 }

All values are spreadable, even undefined and null:

> {...undefined}
{}
> {...null}
{}
> {...123}
{}
> {...'abc'}
{ '0': 'a', '1': 'b', '2': 'c' }
> {...['a', 'b']}
{ '0': 'a', '1': 'b' }

Property .length of strings and Arrays is hidden from this kind of operation (it is not enumerable; see §28.8.1 “Property attributes and property descriptors [ES5]” for more information).

Spreading includes properties whose keys are symbols (which are ignored by Object.keys(), Object.values() and Object.entries()):

const symbolKey = Symbol('symbolKey');
const obj = {
  stringKey: 1,
  [symbolKey]: 2,
};
assert.deepEqual(
  {...obj, anotherStringKey: 3},
  {
    stringKey: 1,
    [symbolKey]: 2,
    anotherStringKey: 3,
  }
);

28.4.1 Use case for spreading: copying objects

We can use spreading to create a copy of an object original:

const copy = {...original};

Caveat – copying is shallow: copy is a fresh object with duplicates of all properties (key-value entries) of original. But if property values are objects, then those are not copied themselves; they are shared between original and copy. Let’s look at an example:

const original = { a: 1, b: {prop: true} };
const copy = {...original};

The first level of copy is really a copy: If we change any properties at that level, it does not affect the original:

copy.a = 2;
assert.deepEqual(
  original, { a: 1, b: {prop: true} }); // no change

However, deeper levels are not copied. For example, the value of .b is shared between original and copy. Changing .b in the copy also changes it in the original.

copy.b.prop = false;
assert.deepEqual(
  original, { a: 1, b: {prop: false} });

28.4.2 Use case for spreading: default values for missing properties

If one of the inputs of our code is an object with data, we can make properties optional by specifying default values that are used if those properties are missing. One technique for doing so is via an object whose properties contain the default values. In the following example, that object is DEFAULTS:

const DEFAULTS = {alpha: 'a', beta: 'b'};
const providedData = {alpha: 1};

const allData = {...DEFAULTS, ...providedData};
assert.deepEqual(allData, {alpha: 1, beta: 'b'});

The result, the object allData, is created by copying DEFAULTS and overriding its properties with those of providedData.

But we don’t need an object to specify the default values; we can also specify them inside the object literal, individually:

const providedData = {alpha: 1};

const allData = {alpha: 'a', beta: 'b', ...providedData};
assert.deepEqual(allData, {alpha: 1, beta: 'b'});

28.4.3 Use case for spreading: non-destructively changing properties

So far, we have encountered one way of changing a property .alpha of an object: We set it (line A) and mutate the object. That is, this way of changing a property is destructive.

const obj = {alpha: 'a', beta: 'b'};
obj.alpha = 1; // (A)
assert.deepEqual(obj, {alpha: 1, beta: 'b'});

With spreading, we can change .alpha non-destructively – we make a copy of obj where .alpha has a different value:

const obj = {alpha: 'a', beta: 'b'};
const updatedObj = {...obj, alpha: 1};
assert.deepEqual(updatedObj, {alpha: 1, beta: 'b'});

28.4.4 “Destructive spreading”: Object.assign() [ES6]

Object.assign() is a tool method:

Object.assign(target, source_1, source_2, ···)

This expression assigns all properties of source_1 to target, then all properties of source_2, etc. At the end, it returns target – for example:

const target = { a: 1 };

const result = Object.assign(
  target,
  {b: 2},
  {c: 3, b: true});

assert.deepEqual(
  result, { a: 1, b: true, c: 3 });
// target was modified and returned:
assert.equal(result, target);

The use cases for Object.assign() are similar to those for spread properties. In a way, it spreads destructively.

28.5 Methods and the special variable this

28.5.1 Methods are properties whose values are functions

Let’s revisit the example that was used to introduce methods:

const jane = {
  first: 'Jane',
  says(text) {
    return `${this.first} says “${text}”`;
  },
};

Somewhat surprisingly, methods are functions:

assert.equal(typeof jane.says, 'function');

Why is that? We learned in the chapter on callable values that ordinary functions play several roles. Method is one of those roles. Therefore, internally, jane roughly looks as follows.

const jane = {
  first: 'Jane',
  says: function (text) {
    return `${this.first} says “${text}”`;
  },
};

28.5.2 The special variable this

Consider the following code:

const obj = {
  someMethod(x, y) {
    assert.equal(this, obj); // (A)
    assert.equal(x, 'a');
    assert.equal(y, 'b');
  }
};
obj.someMethod('a', 'b'); // (B)

In line B, obj is the receiver of a method call. It is passed to the function stored in obj.someMethod via an implicit (hidden) parameter whose name is this (line A).

28.5.3 Methods and .call()

Methods are functions and functions have methods themselves. One of those methods is .call(). Let’s look at an example to understand how this method works.

In the previous section, there was this method invocation:

obj.someMethod('a', 'b')

This invocation is equivalent to:

obj.someMethod.call(obj, 'a', 'b');

Which is also equivalent to:

const func = obj.someMethod;
func.call(obj, 'a', 'b');

.call() makes the normally implicit parameter this explicit: When invoking a function via .call(), the first parameter is this, followed by the regular (explicit) function parameters.

As an aside, this means that there are actually two different dot operators:

1. One for accessing properties: obj.prop
2. Another one for calling methods: obj.prop()

They are different in that (2) is not just (1) followed by the function call operator (). Instead, (2) additionally provides a value for this.

28.5.4 Methods and .bind()

.bind() is another method of function objects. In the following code, we use .bind() to turn method .says() into the stand-alone function func():

const jane = {
  first: 'Jane',
  says(text) {
    return `${this.first} says “${text}”`; // (A)
  },
};

const func = jane.says.bind(jane, 'hello');
assert.equal(func(), 'Jane says “hello”');

Setting this to jane via .bind() is crucial here. Otherwise, func() wouldn’t work properly because this is used in line A. In the next section, we’ll explore why that is.

28.5.5 this pitfall: extracting methods

We now know quite a bit about functions and methods and are ready to take a look at the biggest pitfall involving methods and this: function-calling a method extracted from an object can fail if we are not careful.

In the following example, we fail when we extract method jane.says(), store it in the variable func, and function-call func.

const jane = {
  first: 'Jane',
  says(text) {
    return `${this.first} says “${text}”`;
  },
};
const func = jane.says; // extract the method
assert.throws(
  () => func('hello'), // (A)
  {
    name: 'TypeError',
    message: "Cannot read properties of undefined (reading 'first')",
  });

In line A, we are making a normal function call. And in normal function calls, this is undefined (if strict mode is active, which it almost always is). Line A is therefore equivalent to:

assert.throws(
  () => jane.says.call(undefined, 'hello'), // `this` is undefined!
  {
    name: 'TypeError',
    message: "Cannot read properties of undefined (reading 'first')",
  }
);

How do we fix this? We need to use .bind() to extract method .says():

const func2 = jane.says.bind(jane);
assert.equal(func2('hello'), 'Jane says “hello”');

The .bind() ensures that this is always jane when we call func().

We can also use arrow functions to extract methods:

const func3 = text => jane.says(text);
assert.equal(func3('hello'), 'Jane says “hello”');

28.5.5.1 Example: extracting a method

The following is a simplified version of code that we may see in actual web development:

class ClickHandler {
  constructor(id, elem) {
    this.id = id;
    elem.addEventListener('click', this.handleClick); // (A)
  }
  handleClick(event) {
    alert('Clicked ' + this.id);
  }
}

In line A, we don’t extract the method .handleClick() properly. Instead, we should do:

const listener = this.handleClick.bind(this);
elem.addEventListener('click', listener);

// Later, possibly:
elem.removeEventListener('click', listener);

Each invocation of .bind() creates a new function. That’s why we need to store the result somewhere if we want to remove it later on.

28.5.5.2 How to avoid the pitfall of extracting methods

Alas, there is no simple way around the pitfall of extracting methods: Whenever we extract a method, we have to be careful and do it properly – for example, by binding this or by using an arrow function.

28.5.6 this pitfall: accidentally shadowing this

Consider the following problem: when we are inside an ordinary function, we can’t access the this of the surrounding scope because the ordinary function has its own this. In other words, a variable in an inner scope hides a variable in an outer scope. That is called shadowing. The following code is an example:

const prefixer = {
  prefix: '==> ',
  prefixStringArray(stringArray) {
    return stringArray.map(
      function (x) {
        return this.prefix + x; // (A)
      });
  },
};
assert.throws(
  () => prefixer.prefixStringArray(['a', 'b']),
  {
    name: 'TypeError',
    message: "Cannot read properties of undefined (reading 'prefix')",
  }
);

In line A, we want to access the this of .prefixStringArray(). But we can’t since the surrounding ordinary function has its own this that shadows (and blocks access to) the this of the method. The value of the former this is undefined due to the callback being function-called. That explains the error message.

The simplest way to fix this problem is via an arrow function, which doesn’t have its own this and therefore doesn’t shadow anything:

const prefixer = {
  prefix: '==> ',
  prefixStringArray(stringArray) {
    return stringArray.map(
      (x) => {
        return this.prefix + x;
      });
  },
};
assert.deepEqual(
  prefixer.prefixStringArray(['a', 'b']),
  ['==> a', '==> b']);

We can also store this in a different variable (line A), so that it doesn’t get shadowed:

prefixStringArray(stringArray) {
  const that = this; // (A)
  return stringArray.map(
    function (x) {
      return that.prefix + x;
    });
},

Another option is to specify a fixed this for the callback via .bind() (line A):

prefixStringArray(stringArray) {
  return stringArray.map(
    function (x) {
      return this.prefix + x;
    }.bind(this)); // (A)
},

Lastly, .map() lets us specify a value for this (line A) that it uses when invoking the callback:

prefixStringArray(stringArray) {
  return stringArray.map(
    function (x) {
      return this.prefix + x;
    },
    this); // (A)
},

28.5.6.1 Avoiding the pitfall of accidentally shadowing this

If we follow the advice in §25.3.4 “Recommendation: prefer specialized functions over ordinary functions”, we can avoid the pitfall of accidentally shadowing this. This is a summary:

• Use arrow functions as anonymous inline functions. They don’t have this as an implicit parameter and don’t shadow it.

• For named stand-alone function declarations we can either use arrow functions or function declarations. If we do the latter, we have to make sure this isn’t mentioned in their bodies.

28.5.7 The value of this in various contexts (advanced)

What is the value of this in various contexts?

Inside a callable entity, the value of this depends on how the callable entity is invoked and what kind of callable entity it is:

• Function call:
  • Ordinary functions: this === undefined (in strict mode)
  • Arrow functions: this is same as in surrounding scope (lexical this)
• Method call: this is receiver of call
• new: this refers to the newly created instance

We can also access this in all common top-level scopes:

• <script> element: this === globalThis
• ECMAScript modules: this === undefined
• CommonJS modules: this === module.exports

28.6 Optional chaining for property getting and method calls [ES2020] (advanced)

The following kinds of optional chaining operations exist:

obj?.prop     // optional fixed property getting
obj?.[«expr»] // optional dynamic property getting
func?.(«arg0», «arg1») // optional function or method call

The rough idea is:

• If the value before the question mark is neither undefined nor null, then perform the operation after the question mark.
• Otherwise, return undefined.

Each of the three syntaxes is covered in more detail later. These are a few first examples:

> null?.prop
undefined
> {prop: 1}?.prop
1

> null?.(123)
undefined
> String?.(123)
'123'

28.6.1 Example: optional fixed property getting

Consider the following data:

const persons = [
  {
    surname: 'Zoe',
    address: {
      street: {
        name: 'Sesame Street',
        number: '123',
      },
    },
  },
  {
    surname: 'Mariner',
  },
  {
    surname: 'Carmen',
    address: {
    },
  },
];

We can use optional chaining to safely extract street names:

const streetNames = persons.map(
  p => p.address?.street?.name);
assert.deepEqual(
  streetNames, ['Sesame Street', undefined, undefined]
);

28.6.1.1 Handling defaults via nullish coalescing

The nullish coalescing operator allows us to use the default value '(no name)' instead of undefined:

const streetNames = persons.map(
  p => p.address?.street?.name ?? '(no name)');
assert.deepEqual(
  streetNames, ['Sesame Street', '(no name)', '(no name)']
);

28.6.2 The operators in more detail (advanced)

28.6.2.1 Optional fixed property getting

The following two expressions are equivalent:

o?.prop
(o !== undefined && o !== null) ? o.prop : undefined

Examples:

assert.equal(undefined?.prop, undefined);
assert.equal(null?.prop,      undefined);
assert.equal({prop:1}?.prop,  1);

28.6.2.2 Optional dynamic property getting

The following two expressions are equivalent:

o?.[«expr»]
(o !== undefined && o !== null) ? o[«expr»] : undefined

Examples:

const key = 'prop';
assert.equal(undefined?.[key], undefined);
assert.equal(null?.[key], undefined);
assert.equal({prop:1}?.[key], 1);

28.6.2.3 Optional function or method call

The following two expressions are equivalent:

f?.(arg0, arg1)
(f !== undefined && f !== null) ? f(arg0, arg1) : undefined

Examples:

assert.equal(undefined?.(123), undefined);
assert.equal(null?.(123), undefined);
assert.equal(String?.(123), '123');

Note that this operator produces an error if its left-hand side is not callable:

assert.throws(
  () => true?.(123),
  TypeError);

Why? The idea is that the operator only tolerates deliberate omissions. An uncallable value (other than undefined and null) is probably an error and should be reported, rather than worked around.

28.6.3 Short-circuiting with optional property getting

In a chain of property gettings and method invocations, evaluation stops once the first optional operator encounters undefined or null at its left-hand side:

function invokeM(value) {
  return value?.a.b.m(); // (A)
}

const obj = {
  a: {
    b: {
      m() { return 'result' }
    }
  }
};
assert.equal(
  invokeM(obj), 'result'
);
assert.equal(
  invokeM(undefined), undefined // (B)
);

Consider invokeM(undefined) in line B: undefined?.a is undefined. Therefore we’d expect .b to fail in line A. But it doesn’t: The ?. operator encounters the value undefined and the evaluation of the whole expression immediately returns undefined.

This behavior differs from a normal operator where JavaScript always evaluates all operands before evaluating the operator. It is called short-circuiting. Other short-circuiting operators are:

• (a && b): b is only evaluated if a is truthy.
• (a || b): b is only evaluated if a is falsy.
• (c ? t : e): If c is truthy, t is evaluated. Otherwise, e is evaluated.

28.6.4 Optional chaining: downsides and alternatives

Optional chaining also has downsides:

• Deeply nested structures are more difficult to manage. For example, refactoring is harder if there are many sequences of property names: Each one enforces the structure of multiple objects.
• Being so forgiving when accessing data hides problems that will surface much later and are then harder to debug. For example, a typo early in a sequence of optional property names has more negative effects than a normal typo.

An alternative to optional chaining is to extract the information once, in a single location:

• We can either write a helper function that extracts the data.
• Or we can write a function whose input is deeply nested data and whose output is simpler, normalized data.

With either approach, it is possible to perform checks and to fail early if there are problems.

Further reading:

• “Overly defensive programming” by Carl Vitullo
• Thread on Twitter by Cory House

28.6.5 Frequently asked questions

28.6.5.1 What is a good mnemonic for the optional chaining operator (?.)?

Are you occasionally unsure if the optional chaining operator starts with a dot (.?) or a question mark (?.)? Then this mnemonic may help you:

• IF (?) the left-hand side is not nullish
• THEN (.) access a property.

28.6.5.2 Why are there dots in o?.[x] and f?.()?

The syntaxes of the following two optional operator are not ideal:

obj?.[«expr»]          // better: obj?[«expr»]
func?.(«arg0», «arg1») // better: func?(«arg0», «arg1»)

Alas, the less elegant syntax is necessary because distinguishing the ideal syntax (first expression) from the conditional operator (second expression) is too complicated:

obj?['a', 'b', 'c'].map(x => x+x)
obj ? ['a', 'b', 'c'].map(x => x+x) : []

28.6.5.3 Why does null?.prop evaluate to undefined and not null?

The operator ?. is mainly about its right-hand side: Does property .prop exist? If not, stop early. Therefore, keeping information about its left-hand side is rarely useful. However, only having a single “early termination” value does simplify things.

28.7 Dictionary objects (advanced)

Objects work best as fixed-layout objects. But before ES6, JavaScript did not have a data structure for dictionaries (ES6 brought Maps). Therefore, objects had to be used as dictionaries, which imposed a signficant constraint: Dictionary keys had to be strings (symbols were also introduced with ES6).

We first look at features of objects that are related to dictionaries but also useful for fixed-layout objects. This section concludes with tips for actually using objects as dictionaries. (Spoiler: If possible, it’s better to use Maps.)

28.7.1 Quoted keys in object literals

So far, we have always used fixed-layout objects. Property keys were fixed tokens that had to be valid identifiers and internally became strings:

const obj = {
  mustBeAnIdentifier: 123,
};

// Get property
assert.equal(obj.mustBeAnIdentifier, 123);

// Set property
obj.mustBeAnIdentifier = 'abc';
assert.equal(obj.mustBeAnIdentifier, 'abc');

As a next step, we’ll go beyond this limitation for property keys: In this subsection, we’ll use arbitrary fixed strings as keys. In the next subsection, we’ll dynamically compute keys.

Two syntaxes enable us to use arbitrary strings as property keys.

First, when creating property keys via object literals, we can quote property keys (with single or double quotes):

const obj = {
  'Can be any string!': 123,
};

Second, when getting or setting properties, we can use square brackets with strings inside them:

// Get property
assert.equal(obj['Can be any string!'], 123);

// Set property
obj['Can be any string!'] = 'abc';
assert.equal(obj['Can be any string!'], 'abc');

We can also use these syntaxes for methods:

const obj = {
  'A nice method'() {
    return 'Yes!';
  },
};

assert.equal(obj['A nice method'](), 'Yes!');

28.7.2 Computed keys in object literals

In the previous subsection, property keys were specified via fixed strings inside object literals. In this section we learn how to dynamically compute property keys. That enables us to use either arbitrary strings or symbols.

The syntax of dynamically computed property keys in object literals is inspired by dynamically accessing properties. That is, we can use square brackets to wrap expressions:

const obj = {
  ['Hello world!']: true,
  ['p'+'r'+'o'+'p']: 123,
  [Symbol.toStringTag]: 'Goodbye', // (A)
};

assert.equal(obj['Hello world!'], true);
assert.equal(obj.prop, 123);
assert.equal(obj[Symbol.toStringTag], 'Goodbye');

The main use case for computed keys is having symbols as property keys (line A).

Note that the square brackets operator for getting and setting properties works with arbitrary expressions:

assert.equal(obj['p'+'r'+'o'+'p'], 123);
assert.equal(obj['==> prop'.slice(4)], 123);

Methods can have computed property keys, too:

const methodKey = Symbol();
const obj = {
  [methodKey]() {
    return 'Yes!';
  },
};

assert.equal(obj[methodKey](), 'Yes!');

For the remainder of this chapter, we’ll mostly use fixed property keys again (because they are syntactically more convenient). But all features are also available for arbitrary strings and symbols.

28.7.3 The in operator: is there a property with a given key?

The in operator checks if an object has a property with a given key:

const obj = {
  alpha: 'abc',
  beta: false,
};

assert.equal('alpha' in obj, true);
assert.equal('beta' in obj, true);
assert.equal('unknownKey' in obj, false);

28.7.3.1 Checking if a property exists via truthiness

We can also use a truthiness check to determine if a property exists:

assert.equal(
  obj.alpha ? 'exists' : 'does not exist',
  'exists');
assert.equal(
  obj.unknownKey ? 'exists' : 'does not exist',
  'does not exist');

The previous checks work because obj.alpha is truthy and because reading a missing property returns undefined (which is falsy).

There is, however, one important caveat: truthiness checks fail if the property exists, but has a falsy value (undefined, null, false, 0, "", etc.):

assert.equal(
  obj.beta ? 'exists' : 'does not exist',
  'does not exist'); // should be: 'exists'

28.7.4 Deleting properties

We can delete properties via the delete operator:

const obj = {
  myProp: 123,
};

assert.deepEqual(Object.keys(obj), ['myProp']);
delete obj.myProp;
assert.deepEqual(Object.keys(obj), []);

28.7.5 Enumerability

Enumerability is an attribute of a property. Non-enumerable properties are ignored by some operations – for example, by Object.keys() and when spreading properties. By default, most properties are enumerable. The next example shows how to change that and how it affects spreading.

const enumerableSymbolKey = Symbol('enumerableSymbolKey');
const nonEnumSymbolKey = Symbol('nonEnumSymbolKey');

// We create enumerable properties via an object literal
const obj = {
  enumerableStringKey: 1,
  [enumerableSymbolKey]: 2,
}

// For non-enumerable properties, we need a more powerful tool
Object.defineProperties(obj, {
  nonEnumStringKey: {
    value: 3,
    enumerable: false,
  },
  [nonEnumSymbolKey]: {
    value: 4,
    enumerable: false,
  },
});

// Non-enumerable properties are ignored by spreading:
assert.deepEqual(
  {...obj},
  {
    enumerableStringKey: 1,
    [enumerableSymbolKey]: 2,
  }
);

Object.defineProperties() is explained later in this chapter. The next subsection shows how these operations are affected by enumerability:

28.7.6 Listing property keys via Object.keys() etc.

Each of the methods in tbl. 19 returns an Array with the own property keys of the parameter. In the names of the methods, we can see that the following distinction is made:

• A property key can be either a string or a symbol.
• A property name is a property key whose value is a string.
• A property symbol is a property key whose value is a symbol.

To demonstrate the four operations, we revisit the example from the previous subsection:

const enumerableSymbolKey = Symbol('enumerableSymbolKey');
const nonEnumSymbolKey = Symbol('nonEnumSymbolKey');

const obj = {
  enumerableStringKey: 1,
  [enumerableSymbolKey]: 2,
}
Object.defineProperties(obj, {
  nonEnumStringKey: {
    value: 3,
    enumerable: false,
  },
  [nonEnumSymbolKey]: {
    value: 4,
    enumerable: false,
  },
});

assert.deepEqual(
  Object.keys(obj),
  ['enumerableStringKey']
);
assert.deepEqual(
  Object.getOwnPropertyNames(obj),
  ['enumerableStringKey', 'nonEnumStringKey']
);
assert.deepEqual(
  Object.getOwnPropertySymbols(obj),
  [enumerableSymbolKey, nonEnumSymbolKey]
);
assert.deepEqual(
  Reflect.ownKeys(obj),
  [
    'enumerableStringKey', 'nonEnumStringKey',
    enumerableSymbolKey, nonEnumSymbolKey,
  ]
);

28.7.7 Listing property values via Object.values()

Object.values() lists the values of all enumerable string-keyed properties of an object:

const firstName = Symbol('firstName');
const obj = {
  [firstName]: 'Jane',
  lastName: 'Doe',
};
assert.deepEqual(
  Object.values(obj),
  ['Doe']);

28.7.8 Listing property entries via Object.entries() [ES2017]

Object.entries() lists all enumerable string-keyed properties as key-value pairs. Each pair is encoded as a two-element Array:

const firstName = Symbol('firstName');
const obj = {
  [firstName]: 'Jane',
  lastName: 'Doe',
};
assert.deepEqual(
  Object.entries(obj),
  [
    ['lastName', 'Doe'],
  ]);

28.7.8.1 A simple implementation of Object.entries()

The following function is a simplified version of Object.entries():

function entries(obj) {
  return Object.keys(obj)
  .map(key => [key, obj[key]]);
}

28.7.9 Properties are listed deterministically

Own (non-inherited) properties of objects are always listed in the following order:

1. Properties with string keys that contain integer indices (that includes Array indices):
   In ascending numeric order
2. Remaining properties with string keys:
   In the order in which they were added
3. Properties with symbol keys:
   In the order in which they were added

The following example demonstrates how property keys are sorted according to these rules:

> Object.keys({b:0,a:0, 10:0,2:0})
[ '2', '10', 'b', 'a' ]

28.7.10 Assembling objects via Object.fromEntries() [ES2019]

Given an iterable over [key, value] pairs, Object.fromEntries() creates an object:

const symbolKey = Symbol('symbolKey');
assert.deepEqual(
  Object.fromEntries(
    [
      ['stringKey', 1],
      [symbolKey, 2],
    ]
  ),
  {
    stringKey: 1,
    [symbolKey]: 2,
  }
);

Object.fromEntries() does the opposite of Object.entries(). However, while Object.entries() ignores symbol-keyed properties, Object.fromEntries() doesn’t (see previous example).

To demonstrate both, we’ll use them to implement two tool functions from the library Underscore in the next subsubsections.

28.7.10.1 Example: pick()

The Underscore function pick() has the following signature:

pick(object, ...keys)

It returns a copy of object that has only those properties whose keys are mentioned in the trailing arguments:

const address = {
  street: 'Evergreen Terrace',
  number: '742',
  city: 'Springfield',
  state: 'NT',
  zip: '49007',
};
assert.deepEqual(
  pick(address, 'street', 'number'),
  {
    street: 'Evergreen Terrace',
    number: '742',
  }
);

We can implement pick() as follows:

function pick(object, ...keys) {
  const filteredEntries = Object.entries(object)
    .filter(([key, _value]) => keys.includes(key));
  return Object.fromEntries(filteredEntries);
}

28.7.10.2 Example: invert()

The Underscore function invert() has the following signature:

invert(object)

It returns a copy of object where the keys and values of all properties are swapped:

assert.deepEqual(
  invert({a: 1, b: 2, c: 3}),
  {1: 'a', 2: 'b', 3: 'c'}
);

We can implement invert() like this:

function invert(object) {
  const reversedEntries = Object.entries(object)
    .map(([key, value]) => [value, key]);
  return Object.fromEntries(reversedEntries);
}

28.7.10.3 A simple implementation of Object.fromEntries()

The following function is a simplified version of Object.fromEntries():

function fromEntries(iterable) {
  const result = {};
  for (const [key, value] of iterable) {
    let coercedKey;
    if (typeof key === 'string' || typeof key === 'symbol') {
      coercedKey = key;
    } else {
      coercedKey = String(key);
    }
    result[coercedKey] = value;
  }
  return result;
}

28.7.11 The pitfalls of using an object as a dictionary

If we use plain objects (created via object literals) as dictionaries, we have to look out for two pitfalls.

The first pitfall is that the in operator also finds inherited properties:

const dict = {};
assert.equal('toString' in dict, true);

We want dict to be treated as empty, but the in operator detects the properties it inherits from its prototype, Object.prototype.

The second pitfall is that we can’t use the property key __proto__ because it has special powers (it sets the prototype of the object):

const dict = {};

dict['__proto__'] = 123;
// No property was added to dict:
assert.deepEqual(Object.keys(dict), []);

28.7.11.1 Safely using objects as dictionaries

So how do we avoid the two pitfalls?

• If we can, we use Maps. They are the best solution for dictionaries.
• If we can’t, we use a library for objects-as-dictionaries that protects us from making mistakes.
• If that’s not possible or desired, we use an object without a prototype.

The following code demonstrates using prototype-less objects as dictionaries:

const dict = Object.create(null); // prototype is `null`

assert.equal('toString' in dict, false); // (A)

dict['__proto__'] = 123;
assert.deepEqual(Object.keys(dict), ['__proto__']);

We avoided both pitfalls:

• First, a property without a prototype does not inherit any properties (line A).
• Second, in modern JavaScript, __proto__ is implemented via Object.prototype. That means that it is switched off if Object.prototype is not in the prototype chain.

28.8 Property attributes and freezing objects (advanced)

28.8.1 Property attributes and property descriptors [ES5]

Just as objects are composed of properties, properties are composed of attributes. The value of a property is only one of several attributes. Others include:

• writable: Is it possible to change the value of the property?
• enumerable: Is the property considered by Object.keys(), spreading, etc.?

When we are using one of the operations for handling property attributes, attributes are specified via property descriptors: objects where each property represents one attribute. For example, this is how we read the attributes of a property obj.myProp:

const obj = { myProp: 123 };
assert.deepEqual(
  Object.getOwnPropertyDescriptor(obj, 'myProp'),
  {
    value: 123,
    writable: true,
    enumerable: true,
    configurable: true,
  });

And this is how we change the attributes of obj.myProp:

assert.deepEqual(Object.keys(obj), ['myProp']);

// Hide property `myProp` from Object.keys()
// by making it non-enumerable
Object.defineProperty(obj, 'myProp', {
  enumerable: false,
});

assert.deepEqual(Object.keys(obj), []);

Further reading:

• Enumerability is covered in greater detail earlier in this chapter.
• For more information on property attributes and property descriptors, see Deep JavaScript.

28.8.2 Freezing objects [ES5]

Object.freeze(obj) makes obj completely immutable: We can’t change properties, add properties, or change its prototype – for example:

const frozen = Object.freeze({ x: 2, y: 5 });
assert.throws(
  () => { frozen.x = 7 },
  {
    name: 'TypeError',
    message: /^Cannot assign to read only property 'x'/,
  });

Under the hood, Object.freeze() changes the attributes of properties (e.g., it makes them non-writable) and objects (e.g., it makes them non-extensible, meaning that no properties can be added anymore).

There is one caveat: Object.freeze(obj) freezes shallowly. That is, only the properties of obj are frozen but not objects stored in properties.

28.9 Prototype chains

Prototypes are JavaScript’s only inheritance mechanism: Each object has a prototype that is either null or an object. In the latter case, the object inherits all of the prototype’s properties.

In an object literal, we can set the prototype via the special property __proto__:

const proto = {
  protoProp: 'a',
};
const obj = {
  __proto__: proto,
  objProp: 'b',
};

// obj inherits .protoProp:
assert.equal(obj.protoProp, 'a');
assert.equal('protoProp' in obj, true);

Given that a prototype object can have a prototype itself, we get a chain of objects – the so-called prototype chain. Inheritance gives us the impression that we are dealing with single objects, but we are actually dealing with chains of objects.

Fig. 9 shows what the prototype chain of obj looks like.

Non-inherited properties are called own properties. obj has one own property, .objProp.

28.9.1 JavaScript’s operations: all properties vs. own properties

Some operations consider all properties (own and inherited) – for example, getting properties:

> const obj = { one: 1 };
> typeof obj.one // own
'number'
> typeof obj.toString // inherited
'function'

Other operations only consider own properties – for example, Object.keys():

> Object.keys(obj)
[ 'one' ]

Read on for another operation that also only considers own properties: setting properties.

28.9.2 Pitfall: only the first member of a prototype chain is mutated

Given an object obj with a chain of prototype objects, it makes sense that setting an own property of obj only changes obj. However, setting an inherited property via obj also only changes obj. It creates a new own property in obj that overrides the inherited property. Let’s explore how that works with the following object:

const proto = {
  protoProp: 'a',
};
const obj = {
  __proto__: proto,
  objProp: 'b',
};

In the next code snippet, we set the inherited property obj.protoProp (line A). That “changes” it by creating an own property: When reading obj.protoProp, the own property is found first and its value overrides the value of the inherited property.

// In the beginning, obj has one own property
assert.deepEqual(Object.keys(obj), ['objProp']);

obj.protoProp = 'x'; // (A)

// We created a new own property:
assert.deepEqual(Object.keys(obj), ['objProp', 'protoProp']);

// The inherited property itself is unchanged:
assert.equal(proto.protoProp, 'a');

// The own property overrides the inherited property:
assert.equal(obj.protoProp, 'x');

The prototype chain of obj is depicted in fig. 10.

28.9.3 Tips for working with prototypes (advanced)

28.9.3.1 Getting and setting prototypes

Recommendations for __proto__:

• Don’t use __proto__ as a pseudo-property (a setter of all instances of Object):

  • It can’t be used with all objects (e.g. objects that are not instances of Object).
  • The language specification has deprecated it.

  For more information on this feature see §29.8.7 “Object.prototype.__proto__ (accessor)”.

• Using __proto__ in object literals to set prototypes is different: It’s a feature of object literals that has no pitfalls.

The recommended ways of getting and setting prototypes are:

• Getting the prototype of an object:

  Object.getPrototypeOf(obj: Object) : Object

• The best time to set the prototype of an object is when we are creating it. We can do so via __proto__ in an object literal or via:

  Object.create(proto: Object) : Object

  If we have to, we can use Object.setPrototypeOf() to change the prototype of an existing object. But that may affect performance negatively.

This is how these features are used:

const proto1 = {};
const proto2a = {};
const proto2b = {};

const obj1 = {
  __proto__: proto1,
  a: 1,
  b: 2,
};
assert.equal(Object.getPrototypeOf(obj1), proto1);

const obj2 = Object.create(
  proto2a,
  {
    a: {
      value: 1,
      writable: true,
      enumerable: true,
      configurable: true,
    },
    b: {
      value: 2,
      writable: true,
      enumerable: true,
      configurable: true,
    },  
  }
);
assert.equal(Object.getPrototypeOf(obj2), proto2a);

Object.setPrototypeOf(obj2, proto2b);
assert.equal(Object.getPrototypeOf(obj2), proto2b);

28.9.3.2 Checking if an object is in the prototype chain of another object

So far, “proto is a prototype of obj” always meant “proto is a direct prototype of obj”. But it can also be used more loosely and mean that proto is in the prototype chain of obj. That looser relationship can be checked via .isPrototypeOf():

For example:

const a = {};
const b = {__proto__: a};
const c = {__proto__: b};

assert.equal(a.isPrototypeOf(b), true);
assert.equal(a.isPrototypeOf(c), true);

assert.equal(c.isPrototypeOf(a), false);
assert.equal(a.isPrototypeOf(a), false);

For more information on this method see §29.8.5 “Object.prototype.isPrototypeOf()”.

28.9.4 Object.hasOwn(): Is a given property own (non-inherited)? [ES2022]

The in operator (line A) checks if an object has a given property. In contrast, Object.hasOwn() (lines B and C) checks if a property is own.

const proto = {
  protoProp: 'protoProp',
};
const obj = {
  __proto__: proto,
  objProp: 'objProp',
}
assert.equal('protoProp' in obj, true); // (A)
assert.equal(Object.hasOwn(obj, 'protoProp'), false); // (B)
assert.equal(Object.hasOwn(proto, 'protoProp'), true); // (C)

28.9.5 Sharing data via prototypes

Consider the following code:

const jane = {
  firstName: 'Jane',
  describe() {
    return 'Person named '+this.firstName;
  },
};
const tarzan = {
  firstName: 'Tarzan',
  describe() {
    return 'Person named '+this.firstName;
  },
};

assert.equal(jane.describe(), 'Person named Jane');
assert.equal(tarzan.describe(), 'Person named Tarzan');

We have two objects that are very similar. Both have two properties whose names are .firstName and .describe. Additionally, method .describe() is the same. How can we avoid duplicating that method?

We can move it to an object PersonProto and make that object a prototype of both jane and tarzan:

const PersonProto = {
  describe() {
    return 'Person named ' + this.firstName;
  },
};
const jane = {
  __proto__: PersonProto,
  firstName: 'Jane',
};
const tarzan = {
  __proto__: PersonProto,
  firstName: 'Tarzan',
};

The name of the prototype reflects that both jane and tarzan are persons.

Fig. 11 illustrates how the three objects are connected: The objects at the bottom now contain the properties that are specific to jane and tarzan. The object at the top contains the properties that are shared between them.

When we make the method call jane.describe(), this points to the receiver of that method call, jane (in the bottom-left corner of the diagram). That’s why the method still works. tarzan.describe() works similarly.

assert.equal(jane.describe(), 'Person named Jane');
assert.equal(tarzan.describe(), 'Person named Tarzan');

Looking ahead to the next chapter on classes – this is how classes are organized internally:

• All instances share a common prototype with methods.
• Instance-specific data is stored in own properties in each instance.

§29.3 “The internals of classes” explains this in more detail.

28.10 FAQ: objects

28.10.1 Why do objects preserve the insertion order of properties?

In principle, objects are unordered. The main reason for ordering properties is so that operations that list entries, keys, or values are deterministic. That helps, e.g., with testing.