🌱 Tim's Dev Wiki

Search IconIcon to open search

TypeScript

Last updated September 19, 2022.

TypeScript is a programming language made by Microsoft that is superset of JavaScript. The reason that TypeScript exists is to make complex JavaScript projects more maintainable and less error-prone by introducing a static and strong type system. Essentially, it just gives developers a lot of quality-of-life improvements over JavaScript.

Note: TypeScript gets compiled (or more precisely, ‘transpiled’) to JavaScript in the end. This is not new, languages like CoffeeScript, Dart, Scala, etc. can all have JavaScript as what we call a compilation target.

I learned TypeScript from the official docs and from the ‘Programming TypeScript’ textbook by Boris Cherny.

Why TypeScript is loved

# Setup

Install Node.js, npm or yarn. Then, install tsc, the open-source typescript compiler, as a dev dependency in a javascript project:

1
2
yarn init   # Inside the project root directory, if it's a new project.
yarn add --dev typescript @types/node ts-node

# tsconfig.json

Every typescript project should have a tsconfig.json file at the root of the project directory. It tells tsc which files to compile, where to dump the resulting javascript, and so on. A basic config looks like this (but there are so many more options):

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
{
	"compilerOptions": {
		"lib": [   // Which APIs are available to the transpiled JS code? Eg. es2015 has Function.prototype.bind, etc.
			"es2015"
		],
		"module": "commonjs", // Which module system should be used by the transpiled JS code?
		"outDir": "dist",     // Where should the output JS files go?
		"sourceMap": true,    // Whether to generate a source map.
		"strict": true,       // All code must be properly typed.
		"target": "es2015"    // What JS version to compile to.
	},
	"include": [   // Directories containing .ts files we want to transpile.
		"src"
	]
}

Alternatively, you can generate a tsconfig.json with tsc --init.

Some recommended flags include:

# Typing

Broadly speaking, in programming languages, a type is a set of values, plus the properties/methods available to them.

# Assigning Types

Assigning types is straightforward in TypeScript, you just postfix a variable or parameter with a colon and a type.

1
2
3
// Typing variables.
let a: number = 42;
let b: string;

# Defining Types

Remember, types are just sets of values. When you define a type, you are just defining a set of values. The following are all examples of custom types you can define:

1
2
3
type TwoOrFour = 2 | 4;                           // The set consisting of 2 and 4.
type Value = string | number;                     // The set of all strings and all numbers.
type RandomThings = "Hello" | 42 | null | RegExp; // The set consisting of "Hello", 42, null and all instances of `RegExp`.

An important thing to understand about TypeScript (and many other statically-typed languages) is that it has separate namespaces for values and types. This means that in the following example, a variable identifier with the same name as a type alias are not in conflict. TypeScript can infer if you meant the value or the type.

1
2
type hello = "world";
const hello: hello = "world";

# Type Alias

You can declare type aliases in a very similar way as to how you define variables. Type aliases are block-scoped, just like local variables.

1
2
3
4
5
6
7
type Person = { age: number, name: string };

// Typescript will never infer that `me` is of type `Person` unless you explicitly say.
const me: Person = {
	age: 21,
	name: "Tim"
};

# Interface

Interfaces are basically an alternative to type aliases, but are mostly better suited for defining object shapes. You can’t use & or | for interfaces, but you can use extends.

1
2
3
4
5
6
7
// With type aliases, to add additional fields on top another type, you'd use `&`.
type Employee = { id: string; }
type SoftwareEngineer = Employee & { techStack: string[]; }

// With interfaces, you just use `extends`, similar to how you do class inheritance.
interface Employee { id: string }
interface SoftwareEngineer extends Employee { techStack: string[]; }

# Classes vs. Interfaces

Using interface does not actually generate any javascript code when transpiled. Using class, however, will generate JavaScript code, which enables instanceof to work at runtime. A class Foo { ... } definition actually creates a value Foo that can be used in expressions, and a type Foo that can be used as a type.

Interfaces don’t let you use access modifiers. You can’t supply implementations either.

Keep these critical differences in mind when deciding between class or interface.

# Assignability

Assignability is about what the rules are around an assignment like this: const a: A = b;, where b is of type B. For assignment to be valid, it must be the case that:

  1. B is a subtype of A (basically that $B \subseteq A$),
  2. … or B is any. Note: this rule only exists to make it easier to interoperate with javascript code.

# Union and Intersection

Again, types are just sets of values. To expand the size of a set, you can union it with other sets, and to narrow the size of a set, you can intersect it other sets. In TypeScript, we use | to union two types and & to intersect two types.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
type Student = { id: string; degree: string };
type Tutor = { id: string; courses: string[] };

type StudentOrTutor = Student | Tutor;    // Set of all objects that are either `Student`, `Tutor`, or both.
type StudentAndTutor = Student & Tutor;   // Set of all objects containing all fields of `Student` AND `Tutor`.

const student: Student = { id: '111', degree: 'Bachelor of Science' };
const tutor: Tutor = { id: '222', courses: ['CS101'] };
const studentTutor: StudentAndTutor = {
	id: '333',
	degree: 'Bachelor of Computer Science',
	courses: ['CS201'],
};

// For `Student | Tutor`: you can assign any of: `Student`, `Tutor`, or `Student & Tutor`
const uniGoers: StudentOrTutor[] = [student, tutor, studentTutor];

// For `Student & Tutor`: you can only have people who are simultaneously `Student` and `Tutor`.
const studentTutors: StudentAndTutor[] = [studentTutor];

To make a type nullable, you can union it with null like this: type MiddleName = string | null;

# TypeScript Built-In Types

TypeScript introduces some new built-in data types that aren’t present in JavaScript. These are: any, unknown, void, never.

# any

The any type represents the set of all values. You can assign a variable of type any to a number, a string, a WebServer object, etc. Only use any as a last resort – always prefer assigning the most specific type that you can. Often, people treat any as a way to ‘opt’ out of TypeScript for a small part of the code. When something is any, you are free to do erroneous things on it such as invoking methods on it that don’t exist, using in arithmetic expressions, accessing undefined properties, etc.

# unknown

The unknown type represents the set of all values, just like any. The difference is that TypeScript does not allow you to use an unknown value until you perform type checks or refinement to narrow down what specific type the unknown value is. For this reason, unknown is considered the type-safe version of any.

1
2
3
4
5
6
const a: unknown = 30;
if (typeof a === 'number') {
	const b: number = a + 1; // Now that we are certain `a` is a number, we can use it as one.
} else {
	const b: number = a + 1; // Error.
}
1
2
3
4
5
const a: any = "Hello";
const b: unknown = "World";

a.toLowerCase();   // This is fine since `a` is `any`.
b.toLowerCase();   // Error. We need a type check before 

A useful way to think about any and unknown is: any means “I don’t care”, unknown means “I don’t know (yet)”.

# Object Shape

An important part of using TypeScript effectively is in defining the shape of object values, or in other words, what properties they have and which ones must be assigned a value or are optional.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
// Let typescript infer the object shape.
// `person` will be of type: { age: number, name: string }.
const person = {
	age: 42,
	name: "Andrew"
};

// You can also explicitly specify the object type. You might do this if you want to set narrower types for the properties.
const person: { age: number, name: "Andrew" } = {
	age: 42,
	name: "Andrew"
};

// Or more cleanly, define a type:
type Andrew = { age: number, name: "Andrew" };
const person: Andrew = {...};

Note: the built-in object type is the set of all objects, whether it’s {}, {hello: "world"}, [], new Date(), etc. It’s only slightly more restrictive than any.

# Optional Properties

By default, all properties are treated as compulsory. To allow an object to not define a property, just postfix the property name with ?.

1
2
3
4
5
type Person = {
	firstName: string;
	lastName: string;	
	middleName?: string;   // Objects of type `Person` can optionally set the `middleName` property.
};

# Immutable Properties

You can make properties immutable by prefixing the property name with the readonly property. It’s like as if you declared a property as const, so once it’s assigned for the first time, it cannot be reassigned.

1
2
3
4
5
6
7
8
type Person = {
	name: string;
	readonly dateOfBirth: string;
};

const me: Person = { name: 'Tim', dateOfBirth: '09/03/2001' };
me.name = 'Andrew';             // This is fine.
me.dateOfBirth = '01/01/1970';  // Error.

# Index Signatures

When you want the flexibility for an object to contain more arbitrary properties with a certain key-value pair type, you can use the index signature syntax, [key: K]: V, where key can be any name you choose.

1
2
3
4
5
6
7
8
9
type FruitRatings = {
	[fruitName: string]: number;
};

const myFruitRatings: FruitRatings = {
	apple: 6,
	banana: 7,
	cherry: 9
};

# Indexed Access Types

See indexed access types. When you define a type for an object shape, if you want to access a nested part of that shape as a type, you can just use the subscript operator [].

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
interface Theme {
	colour: {
		mode: 'light' | 'dark';
		palette: {
			primary: ['#0000FF', '#0044FF', '#0066FF', '#3388FF'];
			secondary: ['#464646', '#616161', '#7E7E7E', '#AAAAAA'];
		};
	};
}

type ColourOptions = Theme['colour'];
type Palette = Theme['colour']['palette'];

Nope, you can’t use the . operator as if you were accessing object properties.

# keyof

The keyof unary operator evaluates to the union of a type’s keys.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
interface Theme {
	colour: {
		mode: 'light' | 'dark';
		palette: {
			primary: string[];
			secondary: string[];
		};
	};
}

type ColourProperties = keyof Theme['colour'];              // β†’ 'mode' | 'palette'
type PaletteProperties = keyof Theme['colour']['palette'];  // β†’ 'primary' | 'secondary'

This is helpful for writing getter functions that retrieve values nested in an object:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
const theme: Theme = {
	colour: {
		mode: 'dark',
		palette: {
			primary: ['#0000FF', '#0044FF', '#0066FF', '#3388FF'],
			secondary: ['#464646', '#616161', '#7E7E7E', '#AAAAAA'],
		},
	},
};

const getPalette = (theme: Theme, palette: keyof Theme['colour']['palette']): string[] => {
	return theme['colour']['palette'][palette];
};

console.log(getPalette(theme, 'primary'));

# Arrays

To denote a type as an array of items of type T, you can do T[] or Array<T> (they’re exactly the same).

1
2
3
4
// Array type declaration examples.
type Names = string[];
type People = { name: string; }[];
type Values = (string | number)[];

In general, try to keep arrays homogenous, that is, of a singular type, otherwise you’d have to do type-checking before TypeScript can let you use the items of the array.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
// When arrays are not homogenous, that is, of one type, you'd have to do some type checking to work with its items.
const arr: (string | number)[] = [42, "Hello"];

arr.forEach(item => {
	if (typeof item === 'number') {
		// You can use `item` as a number after you've type-checked it.
		const num = item * 3;
		console.log('Number: ', num);
	} else {
		// `item` must be a string
		const s = item + " world";
		console.log('String: ', s);
	}
})

Arrays are always mutable by default in TypeScript, that is, you can reassign the value at any index and invoke methods that make in-place modifications like push or reverse. To make them immutable, you would prefix them with the readonly modifier.

1
2
3
4
type Names = readonly string[];
const radioheadMembers: Names = ['Thom', 'Johnny', 'Colin', 'Phil', 'Ed'];

radioheadMembers[1] = 'Tim';  // Error.

# Tuples

Remember, tuples are just fixed-length arrays. In typescript, we define a tuple by specifying the type of each item like: [T1, T2, ...]:

1
2
3
4
type FullName = [string, string, string?];  // You can make items optional in a tuple, just like for objects.
                                            // Note: this is basically the same as: `[string, string] | [string, string, string]`.
const elon: FullName = ["Elon", "Reeve", "Musk"];
const jeff: FullName = ["Jeff", "Bezos"];

You can also make use of the rest operator, ..., to allow for tuples of arbitrary lengths.

1
2
type FullName = [string, string, ...string[]];
const queenElizabethII = ['Elizabeth', 'Alexandra', 'Mary', 'Windsor'];

Caveat: tuples are not immutable by default. Unlike in other languages like Python, the items of a tuple in TypeScript can be mutated, that is, reassigned after definition. To make tuples immutable, you would do the same thing as you would for making arrays immutable: prefix it with the readonly modifier.

1
2
3
4
type Coordinate = readonly [number, number];
const point: Coordinate = [1, 2];

point[0] = 3;   // Error.

# Enums

Enums are data types that have a fixed set of constant values. They’re a great way to group together a lot of related constants. Note: JavaScript doesn’t have enums.

1
2
3
4
5
6
7
8
9
enum Theme {
	Light,
	Dark,
	HighContrastLight,
	HighContrastDark
};

// You access the fields of the enum in the same way that you'd access an object.
const theme: Theme = Theme.Dark;

In the example above, every enum value gets implicitly assigned a counter value starting from 0. It’s equivalent to doing the following:

1
2
3
4
5
6
enum Theme {
	Light = 0,
	Dark = 1,
	HighContrastLight = 2,
	HighContrastDark = 3
}

You can also map enum keys to string values instead of integers.

1
2
3
4
5
6
enum Theme {
	Light = '#CCCCCC',
	Dark = '#333333',
	HighContrastLight = '#FFFFFF',
	HighContrastDark = '#000000',
}

# Caveats

One annoying issue is that you can freely assign numbers to enum types that are clearly outside the bounds of the enum values.

1
2
3
4
5
6
7
enum Theme {
	Light,
	Dark,
	HighContrastLight,
	HighContrastDark,
}
const theme: Theme = 6; // No complaints from TypeScript.

In general, the official docs advise you to avoid enums unless they help significantly with readability. Alternatives to enums include string literals, eg. type Theme = "Light" | "Dark" | ..., or object literals, eg. const Theme = { Light: "Light", Dark: "Dark", ... }.

# Type Inference

You don’t have to always supply an explicit type. Often, there’ll be enough context for TypeScript to figure it out without ambiguity. In general, we prefer type inference over explict assigning types to varaibles/parameters/etc. for conciseness.

1
2
let a: number = 42;   // There is no need to do this. It's clear what type `a` is from the RHS of the assignment.
let b = 42;           // Equivalent to above, but it lets typescript assign the type implicitly.

This extends to functions as well, meaning that often you won’t have to specify the return value.

# Type Widening/Narrowing

An important implicit rule in TypeScript is that when you let type inference happen for const variables, TypeScript will assign it the narrowest type possible since it knows that a const variable cannot possibly take any other value after its defined. Otherwise, TypeScript will infer the type to be wider than it might be.

1
2
let a = 2;    // `a` is of type `number`.
const b = 2;  // `b` is of type `2`, a specific member of `number`.

# typeof and instanceof

Although type checking is done for you statically, there are times when you must perform run-time type checks such as when you’re fetching external data. In these times, rely on JavaScript’s operators: typeof and instanceof.

  1. Use the instanceof binary operator to check some value is of a custom type, or a complex built-in type like RegExp.
    • Note that val instanceof T works by checking if T exists anywhere along val’s prototype chain. This is why you get unintuitive results when you use instanceof on primitive types. For example, 42 instanceof Number is false, but new Number(42) instanceof Number is true.
  2. Use the typeof unary operator to check some value is some built-in primitive type such as undefined, number, string, boolean, etc.

# Type Assertions

When you’re confident that some value should be a certain type but TypeScript isn’t, you can make a type assertion with the as keyword.

1
2
3
4
let someVal: any = 123;

// Here, you're basically telling TypeScript: "I am 100% sure this is a number. Trust me."
const val = someVal as number;

You can also make type assertions by prefixing an expression with <T>, eg. <Person>person which is exactly the same as person as Person.

Aim to minimise your usage of type assertions like above. They’re considered ’escape hatches’ from the language and can prevent you from maximising the benefits of using a type system.

# Const Assertion

Use as const to tell TypeScript to infer the value to its narrowest possible type.

1
2
const a = [1, 2, 3];           // β†’ Type: `number[]`
const b = [1, 2, 3] as const;  // β†’ Type: `readonly [1, 2, 3]`

# Nonnull Assertion

When you’re confident a value is not null, you can postfix that value with ! to assure TypeScript. If you cannot be confident, then just use a standard null-check: if (_ === null) ....

1
2
3
4
5
type NullableString = string | null;
const s: NullableString = 'Hello';

// This is basically saying: "Don't worry TypeScript, I'm 100% sure `s` is not null."
console.log(s!.toUpperCase());

# Refinement

TypeScript’s static analysis can handle refinement where, based on the control flow logic, TypeScript can narrow the type of the variable. Refinement can happen when you use if, the optional chaining operator ?., ||, switch, typeof, instanceof, in, etc.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
type CssWidth = number | string | undefined;

const getPixelWidth = (width: CssWidth): number => {
	// At this point, TypeScript knows `width` is `number | string | undefined`.
	if (typeof width === 'undefined') return 0;

	// At this point, TypeScript knows `width` is `number | string`.
	if (typeof width === 'number') return width;

	// At this point, TypeScript knows `width` is `string`. We can therefore use
	// string methods on `width` with confidence.
	return Number(width.slice(0, width.search('px')));
};

console.log(getPixelWidth(undefined)); // 0
console.log(getPixelWidth(10));        // 10
console.log(getPixelWidth('480px'));   // 480

Refinement works with unioned objects, but it’s best to use unique strings to help TypeScript infer types properly.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
interface UserTextEvent {
	type: 'TextEvent';
	value: string;
	target: HTMLInputElement;
}
interface UserMouseEvent {
	type: 'MouseEvent';
	value: [number, number];
	target: HTMLElement;
}

type UserEvent = UserTextEvent | UserMouseEvent;

const handle = (event: UserEvent): void => {
	if (event.type === 'TextEvent') {
		// At this point, TypeScript is certain that `event` is `UserTextEvent`.
		// ...
	} else {
		// At this point, TypeScript is certain that `event` is `UserMouseEvent`.
		// ...
	}
};

This kind of type refinement is very useful when working with [[Knowledge/Engineering/Technologies/Redux#Reducers|Redux reducers]].

# Type Guards

Refinement doesn’t work as expected when you use a function to do the type-checking. Any type-checking only contributes to refinement if it’s in the same scope.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
const isString = (s: unknown): boolean => {
	return typeof s === 'string';
}

const refinementTest = (val: string | number) => {
	if (isString(val)) {
		val.toLowerCase();   // Error. TypeScript still thinks `val` is `string | number`.
		// ...
	} else {
		const num = val * 2; // Error. TypeScript still thinks `val` is `string | number`.
		// ...
	}
}

To fix this, you’d need to define a type guard which is a predicate function that confirms an argument is a given type. It looks like this:

1
2
3
const isString = (s: unknown): s is string => {
	return typeof s === 'string';
};

# Variance

It’s useful to think of types as just sets. When $A$ is a subtype of $B$, it is basically just saying that $A \subseteq B$.

Variance, in programming language theory, is how subtyping works for generic types. It is concerned about the idea of whether a generic type like List<Cat> is a subtype of List<Animal>.

There are 4 kinds of variance:

Every language’s type system has different rules around variance. As a programming language designer, if you were to allow covariance or contravariance over invariance, then you’re allowing for greater flexibility in the type system, but it exposes programmers to greater risk of runtime type errors.

TypeScript tends to be more relaxed by allowing functions to take in covariant arguments. For example, you can pass an argument so long as it is a subtype of the expected parameter, ie. covariant to the expected parameter, but this makes it possible to create run-time type errors like this:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
interface EngineeringStudent {
	name: string;
	discipline: string;
}
interface FirstYearEngineeringStudent {
	name: string;
	discipline?: string;   // This is basically: `string | undefined`.
}                          // This makes `EngineeringStudent` a subtype of `FirstYearEngineeringStudent`!

// Here, `student` can be `FirstYearEngineeringStudent` or any subtype of it.
const clearDiscipline = (student: FirstYearEngineeringStudent) => {
	delete student.discipline;
};

// The dangers of accepting a covariant argument:
// We can delete the non-optional `discipline` field and TypeScript will not complain.
const csStudent: EngineeringStudent = { name: 'Linus', discipline: 'Computer Science' };
clearDiscipline(csStudent);

console.log(csStudent.discipline);  // β†’ undefined

# Functions

Typing Function Declarations To type a function declaration, you just assign types for each parameter it takes in and then specify the return type by postfixing the parameter list with a colon and a type.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
// Typing function parameters and return values.
// To assign a return type to a function, you postfix the parameter list with a colon and a type.

// Regular functions:
function increment(num: number): number {
	return num + 1;
}

// Arrow functions:
const decrement = (num: number): number => num - 1;

Typing Function Expressions or Arrow Functions What if you want to specify the type of a callback rather than a function declaration? You would use the syntax: (param: Type, ...) => RetType. Although the syntax is inspired by arrow functions, it is not actually defining a function.

1
2
3
4
5
6
7
8
type Greeting = (a: string, b: string) => void;

// Note that the parameter names above don't need to match that of the assigned
// callback's parameter names, they're purely for documentation.
const callback: Greeting = (name, message) => {
	console.log(`Hi, I'm ${name}. ${message}`);
}
callback("Linus", "F*** you, Nvidia.");

# Optional & Default Parameters

Just like how you can make object properties optional and tuple items optional, you can make function parameters optional by postfixing the parameter name with a ?. Alternatively, you can set a default value for a parameter by assigning a value directly after the parameter name, which is pretty much the same as making it optional.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
// Optional parameter.
const greet = (name: string, message?: string) => {
	console.log(`Hi ${name}!`);
	if (message) console.log(message);
}

// Default parameter. Notice that the parameter type can often be inferred from the
// default value that you supply.
// You can also choose to explicitly set the type anyway like: 
//   `message: string = "You rock."`
const greet = (name: string, message = "You rock.") => {
	console.log(`Hi ${name}! ${message}`);
}

# Variadic Functions

A variadic function is just one that takes in an arbitrary number of arguments. The vast majority of functions take in a fixed list of parameters, we call these ‘fixed-arity’ functions. Normally in JavaScript, defining a variadic function requires you to make use of the implicit arguments array in the function body. In TypeScript, it’s way more intuitive and can be done in a type-safe way with the rest operator, ....

1
2
3
4
const max = (...nums: number[]): number => {
	if (!nums || nums.length === 0) return -Infinity;
	return nums.reduce((maxSoFar, currNum) => (maxSoFar > currNum) ? maxSoFar : currNum, -Infinity);
}

# this

In JavaScript, this is a nightmare for most programmers to work with because its value is different depending on how it is called. In TypeScript, you can minimise surprises around the value of this by assigning a type to it as the first function parameter.

To illustrate the problem:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
type Person = { name: string, greet: () => void };
const person: Person = {
	name: "Linus Torvalds",
	greet: function() {
		console.log(`Hi, I'm ${this.name}`);
	}
}
person.greet(); // This works as expected since `this` is bound to `person`.

const greet = person.greet;
greet();        // This fails since the `this` is lost and is no longer bound to `person`.

The ‘solution’ is to assign a type for this so that the developer is warned when this takes on the wrong type when they invoke a function that uses it.

1
2
3
4
5
6
7
type Person = { name: string, greet: (this: Person) => void };
const person: Person = {
	name: "Linus Torvalds",
	greet: function(this: Person) {
		console.log(`Hi, I'm ${this.name}`);
	}
}

# Generators

See generators. In JavaScript, you can create a generator function by postfixing function with an asterisk, *. Note: you cannot define arrow functions as generator functions (at least as of 2022’s ES standard).

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
function* fooGenerator() {
	yield 42;
	yield 24;
}

const fooNums = fooGenerator();
console.log(fooNums.next());    // β†’ { value: 42, done: false }
console.log(fooNums.next());    // β†’ { value: 24, done: false }
console.log(fooNums.next());    // β†’ { value: undefined, done: false }

// You can loop through a generator's values with JavaScript's for-of loops.
for (const item of fooGenerator()) {
    console.log(item);
}

TypeScript automatically infers the return type of the generator to be IterableIterator<number>. To assign an explicit type for what gets yielded, do it the same way that you’d specify the return value, but wrap it around with IterableIterator.

1
2
3
function* fooGenerator(): IterableIterator<number> {
	...
}

# Iterators

See iterators. In JavaScript, an iterable is an object containing the Symbol.iterator property with the value being a function that returns an iterator (which can be done by defining Symbol.iterator to be a generator function, which always returns an iterator). An iterator is an object that defines a next method which returns an object of shape: { value: any, done: boolean }.

An object can be both an iterable and an iterator at the same time. When you invoke a generator function, for example, you get an object of type IterableIterator which is both, meaning it has a Symbol.iterator property, whose value is a function that returns an iterator, and the next method.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
const favouriteNums = {
	*[Symbol.iterator]() {
		yield 42;
		yield 2;
		yield 4;
	}
}

for (const item of favouriteNums) {
	console.log(item);
}

Note: the syntax for defining Symbol.iterator as a generator function might seem strange. See this post for clarifications. As for the square brackets around Symbol.iterator, it’s called the computed property name syntax, introduced in ES6.

# Function Overloading

You can define a function type that actually consists of multiple function signatures. See the Function Overloads.

# Generic Functions

See generics. In TypeScript, you can define generic functions by specifying a comma-separated list of generic type parameters in angle brackets <> right before the parameter list of function. You would use generic functions if you wanted a function to be reusable across multiple types without giving up type safety by resorting to any.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
type Filter = <T>(array: T[], predicate: (elem: T) => boolean) => T[];

const filter: Filter = <T>(array: T[], predicate: (elem: T) => boolean) => {
	const arr: T[] = [];
	array.forEach((elem) => {
		if (predicate(elem)) arr.push(elem);
	});
	return arr;
};

// TypeScript can infer that `T` should be `number`.
console.log(filter([1, 2, 3, 4, 5], (num) => num % 2 === 0));

// To explicitly set `T`, use angle brackets after the function name.
console.log(filter<number>([1, 2, 3, 4, 5], (num) => num % 2 === 0));

To make regular function declarations generic, you also place the generic type parameters between angle brackets right before the parameter list:

1
function filter<T>(array: T[], predicate: (elem: T) => boolean) => { ... }

# Bind on Reference

In the above example, T gets bound when the function gets invoked, but you could also bind T whenever the type alias gets referenced by placing the generic type parameters after the type name instead of before the function parameter list:

1
2
3
4
type Filter<T> = (array: T[], predicate: (elem: T) => boolean) => T[];

// Wherever you use `Filter`, you have to explicitly bind `T` like `Filter<T>`.
const filter: Filter<number> = (array, predicate) => ...;

# Bounded Polymorphism

Sometimes, saying that a generic function takes in type parameter of T is too permissive. Instead, we might want T to be a subtype of U, that is, we should accept type parameters that are ‘at least’ U. This is called bounded polymorphism or constrained genericity.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
type Enemy = { health: number };
type Alien = Enemy & { galaxy: string };
type Cyborg = Enemy & { model: string };

type AttackEnemy = <T extends Enemy>(enemy: T, damage: number) => void;
const attackEnemy: AttackEnemy = <T extends Enemy>(enemy: T, damage: number) => {
	enemy.health -= damage;
	console.log(`Dealt ${damage}. Enemy now has ${enemy.health} HP left.`);
};

const enemy: Enemy = { health: 20 };
const alienEnemy: Alien = { health: 50, galaxy: 'Andromed' };
const cyborgEnemy: Cyborg = { health: 100, model: 'Terminator Mk. II' };

attackEnemy(enemy, 15);
attackEnemy(alienEnemy, 10);
attackEnemy(cyborgEnemy, 8);
attackEnemy("Hello world", 5);  // Fails because "Hello world" is not a subtype of `Enemy`.

# Object-Oriented Programming

See Object Oriented Programming.

# Access Modifiers

See encapsulation. TypeScript offers 3 access modifiers, which can be prefixed to any class field declaration:

When prefixing a constructor’s parameter with an access modifier, it’ll declare the field and assign the given value implicitly.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
class Person {
	constructor(public name: string) {}
}
// ... is a shorthand that's equivalent to:
class Person {
	public name: string;
	constructor(name: string) { this.name = name; }
}

const person: Person = new Person('Linus Torvalds');
console.log(person.name);

# Inheritance

See inheritance. In TypeScript, inheritance works in the same way and uses the same syntax as JavaScript’s inheritance.

# Method Overriding

See method overriding. By default, every method is ‘virtual’, so you can override them all. To override a method in TypeScript, just copy the method signature and supply the new method body. As good practice, use the optional override modifier so that you’re warned when you’ve got the base class’ method signature wrong.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
class Base {
    // Methods are virtual by default.
	public foo(): void {
		console.log('Foo');
	}
}

class Sub extends Base {
    // Explicitly re-implementing the parent's `foo` method.
	public override foo(): void {
		console.log('Bar');
	}
}

# Abstract Classes

See abstract classes. To make a class abstract, just prefix it with the abstract keyword.

1
2
abstract class Employee { ... }
class SoftwareEngineer extends Employee { ... }

# Abstract Methods

See abstract methods. Abstract methods must be inside abstract classes. To make a method abstract, use the abstract modifier, explicitly type the method signature and do not provide a body.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
abstract class Employee {
	constructor(public salary: number) {}
	public getSalary(): number { return this.salary; }
	public abstract slackOff(): void;
}

class SoftwareEngineer extends Employee {
	constructor() { super(100000); }
	public override slackOff() { console.log('Time to browse r/ProgrammerHumor...'); }
}

const linus: Employee = new SoftwareEngineer();
linus.slackOff();

# Generic Types in Classes/Interfaces

You can set class-scoped or interface-scoped generic type parameters:

1
2
class HashMap<K, V> { ... }
interface HashMap<K, V> { ... }

# Modules

See JavaScript modules. With TypeScript, you can additionally import/export type aliases and interfaces.

Note: in import statements, you don’t need to specify the .ts file extension. This means you can easily import type declaration files with the extensionless name.

In thing.ts:

1
2
3
4
5
6
7
8
9
// Notice that this file exports a value `Thing` and a type `Thing`, but
// no name collision happens because 'values' and 'types' are tracked in
// separate namespaces by the TypeScript compiler.
export type Thing = {
	val: number;
};
export const Thing = {
	val: 42,
};

In main.ts:

1
2
3
4
5
// Notice that you don't need to write the extension in the path: './thing.ts'.
import { Thing } from './thing';

const thing: Thing = Thing;
console.log(thing);

# Error Handling

See JavaScript error handling. TypeScript doesn’t introduce any new syntax for error handling over JavaScript, but the type system allows for streamlining how errors are treated in a project by developers.

# Ways of Error Handling

There are 4 common patterns for handling errors in TypeScript projects, which are also mostly applicable to non-TypeScript projects:

  1. Just return null. This reveals the least information in the event of an error, but it’s the easiest to do. Constant null-checking is required throughout the code however, which can be laborious and verbose.
  2. Throw an exception. When an exception is thrown, it must be caught by the caller in a try-catch block (or a catch callback if using promises) otherwise a full crash occurs. Making and throwing custom subclasses of Error would offer specific information to help with debugging and informing the user about the problem. The main issue is that it’s hard to enforce that programmers write the error-handling try-catch logic when they’re lazy.
  3. Return exceptions (rather than throw them). This means a function will specify in its return type a union of the expected return type and the error classes that it could throw, such as in the following:
    1
    
     const getData = (): Data | NetworkError => {};
    
    By putting the error as part of the return type, the user of the function is unlikely to ignore error cases. The idea here is very similar in spirit to Java’s throws keyword. The downside to this approach is that it’ll lead to more verbose function signatures, especially if errors are simply ‘bubbled’ up the call stack.
  4. Define and use the Option type. The idea comes from languages like Rust and Haskell. See Rust’s documentation on std::option.

# Utility Types

TypeScript gives you a bunch of very useful built-in utility types that you can use to make working with complex types a breeze 🌬️.

# Mapping Types

Here are some of the most useful utility types for sourcing types from other types:

Usage examples:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
interface Human {
    limbs: string[];
    organs: string[];
    memories: string[];
    soul: boolean;
}

type SubHuman = Partial<Human>;                                 // Human, but all properties are optional.
type Husk = Omit<Human, 'soul' | 'memories'>;                   // Human, but without the soul or memories.
type SentimentalProperties = Pick<Human, 'soul' | 'memories'>;  // Only the soul and memories of a human.
type FrozenHuman = Readonly<Human>;                             // Human, but every property is immutable.

// After experiencing Java programming, I am just a husk ;(
const me: Husk = {
    limbs: ["arms", "legs", "..."],
    organs: ["half a brain", "heart", '...'],
};

Note: behind the scenes, utility types such as the ones above are realised through ‘mapped types’.

1
2
3
4
5
// This is the `Partial` type, implemented using mapped types.
// Many other utility types are implemented in a very similar manner!
type MyPartial<T> = {
    [K in keyof T]?: T[K];
};

# Conditional Types

Here are some of the most useful utility types that leverage conditional typing, a TypeScript innovation.

Note: just like how you can use the ternary operator, (condition) ? expr1 : expr2 for conditional expression evaluation, you can use the ternary operator for conditional type evalution. This is what’s used to implement those conditional utility types above.

# Asynchronous Programming

See JavaScript asynchronous programming.

# JavaScript Interoperability

An excellent reason to adopt TypeScript is that you don’t have to rewrite your JavaScript codebase to begin benefiting from a type system.

# Type Declaration Files

A type declaration file, which goes with the extension .d.ts, associates types to JavaScript code. It’s a file consisting only of type-level code, meaning you can’t use expressions in there (which means no function implementations, variables, class implementations, etc. can be defined within). As a very loose analogy, .d.ts files are kind of like the .h header files in C or C++.

If you have a hello-world.js file, then the type declaration file must have the name, hello-world.d.ts.

A type declaration is a way to tell TypeScript, β€œThere exists this thing that’s defined in JavaScript, and I’m going to describe it to you.” (Programming TypeScript).

NPM packages that once were intended only for JavaScript developers (eg. jQuery) can be made consumable by TypeScript developers by having these type declaration files. As a TypeScript dev, you’d be able to use pure JS libraries as if they were written in TypeScript.

When type declarations don’t ship with an NPM package, they can usually be install individually in the @types organisation on npm. The type declarations in DefinitelyTyped, a big community effort to bring types to popular JS libraries, are automatically published to the @types organisation.

Eg. to bring jQuery into a TypeScript frontend project, you’d do:

1
2
npm install jquery --save              
npm install @types/jquery --save-dev  # Brings in all the type declaration files.