Install

npm install --save-dev ts-essentials

👉 We require typescript>=3.7. If you're looking for support for older TS versions use ts-essentials@3 (for 3.6>=) or ts-essentials@2 instead. If you use any functions you should add ts-essentials to your dependencies (npm install --save ts-essentials) to avoid runtime errors in production.

What's inside?

ts-essentials is a set of high-quality, useful TypeScript types that make writing type-safe code easier.

• Install
• What's inside?
  • Basic
  • Dictionaries
  • Deep* wrapper types
    • DeepPartial
    • DeepRequired
    • DeepReadonly
    • DeepNonNullable
    • DeepNullable
    • DeepUndefinable
  • Writable & DeepWritable
  • Buildable
  • Omit
  • StrictOmit
    • Comparison between Omit and StrictOmit
  • DeepOmit
  • OmitProperties
  • PickProperties
  • NonNever
  • Merge
  • MarkRequired
  • MarkOptional
  • ReadonlyKeys
  • WritableKeys
  • OptionalKeys
  • RequiredKeys
  • PickKeys
  • UnionToIntersection
  • Opaque types
  • Tuple constraint
  • Exhaustive switch cases
  • ValueOf type
  • ElementOf type
  • AsyncOrSync type
  • Awaited type
  • Newable
  • Assertions
  • Exact
  • XOR
  • Functional type essentials
    • Head
    • Tail
• Contributors

Basic

• Primitive type matching all primitive values.
• noop function that takes any arguments and returns nothing, as a placeholder for e.g. callbacks.

Dictionaries

keywords: map

const stringDict: Dictionary<string> = {
  a: "A",
  b: "B",
};

// Specify second type argument to change dictionary keys type
const dictOfNumbers: Dictionary<string, number> = {
  420: "four twenty",
  1337: "HAX",
};

// You may specify union types as key to cover all possible cases. It acts the same as Record from TS's standard library
export type DummyOptions = "open" | "closed" | "unknown";
const dictFromUnionType: Dictionary<number, DummyOptions> = {
  closed: 1,
  open: 2,
  unknown: 3,
};

// and get dictionary values
type stringDictValues = DictionaryValues<typeof stringDict>;
// Result: string

// When building a map using JS objects consider using SafeDictionary
const safeDict: SafeDictionary<number> = {};
const value: number | undefined = safeDict["foo"];

// With SafeDictionary you don't need to use all of the sub-types of a finite type.
// If you care about the key exhaustiveness, use a regular Dictionary.
type ConfigKeys = "LOGLEVEL" | "PORT" | "DEBUG";
const configSafeDict: SafeDictionary<number, ConfigKeys> = {
  LOGLEVEL: 2,
};
const maybePort: number | undefined = configSafeDict["PORT"];

const configDict: Dictionary<number, ConfigKeys> = {
  LOGLEVEL: 2,
  PORT: 8080,
  DEBUG: 1,
};
const port: number = configDict["PORT"];

Deep* wrapper types

• DeepPartial
• DeepRequired
• DeepReadonly
• DeepNonNullable
• DeepNullable
• DeepUndefinable

keywords: recursive, nested, optional

type ComplexObject = {
  simple: number;
  nested: {
    a: string;
    array: [{ bar: number }];
  };
};

type ComplexObjectPartial = DeepPartial<ComplexObject>;
const samplePartial: ComplexObjectPartial = {
  nested: {
    array: [{}],
  },
};

type ComplexObjectAgain = DeepRequired<ComplexObjectPartial>;
const sampleRequired: ComplexObjectAgain = {
  simple: 5,
  nested: {
    a: "test",
    array: [{ bar: 1 }],
  },
};

type ComplexObjectReadonly = DeepReadonly<ComplexObject>;

type ComplexNullableObject = {
  simple: number | null | undefined;
  nested: {
    a: string | null | undefined;
    array: [{ bar: number | null | undefined }] | null | undefined;
  };
};

type ComplexObjectNonNullable = DeepNonNullable<ComplexNullableObject>;
const sampleNonNullable: ComplexObjectNonNullable = {
  simple: 5,
  nested: {
    a: "test",
    array: [{ bar: null }], // Error: Type 'null' is not assignable to type 'number'
  },
};

type ComplexObjectNullable = DeepNullable<ComplexObject>;
const sampleDeepNullable1: ComplexObjectNullable = {
  simple: null,
  nested: {
    a: null,
    array: [{ bar: null }],
  },
};
const sampleDeepNullable2: ComplexObjectNullable = {
  simple: 1,
  nested: {
    array: [null], // OK
    // error -- property `a` missing, should be `number | null`
  },
};

// DeepUndefinable will come in handy if:
//  - you want to explicitly assign values to all of the properties
//  AND
//  - the expression used for the assignment can return an `undefined` value
// In most situations DeepPartial will suffice.
declare function tryGet(name: string): string | undefined;
type ComplexObjectUndefinable = DeepUndefinable<ComplexObject>;
const sampleDeepUndefinable1: ComplexObjectUndefinable = {
  simple: undefined,
  nested: {
    a: tryGet("a-value"),
    array: [{ bar: tryGet("bar-value") }],
  },
};
const sampleDeepUndefinable2: ComplexObjectUndefinable = {
  // error -- property `simple` missing, should be `number | undefined`
  nested: {
    array: [[{ bar: undefined }]],
    // error -- property `a` missing, should be `string | undefined`
  },
};

Writable

Make all attributes of object writable.

type Foo = {
  readonly a: number;
  readonly b: string;
};

const foo: Foo = { a: 1, b: "b" };
(foo as Writable<typeof foo>).a = 42;

type Foo = {
  readonly foo: string;
  bar: {
    readonly x: number;
  };
}[];

const test: DeepWritable<Foo> = [
  {
    foo: "a",
    bar: {
      x: 5,
    },
  },
];

// we can freely write to this object
test[0].foo = "b";
test[0].bar.x = 2;

Buildable

keywords: builder

A combination of both DeepWritable and DeepPartial. This type allows building an object step-by-step by assigning values to its attributes in multiple statements.

interface ReadonlyObject
  extends Readonly<{
    simple: number;
    nested: Readonly<{
      a: string;
      array: ReadonlyArray<Readonly<{ bar: number }>>;
    }>;
  }> {}

const buildable: Buildable<ReadonlyObject> = {};
buildable.simple = 7;
buildable.nested = {};
buildable.nested.a = "test";
buildable.nested.array = [];
buildable.nested.array.push({ bar: 1 });
const finished = buildable as ReadonlyObject;

Omit

Our version of Omit is renamed to StrictOmit in v3, since the builtin Omit has become part of TypeScript 3.5

StrictOmit

Usage is similar to the builtin version, but checks the filter type more strictly.

type ComplexObject = {
  simple: number;
  nested: {
    a: string;
    array: [{ bar: number }];
  };
};

type SimplifiedComplexObject = StrictOmit<ComplexObject, "nested">;

// Result:
// {
//  simple: number
// }

// if you want to Omit multiple properties just use union type:
type SimplifiedComplexObject = StrictOmit<ComplexObject, "nested" | "simple">;

// Result:
// { } (empty type)

Comparison between Omit and StrictOmit

Following the code above, we can compare the behavior of Omit and StrictOmit.

type SimplifiedComplexObjectWithStrictOmit = StrictOmit<ComplexObject, "nested" | "simple" | "nonexistent">;

// Result: error
// Type '"simple" | "nested" | "nonexistent"' does not satisfy the constraint '"simple" | "nested"'.
// Type '"nonexistent"' is not assignable to type '"simple" | "nested"'.

type SimplifiedComplexObjectWithOmit = Omit<ComplexObject, "nested" | "simple" | "nonexistent">;

// Result: no error

As is shown in the example, StrictOmit ensures that no extra key is specified in the filter.

DeepOmit

Recursively omit deep properties according to key names.

Here is the Teacher interface.

interface Teacher {
  name: string;
  gender: string;
  students: { name: string; score: number }[];
}

Now suppose you want to omit gender property of Teacher, and score property of students. You can achieve this with a simple type filter.

In the filter, the properties to be omitted completely should be defined as never. For the properties you want to partially omit, you should recursively define the sub-properties to be omitted.

type TeacherSimple = DeepOmit<
  Teacher,
  {
    gender: never;
    students: {
      score: never;
    };
  }
>;

// The result will be:
// {
//  name: string,
//  students: {name: string}[]
// }

NOTE

• DeepOmit works fine with Arrays and Sets. When applied to a Map, the filter is only applied to its value.
• If there exists any property in the filter which is not in the original type, an error will occur.

OmitProperties

keywords: filter, props

Removes all properties extending type P in type T. NOTE: it works opposite to filtering.

interface Example {
  log(): void;
  version: string;
}

type ExampleWithoutMethods = OmitProperties<Example, Function>;

// Result:
// {
//   version: string;
// }

// if you want to Omit multiple properties just use union type like

type ExampleWithoutMethods = OmitProperties<Example, Function | string>;
// Result:
// { } (empty type)

PickProperties

Pick only properties extending type P in type T.

interface Example {
  log(): void;
  version: string;
  versionNumber: number;
}

type ExampleOnlyMethods = PickProperties<Example, Function>;

// Result:
// {
//   log(): void;
// }

// if you want to pick multiple properties just use union type like

type ExampleOnlyMethodsAndString = PickProperties<Example, Function | string>;
// Result:
// {
//   log(): void;
//   version: string;
// }

NonNever

Useful for purifying object types. It improves intellisense but also allows for extracting keys satisfying a conditional type.

type GetDefined<TypesMap extends { [key: string]: any }> = keyof NonNever<
  { [T in keyof TypesMap]: TypesMap[T] extends undefined ? never : TypesMap[T] }
>;

NonEmptyObject

Useful for accepting only objects with keys, great after a filter like OmitProperties or PickProperties.

/* return never if the object doesn't have any number value*/
type NumberDictionary<T> = NonEmptyObject<PickProperties<T, number>>;

// return { a: number }
type SomeObject = NumberDictionary<{ a: number; b: string }>;

// return never
type EmptyObject = NumberDictionary<{}>;

Merge

keywords: override

type Foo = {
  a: number;
  b: string;
};

type Bar = {
  b: number;
};

const xyz: Merge<Foo, Bar> = { a: 4, b: 2 };
// Result:
// {
//   a: number,
//   b: number,
// }

MarkRequired

Useful when you're sure some optional properties will be set. A real life example: when selecting an object with its related entities from an ORM.

class User {
  id: number;
  posts?: Post[];
  photos?: Photo[];
}
type UserWithPosts = MarkRequired<User, "posts">;

// example usage with a TypeORM repository -- `posts` are now required, `photos` are still optional
async function getUserWithPosts(id: number): Promise<UserWithPosts> {
  return userRepo.findOneOrFail({ id }, { relations: ["posts"] }) as Promise<UserWithPosts>;
}

MarkOptional

Useful when you want to make some properties optional without creating a separate type.

interface User {
  id: number;
  name: string;
  email: string;
  password: string;
}

type UserWithoutPassword = MarkOptional<User, "password">;

// Result:

// {
//   id: number;
//   name: string;
//   email: string;
//   password?: string;
// }

ReadonlyKeys

Gets keys of an object which are readonly.

type T = {
  readonly a: number;
  b: string;
};
type Result = ReadonlyKeys<T>;
// Result:
// "a"

WritableKeys

Gets keys of an object which are writable.

type T = {
  readonly a: number;
  b: string;
};
type Result = WritableKeys<T>;
// Result:
// "b"

OptionalKeys

Gets keys of an object which are optional.

type T = {
  a: number;
  b?: string;
  c: string | undefined;
  d?: string;
};
type Result = OptionalKeys<T>;
// Result:
// "b" | "d"

RequiredKeys

Gets keys of an object which are required.

type T = {
  a: number;
  b?: string;
  c: string | undefined;
  d?: string;
};
type Result = RequiredKeys<T>;
// Result:
// "a" | "c"

PickKeys

Gets keys of properties of given type in object type.

type T = {
  a: number;
  b?: string;
  c: string | undefined;
  d: string;
};
type Result1 = PickKeys<T, string>;
// Result1:
// "d"
type Result2 = PickKeys<T, string | undefined>;
// Result2:
// "b" | "c" | "d"

UnionToIntersection

Useful for converting mapped types with function values to intersection type (so in this case - overloaded function).

type Foo = {
  bar: string;
  xyz: number;
};

type Fn = UnionToIntersection<{ [K in keyof Foo]: (type: K, arg: Foo[K]) => any }[keyof Foo]>;

Opaque types

Opaque types allow you to create unique type that can't be assigned to base type by accident. Good examples of opaque types include:

• JWTs or other tokens - these are special kinds of string used for authorization purposes. If your app uses multiple types of tokens each should be a separate opaque type to avoid confusion.
• specific currencies - amount of different currencies shouldn't be mixed
• bitcoin address - special kind of string

It's critical to understand that each token (second argument to Opaque) has to be unique across your codebase.

We encourage you to leverage a pattern where you have single function to validate base type and create opaque type.

type PositiveNumber = Opaque<number, "PositiveNumber">;
function makePositiveNumber(n: number): PositiveNumber {
  if (n <= 0) {
    throw new Error(`Value ${n} is not positive !`);
  }
  return (n as any) as PositiveNumber; // this ugly cast is required but only when "producing" opaque types
}

type NegativeNumber = Opaque<number, "NegativeNumber">;
function makeNegativeNumber(n: number): NegativeNumber {
  if (n >= 0) {
    throw new Error(`Value ${n} is not negative !`);
  }
  return (n as any) as NegativeNumber; // this ugly cast is required but only when "producing" opaque types
}

let a = makePositiveNumber(5); // runtime check
let b = makeNegativeNumber(-10); // runtime check

a = b; // error at compile time

Tuple constraint

function foo<T extends Tuple>(tuple: T): T {
  return tuple;
}

const ret = foo(["s", 1]);
// return type of [string, number]

You can also parametrize Tuple type with a type argument to constraint it to certain types, i.e. Tuple<string | number>.

Exhaustive switch cases

function actOnDummyOptions(options: DummyOptions): string {
  switch (options) {
    case "open":
      return "it's open!";
    case "closed":
      return "it's closed";
    case "unknown":
      return "i have no idea";
    default:
      // if you would add another option to DummyOptions, you'll get error here!
      throw new UnreachableCaseError(options);
  }
}

ValueOf type

const obj = {
  id: "123e4567-e89b-12d3-a456-426655440000",
  name: "Test object",
  timestamp: 1548768231486,
};

type objKeys = ValueOf<typeof obj>;
// Result: string | number

ElementOf type

const array = [1, 2, true, false];
type arrayElement = ElementOf<typeof array>;
// Result: number | boolean

AsyncOrSync type

Useful as a return type in interfaces or abstract classes with missing implementation

interface CiProvider {
  getSHA(): AsyncOrSync<string>;
  // same as
  getSHA(): Promise<string> | string;
}

class Circle implements CiProvider {
  // implementation can use sync version
  getSHA() {
    return "abc";
  }
}

class Travis implements CiProvider {
  // implementation can use async version when needed
  async getSHA() {
    // do async call
    return "def";
  }
}

// to get original type use AsyncOrSyncType
AsyncOrSyncType<AsyncOrSync<number>> // return 'number'

Awaited type

Unwrap promised type:

Awaited<Promise<number>> // number

Newable

keywords: constructor, class

Type useful when working with classes (not their instances).

class TestCls {
  constructor(arg1: string) {}
}

const t1: Newable<any> = TestCls;

Assertions

keywords: invariant

Simple runtime assertion that narrows involved types using assertion functions.

Note: This function is not purely type level and leaves minimal runtime trace in generated code.

const something: string | undefined = "abc" as any;
assert(something, "Something has to be defined!");
// from now on `something` is string, if this wouldn't be a case, assert would throw

const anything = "abc" as any;
assert(anything instanceof String, "anything has to be a string!");
// from now on `anything` is string

Exact

keywords: same, equals, equality

Exact<TYPE, SHAPE> Checks if TYPE is exactly the same as SHAPE, if yes than TYPE is returned otherwise never.

type ABC = { a: number; b: number; c: number }
type BC = { b: number; c: number }
type C = { c: number }

Exact<ABC, C> // returns NEVER
Exact<C, C> // returns C

XOR

Gets the XOR (Exclusive-OR) type which could make 2 types exclude each other.

type A = { a: string };
type B = { a: number; b: boolean };
type C = { c: number };

let A_XOR_B: XOR<A, B>;
let A_XOR_C: XOR<A, C>;

// fail
A_XOR_B = { a: 0 };
A_XOR_B = { b: true };
A_XOR_B = { a: "", b: true };
A_XOR_C = { a: "", c: 0 }; // would be allowed with `A | C` type

// ok
A_XOR_B = { a: 0, b: true };
A_XOR_B = { a: "" };
A_XOR_C = { c: 0 };

Functional type essentials

Head & Tail: useful for functional programming, or as building blocks for more complex functional types.

function tail<T extends any[]>(array: T): Tail<T> {
  return array.slice(1) as Tail<T>;
}

type FirstParameter<FnT extends (...args: any) => any> = FnT extends (...args: infer ArgsT) => any
  ? Head<ArgsT>
  : never;

Contributors

Thanks goes to these wonderful people (emoji key):

This project follows the all-contributors specification. Contributions of any kind welcome! Read more