type-guards

This module allows you to write run-time validation in a clear way in a strongly-typed manner. In other words, it gives you a nice way to keep your code DRY by writing the validation function, and get the TypeScript types with it.

Installation

$ yarn add type-guards

Examples

import * as tg from 'type-guards'
const isUser = tg.isOfShape({
  name: tg.isString,
  age: tg.isNumber,
})
 
// we purposely mark it as "any" to imitate API response or user's input
// (anything generated at runtime which we cannot give a static type to)
const john: any = { name: 'John', age: 21 }
 
if (isUser(john)) {
  john.name // typesafe, this is a string
  john.age // typesafe, this is a number
  john.years // error
}

List of functions

is

Create a validator that asserts that the given argument is strictly equal (===) to something. Not very useful on its own.

isOneOf

Create a validator that asserts that at least one of the validators passed as arguments are passing.

const isAbc = isOneOf(is('a'), is('b'), is('c'))
isAbc('a') // => true
isAbc('b') // => true
isAbc('c') // => true
isAbc('d') // => false

isEnum

Create a validator that asserts that the given arguments is exactly one of the given arguments. For example, the validator form the previous example can be written in a more simple way.

const isAbc = isEnum('a', 'b', 'c')

If all the arguments are of the same type, it will be inferred; so the above example will assert for string. If you have an enum, list all its values somehow in an array, and use the spread operator to pass the values in.

enum Gender { Male = 'm', Female = 'f' }
const GENDERS = [ Gender.Male, Gender.Female ]
const isGender = isEnum(...GENDERS) // a guard for type Gender

isNull

A validator that asserts that the given argument is null. Short for is(null)

isUndefined

A validator that asserts that the given argument is undefined. Short for is(undefined).

isNullOrUndefined, isNullish

A validator that asserts that the given argument is null or undefined (like doing arg == null). Short for isOneOf(is(null), is(undefined)).

An alias with a shorten yet recognizable name is isNullish.

isNotNull, isNotUndefined, isNotNullOrUndefined, isNotNullish

The opposite of the previous three validators.

A common use-case is filtering an array to get rid of nullish values:

const array: Array<number | null | undefined> = [0, null, 1, undefined, 2]
const filtered = array.filter(tg.isNotNullish)
// type of `filtered` is `Array<number>`

Doesn't work perfectly with the else branch, but this is a less common use-case. Either way, help is appreciated in the SO thread if you know more about this.

isOfBasicType

Create a validator that asserts that the given argument is of a certain type. This is a wrapper around typeof checks and works with string, number, boolean, symbol, function and object.

isString, isNumber, isBoolean, isSymbol, isObject, isFunction.

Validators that assert that the given argument is of the correct type. Short for isOfBasicType('string'), isOfBasicType('number'), etc.

Instead of isObject, you probably need isShapeOf instead, which gives you more control over the type.

isInstanceOf

Create a validator that asserts that utilized the mechanism of instanceof keyword in JavaScript.

isArrayOf

Create a validator that asserts the given argument is an array, where each of its item is of a certain type. The type of the items is passed as the argument of isArrayOf.

const areNumbers = isArrayOf(isNumber)
areNumbers([1, 2, 3]) // => true
areNumbers(1) // => false
areNumbers([1, 2, '3']) // => false
areNumbers([1, 2, undefined]) // => false

Of course, feel free to combine validators.

const areKindaNumbers = isArrayOf(isOneOf(isNumber, isNullOrUndefined))
areNumbers([1, 2, 3]) // => true
areNumbers([1, 2, null, 4, undefined]) // => true

isOfShape

Create a validator that asserts that the given argument is an object, where each of the values of its keys correspond to the given shape. The shape is an object where the values are either new shapes or simple type checks. isOfShape allows objects that have extra keys. See isOfExactShape to exclude objects having extra keys not defined by the shape.

const isUser = isOfShape({ name: isString, age: isNumber })
isUser({name: 'John', age: 21}) // => true
isUser({name: 'John', years: 21}) // => false
isUser({name: 'John', age: 21, years: 21}) // => true
isUser({name: 'John'}) // => false
 
const isCompany = isOfShape({
  name: isString,
  users: isArrayOf(isUser),
})
isCompany({name: 'Untitled', users: [{name: 'John', age: 21}]) // => true

isOfExactShape

The same as isOfShape, except that it excludes objects that have extra keys not defined by the shape.

const isUser = isOfExactShape({ name: isString, age: isNumber })
isUser({name: 'John', age: 21}) // => true
isUser({name: 'John', years: 21}) // => false
isUser({name: 'John', age: 21, years: 21}) // => false
isUser({name: 'John'}) // => false

isTuple

Create a validator that asserts that passed argument is a tuple of certain elements.

const isNamePair = isTuple(isString, isString)
 
isNamePair(['Walter', 'Jessie']) // => true
isNamePair('Gustavo') // => false
isNamePair(['Hector']) // => false
isNamePair(['Walter', 'Jessie', 'Mike']) // => false

pick

Create a validator which utilizes an already created validator and picks only a part of it.

const fullUser = isOfShape({ name: isString, age: isNumber })
const partOfUser1 = pick(fullUser, 'name')
const partOfUser2 = isOfShape({ name: isString })
// the two consts above produce the same validator

omit

Create a validator which utilizes an already created validator and omits a part of it.

const fullUser = isOfShape({ name: isString, age: isNumber })
const partOfUser1 = omit(fullUser, 'age')
const partOfUser2 = isOfShape({ name: isString })
// the two consts above produce the same validator

partial

Create a validator which utilizes an already created validator but allows undefined for every value.

const fullUser = isOfShape({ name: isString, age: isNumber })
const partOfUser1 = partial(fullUser)
const partOfUser2 = isOfShape({ name: one(isUndefined, isString), age: isOneOf(isUndefined, isNumber) })
// the two consts above produce the same validator

Note, however:

const partOfUser = partial(fullUser)
partOfUser({ name: 'John', age: 21 }) // => true
partOfUser({ name: 'John' }) // => false (missing "age")
partOfUser({ name: 'John', age: undefined }) => true 

Currently working on making the second one return true as well.

Using the type only

If you do not need to use the validator, but only want the type information, you can do that as well.

type Number = FromGuard<typeof isNumber> // : number
 
const isUser = isOfShape({ name: isString, age: isNumber })
type User = FromGuard<typeof isUser> // : { name: string, age: number }

This means that your codebase can have validators "just in case", but if you never use them, it will not increase your bundle size. You could also set up your build pipeline in such way that the validators are run only in development mode.

Run-time assertion

You usually want to throw an exception at run-time in case the state of the application becomes unexpected. For example, you might have public foo?: string in the class, but at some place you're certain that foo must be defined. Instead of doing this.foo!, which is just a build-time assertion, you might want to perform a run-time assertion such as the following.

if (this.foo === undefined) {
  throw new Error(`Unexpected value "undefined".`) 
}

TypeScript will properly assert here that this.foo is Exclude<string | undefined, undefined> below the if block, which boils down to string.

However, this becomes quite annoying to write all the time. Hence, throwIf helper.

const foo = tg.throwIf(tg.isUndefined)(this.foo)

Or, create a reusable function. This is the recommended way.

const throwIfUndefined = tg.throwIf(tg.isUndefined, `Unexpected "undefined" value.`)
const foo = throwIfUndefined(this.foo, `"this.foo" should've been defined here. Something's wrong.`)