Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

Sign up
Appearance settings

niieani/typescript-vs-flowtype

Folders and files

NameName
Last commit message
Last commit date

Latest commit

History

78 Commits

Repository files navigation

TypeScript vs Flow

Both TypeScript and Flow are very similar products and they share most of their syntax with some important differences. In this document I've tried to compile the list of differences and similarities between Flowtype and TypeScript -- specifically the syntax, usage and usability.

Disclaimer

This document might be incomplete and/or contain mistakes and was last updated to describe TypeScript 3.7.0 and Flow 0.101.

I'm maintaining it in my spare time, so if you find mistakes, or learn about latest additions to either project, please help keep this repo up-to-date by contributing and editing this page.

Thanks!

Differences in usage and usability

Some of these differences are subjective (e.g. error readability), and I'd love to make this as scientific as possible — so please contribute to make it better. :)

TypeScript Flow
Leading Design Goal / North Star identify errors in programs through a balance between correctness and productivity enforce type soundness / safety
IDE integrations top-notch: language server, built-in refactorings, type and typedoc information on hover, snappy go-to-definition language server is a work in progress, some IDEs use the CLI and require saving the file to run the type-check, refactorings in alpha, only type information on hover, sketchy go-to-definition
type-checking speed (excluding transpilation) benchmarks needed benchmarks needed, in-depth description numberOfFiles · O( (LoCperFile + SizeOfTypesOfExports) ^ k )
autocomplete
  • both during declaration and usage
  • feels instantaneous
  • feels reliable
  • only for usage
  • feels sluggish (often a second or more of delay)
  • feels unreliable (sometimes does not show up at all)
expressiveness great (since TS @ 2.1) great
type safety very good (7 / 10) great (8 / 10)
specifying generic parameters during call-time (f<T>(x)) yes e.g. yes (since Flow 0.72)
specifying generic parameters for type definitions yes yes
typings for public libraries plenty of well maintained typings a handful of mostly incomplete typings
unique features
  • autocomplete for object construction
  • declarable this in functions (typing someFunction.bind())
  • large library of typings
  • more flexible type mapping via iteration
  • namespacing
  • variance
  • existential types * (deprecated since 0.72)
  • testing potential code-paths when types not declared for maximum inference
  • $Diff<A, B> type
type spread operator no (planned) shipped >=0.42
support for nullish coalescing proposal shipped > 3.7beta yes
support for decorators proposal yes, legacy proposal only parsing of legacy proposal, no type-checking
support for extending built-in types yes no
userland plugins basic, not effecting emitting yet (planned) no
programmatic hooking architecture prepared, work in progress work in progress
documentation and resources
  • very good docs
  • many books
  • videos
  • e-learning resources
  • incomplete, often vague docs
ease-of-understanding of errors good good in some, vague in other cases
transparency meeting notes, leadership reasoning and roadmap happens mostly publicly low transparency, roadmap developed behind closed doors
commercial support no no
nominal and structural typing structural with plans to support nominal mostly structural, nominal for classes and imported opaque type aliases
dynamic import types import('module-name') since 2.9 undocumented $Exports<'module-name'>
utility size (not emitted JavaScript) (latest version) typescript size flow-bin size

Differences in syntax

bounded polymorphism

Flow

function fooGood<T: { x: number }>(obj: T): T {
 console.log(Math.abs(obj.x));
 return obj;
}

TypeScript

function fooGood<T extends { x: number }>(obj: T): T {
 console.log(Math.abs(obj.x));
 return obj;
}

Reference

https://flow.org/blog/2015/03/12/Bounded-Polymorphism/

maybe & nullable type

Flow

let a: ?string
// equivalent to:
let a: string | null | void

TypeScript

let a: string | null | undefined

Optional parameters implicitly add undefined:

function f(x?: number) { }
// is semantically the same as:
function f(x: number | undefined) { }
// and also same as (the `| undefined` is redundant):
function f(x?: number | undefined) { }

Optional properties implicitly add undefined

class A {
 foo?: string;
}

type casting

Flow

(1 + 1 : number);

TypeScript

(1 + 1) as number;
// OR (old version, not recommended):
<number> (1 + 1);

mapping dynamic module names

Flow

.flowconfig

[options]
module.name_mapper='^\(.*\)\.css$' -> '<PROJECT_ROOT>/CSSModule.js.flow'

CSSModule.js.flow

// @flow
// CSS modules have a `className` export which is a string
declare export var className: string;

TypeScript

declare module "*.css" {
 export const className: string;
}

Reference

Exact/Partial Object Types

By default objects in Flow are not exact, i.e. they can contain more properties than declared, whereas in TypeScript they are always exact (must contain only declared properties). In future versions Flow plans to change this and make objects exact by default.

Flow

When using flow, { name: string } only means "an object with at least a name property".

type ExactUser = {| name: string, age: number |};
type User = { name: string, age: number };
type OptionalUser = $Shape<User>; // all properties become optional

TypeScript

TypeScript is more strict here, in that if you want to use a property which is not declared, you must explicitly say so by defining the indexed property. It is possible to use dotted syntax to access indexed properties since TypeScript 2.2. This is mostly a design decision as it forces you to write the typings upfront.

type ExactUser = { name: string, age: number };
type User = { name: string, age: number, [otherProperty: string]: any };
type OptionalUser = Partial<ExactUser>; // all properties become optional

Reference

Importing types

Flow

import type {UserID, User} from "./User.js";
// equivalent:
import {type UserID, type User} from "./User.js";

TypeScript

TypeScript does not treat Types in any special way when importing.

import {UserID, User} from "./User.js";

typeof

Works the same in both cases, however Flow has an additional syntax to directly import a typeof:

Flow

import typeof {jimiguitar as GuitarT} from "./User";
// OR
import {typeof jimiguitar} from "./User.js";
type GuitarT = jimiguitar;
// OR (below also works in TypeScript)
import {jimiguitar} from "./User.js";
type GuitarT = typeof jimiguitar;

TypeScript

import {jimiguitar} from "./User";
type GuitarT = typeof jimiguitar;

Restrictive type

When you don't know a type, commonly you would use any type. A restrictive type accepts anything, like any but in order to use that variable you must ensure values type by refining it.

Flow

mixed

function stringifyNum(num: number) {
 // Do stuff
}
function stringify(value: mixed) {
 if (typeof value === 'string') {
 return '' + value; // Works!
 }
 if (typeof value === 'number') {
 return stringifyNum(value); // Works!
 }
 return '';
}

Reference: https://flow.org/en/docs/types/mixed/

Typescript

unknown

function stringifyNum(num: number) {
 // Do stuff
}
function stringify(value: unknown) {
 if (typeof value === 'string') {
 return '' + value; // Works!
 }
 if (typeof value === 'number') {
 return stringifyNum(value); // Works!
 }
 return '';
}

Reference: https://github.com/Microsoft/TypeScript/wiki/What%27s-new-in-TypeScript#new-unknown-top-type

Accessing the type of a Class

Classes are typed, so you don't need to define an explicit type for them. If you want to reference the type, you can do it the following way:

Flow

class Test {};
type TestType = typeof Test;
const instance = new Test();
type TestTypeFromInstance = Class<typeof instance>;

TypeScript

class Test {};
type TestType = typeof Test;

Nominal typing

Flow treats classes as nominal types, whereas TypeScript treats them as structural types.

Flow

class Foo {};
class Bar {};
const foo: Foo = new Bar();
// Cannot assign `new Bar()` to `foo` because `Bar` [1] is incompatible with `Foo` [2].

TypeScript

class Foo {};
class Bar {};
const foo: Foo = new Bar();
// No errors!

You can work around this with tricks like the following (declare only works in TypeScript >=3.7.0):

class Foo {
 declare private __nominal: void;
};
class Bar {
 declare private __nominal: void;
};
const foo: Foo = new Bar();
// Type 'Bar' is not assignable to type 'Foo'.
// Types have separate declarations of a private property '__nominal'.(2322)

Keys/Props Of Type

Flow

var props = {
 foo: 1,
 bar: 'two',
 baz: 'three',
}
type PropsType = typeof props;
type KeysOfProps = $Enum<PropsType>;
function getProp<T>(key: KeysOfProps): T {
 return props[key]
}

TypeScript

var props = {
 foo: 1,
 bar: 'two',
 baz: 'three',
}
type PropsType = typeof props
type KeysOfProps = keyof PropsType;
function getProp<T>(key: KeysOfProps): T {
 return props[key]
}

Records

Flow

type $Record<T, U> = {[key: $Enum<T>]: U}
type SomeRecord = $Record<{ a: number }, string>

TypeScript

type SomeRecord = Record<{ a: number }, string>

Lookup Types

Flow

type A = {
 thing: string
}
// when the property is a string constant use $PropertyType (i.e. you know it when typing)
type lookedUpThing = $PropertyType<A, 'thing'>
// when you want the property to be dynamic use $ElementType (since Flow 0.49)
function getProperty<T : Object, Key : string>(obj: T, key: Key): $ElementType<T, Key> {
 return obj[key];
}

Reference:

TypeScript

Arguably, it's a bit easier to type both cases in TS, since they follow the same pattern.

type A = {
 thing: string
}
type lookedUpThing = A['thing']
// and...
function getProperty<T, K extends keyof T>(obj: T, key: K) {
 return obj[key]; // Inferred type is T[K]
}
function setProperty<T, K extends keyof T>(obj: T, key: K, value: T[K]) {
 obj[key] = value;
}

Reference:

Type-narrowing functions

These are functions that return a boolean, performing some logic to assert that a given input parameter is of a certain type.

The implementations differ between Flow and TypeScript:

In TypeScript, it ensures the mapping between: true and value is T, versus in the case of Flow, it ensures the value is "checked" against the logic within the body of the function (i.e. things like typeof, instanceof, value === undefined).

This means you cannot tell Flow that the tested parameter is of an arbitrary type, which closes the door to complex cases, e.g.:

  • reusing logic from a different function
  • library definitions, where there is no body at all (it is possible to specify the body in the declaration, however you are still limited by the type of assertions you may specify)

Flow

function isNil(value: mixed): boolean %checks {
 return value == null;
}
const thing = null;
if (!isNil(thing)) {
 const another = thing.something;
}

Reference:

Caveats

The current implementation in Flow is incomplete, which means you cannot yet use %checks in class methods.

Example showing the limitation in the respective playgrounds: TypeScript vs Flow

TypeScript

Type-narrowing functions are called type guard functions in TypeScript.

function isNil<T>(value: T | null): value is null {
 return value == null;
}
const thing: any = null;
if (!isNil(thing)) {
 const another = thing.something;
}

Getting the type of a function call return value

Flow

$Call utility type:

type Fn1 = <T>(T) => T;
type E = $Call<Fn1, number>;
declare var e: E; // E is number
(42: E); // OK

Reference: https://github.com/facebook/flow/commit/ac7d9ac68acc555973d495f0a3f1f97758eeedb4

TypeScript

ReturnType utility type:

type fn1<T> = (a: T) => T;
type E = ReturnType<fn1<number>>;
var e: E; // E is number

Mapped Types / Foreach Property

Flow

type InputType = { hello: string };
type MappedType = $ObjMap<InputType, ()=>number>;

Reference:

TypeScript

A bit more flexibility here, as you have access to each individual key name and can combine with Lookup types and even do simple transformations.

type InputType = { hello: string };
type MappedType = {
 [P in keyof InputType]: number;
};

Function and method overloading

Flow

It is possible to declare multiple signatures for the same method (also called: overloading). This feature is undocumented, and only available in type declarations (.js.flow files or module statements), not inline/alongside your code.

declare function add(x: string, y: string): string;
declare function add(x: number, y: number): number;
declare class Adder {
 add(x: string, y: string): string;
 add(x: number, y: number): number;
}

However, it's possible to create function overloads inline for functions outside of classes, by using additional declarations.

declare function add(x: string, y: string): string;
declare function add(x: number, y: number): number;
function add(x, y) {
 return x + y;
}
add(1, 1); // Ok
add("1", "1"); // Ok
add(1, "1"); // Error

It is also possible to create function overloads using callable property syntax, see the section Object callable property.

TypeScript

TypeScript supports both function and method overloading, in both: type definitions (.d.ts) and inline alongside code.

class Adder {
 add(x: string, y: string): string;
 add(x: number, y: number): number;
 add(x, y) {
 return x + y;
 }
}
function add(x: string, y: string): string;
function add(x: number, y: number): number;
function add(x, y) {
 return x + y;
}

Read-only Types

Flow

type A = {
 +b: string
}
let a: A = { b: 'something' }
a.b = 'something-else'; // ERROR

TypeScript

type A = {
 readonly b: string
}
let a: A = { b: 'something' }
a.b = 'something-else'; // ERROR

One caveat that makes TypeScript's readonly less safe is that the same non-readonly property in a type is compatible with a readonly property. This essentially means that you can pass an object with readonly properties to a function which expects non-readonly properties and TypeScript will not throw errors: example.

"Impossible flow" type

Flow

empty

function returnsImpossible() {
 throw new Error();
}
// type of returnsImpossible() is 'empty'

TypeScript

never

function returnsImpossible() {
 throw new Error();
}
// type of returnsImpossible() is 'never'

Difference types

Flow

type C = $Diff<{ a: string, b: number }, { a: string }>
// C is { b: number}

It only works properly as lower bound, i.e. you can assign something to it, but can't use it after that.

(source)

Flow also has $Rest<>, which represents the result of the JS object rest operator ({ ...rest }).

type Props = { name: string, age: number };
const props: Props = {name: 'Jon', age: 42};
const {age, ...otherProps} = props;
(otherProps: $Rest<Props, {|age: number|}>);
otherProps.age; // Error, since we removed it

Typescript

You can define your own filter type, but it does not have a helper type for that.

class A {
 a: string;
 b: number;
}
class B {
 a: string;
 c: boolean;
}
type Omit<T, U> = Pick<T, Exclude<keyof T, keyof U>>;
// 
type C = Omit<A, B>;
// C is { b: number }

However, Flow implementation is stricter in this case, as B have a property that A does not have, it would rise an error. In Typescript, however, they would be ignored.

Same syntax

Most of the syntax of Flow and TypeScript is the same. TypeScript is more expressive for certain use-cases (advanced mapped types with keysof, readonly properties), and Flow is more expressive for others (e.g. $Diff).

Object callable property

The basic syntax are the same, except Flow has special syntax for the internal call property slot.

Both can be used to annotate function statics.

Flow

You can use objects with callable properties as functions: Try Flow

type F = {
 (): string
};
const f: F = () => "hello";
const hello: string = f();

An overloaded function is a function with multiple call signatures. This is supported by Flow. And we list out the different syntaxes here: Try Flow

type F = {
 (): string,
 [[call]]: (number) => string,
 [[call]](string): string
}
const f: F = (x?: number | string) => {
 return x ? x.toString() : '';
}

Use call property to annotate function statics: Try Flow

type MemoizedFactorialType = {
 cache: {
 [number]: number,
 },
 [[call]](number): number,
}
const factorial: MemoizedFactorialType = n => {
 if (!factorial.cache) {
 factorial.cache = {}
 }
 if (factorial.cache[n] !== undefined) {
 return factorial.cache[n]
 }
 factorial.cache[n] = n === 0 ? 1 : n * factorial(n - 1)
 return factorial.cache[n]
}

Reference:

TypeScript

You can use objects with callable properties as functions: TypeScript Playground

type F = {
 (): string;
}
const foo: F = () => "hello";
const bar: string = foo();

An overloaded function is a function with multiple call signatures. This is also supported by TypeScript: TypeScript Playground

type F = {
 (): string,
 (x: number): string,
 (x: string): string
}
const f: F = (x?: number | string) => {
 return x ? x.toString() : '';
}

Use call property to annotate function statics: TypeScript Playground

type MemoizedFactorialType = {
 cache?: {
 [n: number]: number,
 },
 (n: number): number,
}
const factorial: MemoizedFactorialType = n => {
 if (!factorial.cache) {
 factorial.cache = {}
 }
 else if (factorial.cache[n] !== undefined) {
 return factorial.cache[n]
 }
 factorial.cache[n] = n === 0 ? 1 : n * factorial(n - 1)
 return factorial.cache[n]
}

Reference:

optional parameters

Flow and TypeScript

The syntax in either tool is the same - question mark: ? suffixing the parameter name:

function(a?: string) {}

call-time generic parameters

In TypeScript and Flow (since version 0.72) you may use specify the type of a generic when calling the generic function or the constructor.

const set = new Set<string>();

Or using a more complex behavior:

function makeTgenerator<T>() {
 return function(next: () => T) {
 const something = next();
 return something;
 }
}
const usage = makeTgenerator<string>()
// 'usage' is of type: (next: () => string) => string

Typing pure JS files (i.e. without transpilation)

Flow

Flow supports a comment-based syntax, by encapsulating type annotations in /* */-style comments:

const f = (x /*: number */, y /*: number */) /*: number */ => x + y

TypeScript

TypeScript can check types with JavaScript files annotated with JSDoc comments:

// JSDoc type syntax
/** @type {function(number, number): number} */
const f = (x, y) => x + y
// equivalent TypeScript type syntax
/** @type {(x: number, y: number) => number} */

JSDoc's overloaded function comment syntax is not supported:

/**
 * @param {string} input
 * @returns {string} result
 *//**
 * @param {number} input
 * @returns {string} result
 */
function notSupported(input) { /* omit */ }

However, we can express function overloading type in TypeScript's form in a tricky way:

/** @type {{
 (): void;
 (code: 0): void;
 (code: 1, msg: string): void
 }} */
const functionOverloads = (
 /** @type {0 | 1} */ code = 0,
 /** @type {string | undefined} */ msg = code === 0 ? undefined : ""
) => { /* omit */ }

However, it still lacks some features:

  1. There is no way to pass type parameter when invoking generic functions.
  2. TypeScript cannot parse conditional types in JSDoc comments correctly. #27424
  3. There is no equivalent form of as const assertion. #30445

TypeScript-only concepts

Declarable arbitrary this in functions (outside of objects)

function something(this: { hello: string }, firstArg: string) {
 return this.hello + firstArg;
}

Private and Public properties in classes

class SomeClass {
 constructor(public prop: string, private prop2: string) {
 // transpiles to:
 // this.prop = prop;
 // this.prop2 = prop2;
 }
 private prop3: string;
}

Add ! to signify we know an object is non-null.

// Compiled with --strictNullChecks
function validateEntity(e?: Entity) {
 // Throw exception if e is null or invalid entity
}
function processEntity(e?: Entity) {
 validateEntity(e);
 let s = e!.name; // Assert that e is non-null and access name
}

Conditional Typing

type XorY<T, U> = T extends U ? X : Y;

This alone, introduces new helper types, or types aliases.

type Exclude<T, U> = T extends U ? never : T;
/**
 * Extract from T those types that are assignable to U
 */
type Extract<T, U> = T extends U ? T : never;
/**
 * Exclude null and undefined from T
 */
type NonNullable<T> = T extends null | undefined ? never : T;
/**
 * Obtain the return type of a function type
 */
type ReturnType<T extends (...args: any[]) => any> =
 T extends (...args: any[]) => infer R ? R : any;

Mapped Type Modifiers

You can use + and - operators to modify mapped types.

type Mutable<T> = {
 -readonly [P in keyof T]: T[P]
}
interface Foo {
 readonly abc: number;
}
// 'abc' is no longer read-only.
type TotallyMutableFoo = Mutable<Foo>

Helper type modifiers

Required is a type mapper to make all properties of an object to be required.

Partial is a type mapper to make all properties of an object to be optional.

Readonly is a type mapper to make all properties of an object to be readonly.

Flow-only concepts

Inferred existential types

* as a type or a generic parameter signifies to the type-checker to infer the type if possible

Array<*>

However this type was deprecated in Flow 0.72.

TypeScript proposal

Variance

https://flow.org/en/docs/lang/variance/

function getLength(o: {+p: ?string}): number {
 return o.p ? o.p.length : 0;
}

TypeScript proposal

Bivariance is among the design decisions driving TypeScript.

Opaque Type Alias

https://flow.org/en/docs/types/opaque-types/

opaque type Alias = Type;
opaque type Alias: SuperType = Type; // with subtyping constrains

Within the same file the opaque type alias is defined, opaque type aliases behave exactly as type aliases.

Outside the defining file, i.e. when importing an opaque type alias, it behaves like a nominal type. If the opaque type alias is defined with subtyping constrains, it can be used as the super type when outside the defining file.

export opaque type Age: number = number;
function newAge(age: number): Age {
 return age; // ok within same file, not ok outside defining file
}
function incAge(age: Age): number {
 return age + 1; // ok
}

TypeScript dose not have opaque type, but we can define an utility type with intersection type to mimic the behavior of Flow's opaque type alias with subtyping constrains used outside the defining file.

type Opaque<T, U> = U & { readonly __TYPE__: T }
type Age = Opaque<'age', number>
function newAge(age: number): Age {
 return age; // not ok
}
function incAge(age: Age): number {
 return age + 1; // ok
}

Object type spread

Object type spread acts as object spread but for types. Unlike intersection types type spreads work with exact object types and overwrite existing properties.

type Foo = {| foo: string, bar: string |}
type Bar = {| bar: number |}
type FooBarIntersection = Foo & Bar
type FooBarSpread = {| ...Foo, ...Bar |}
const fooBarInterect: FooBarIntersection = { foo: '123', bar: 12 } // not ok
const fooBarString: FooBarSpread = { foo: '123', bar: 'string' } // not ok
const fooBar: FooBarSpread = { foo: '123', bar: 12 } // ok

While TypeScript does understand object spread, the support for object type spread is not implemented.

Useful References

About

Differences between Flowtype and TypeScript -- syntax and usability

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 24

AltStyle によって変換されたページ (->オリジナル) /