Generics

Generics (or polymorphic types) let you write functions, classes, and type aliases that work with any type while preserving type relationships.

function identity<T>(value: T): T {
  return value;
}

When to use this​

Use generics when a function, class, or type alias needs to work across multiple types while preserving the relationship between inputs and outputs. If the type doesn't matter and you don't need to track it, use unknown instead. If there are a fixed number of specific types, a union may be clearer.

Warning: Flow does not infer generic types. If you want something to have a generic type, annotate it. Otherwise, Flow may infer a type that is less polymorphic than you expect.

Syntax of generics​

There are a number of different places where generic types appear in syntax.

Functions with generics​

Functions can create generics by adding the type parameter list <T> before the function parameter list.

You can use generics in the same places you'd add any other type in a function (parameter or return types).

function method<T>(param: T): T {
  return param;
}


const f = function<T>(param: T): T {
  return param;
}

Function types with generics​

Function types can create generics in the same way as normal functions, by adding the type parameter list <T> before the function type parameter list.

You can use generics in the same places you'd add any other type in a function type (parameter or return types).

<T>(param: T) => T

Which then gets used as its own type.

function method(func: <T>(param: T) => T) {
  // ...
}

Classes with generics​

Classes can create generics by placing the type parameter list before the body of the class.

class Item<T> {
  // ...
}

You can use generics in the same places you'd add any other type in a class (property types and method parameter/return types).

class Item<T> {
  prop: T;


  constructor(param: T) {
    this.prop = param;
  }


  method(): T {
    return this.prop;
  }
}

Type aliases with generics​

type Item<T> = {
  foo: T,
  bar: T,
};

Interfaces with generics​

interface Item<T> {
  foo: T,
  bar: T,
}

Supplying Type Arguments to Callables​

You can give callable entities type arguments for their generics directly in the call:

function doSomething<T>(param: T): T {
  // ...
  return param;
}


doSomething<number>(3);

You can also give generic classes type arguments directly in the new expression:

class GenericClass<T> {}
const c = new GenericClass<number>();

If you only want to specify some of the type arguments, you can use _ to let flow infer a type for you:

class GenericClass<T, U=string, V=number>{}
const c = new GenericClass<boolean, _, string>();

Warning: For performance purposes, we always recommend you annotate with concrete arguments when you can. _ is not unsafe, but it is slower than explicitly specifying the type arguments.

Behavior of generics​

Generics act like variables​

Generic types work a lot like variables or function parameters except that they are used for types. You can use them whenever they are in scope.

function constant<T>(value: T): () => T {
  return function(): T {
    return value;
  };
}

Create as many generics as you need​

You can have as many of these generics as you need in the type parameter list, naming them whatever you want:

function identity<One, Two, Three>(one: One, two: Two, three: Three) {
  // ...
}

Generics track values around​

When using a generic type for a value, Flow will track the value and make sure that you aren't replacing it with something else.

function identity<T>(value: T): T {
  return "foo"; // Error!
}

function identity<T>(value: T): T {
  value = "foo"; // Error!
  return value;  // Error!
}

Flow tracks the specific type of the value you pass through a generic, letting you use it later.

function identity<T>(value: T): T {
  return value;
}


let one: 1 = identity(1);
let two: 2 = identity(2);
let three: 3 = identity(42); // Error

Adding types to generics​

Similar to unknown, generics have an "unknown" type. You're not allowed to use a generic as if it were a specific type.

function logFoo<T>(obj: T): T {
  console.log(obj.foo); // Error!
  return obj;
}

You could refine the type, but the generic will still allow any type to be passed in.

function logFoo<T>(obj: T): T {
  if (obj && obj.foo) {
    console.log(obj.foo); // Works.
  }
  return obj;
}


logFoo({ foo: 'foo', bar: 'bar' });  // Works.
logFoo({ bar: 'bar' }); // Works. :(

Instead, you could add a bound to your generic using the extends keyword.

function logFoo<T extends {foo: string, ...}>(obj: T): T {
  console.log(obj.foo); // Works!
  return obj;
}


logFoo({ foo: 'foo', bar: 'bar' });  // Works!
logFoo({ bar: 'bar' }); // Error!

This way you can keep the behavior of generics while only allowing certain types to be used.

function identity<T extends number>(value: T): T {
  return value;
}


let one: 1 = identity(1);
let two: 2 = identity(2);
let three: "three" = identity("three"); // Error!

Generic types act as bounds​

function identity<T>(val: T): T {
  return val;
}


let foo: 'foo' = 'foo';           // Works!
let bar: 'bar' = identity('bar'); // Works!

In Flow, most of the time when you pass one type into another you lose the original type. So that when you pass a specific type into a less specific one Flow "forgets" it was once something more specific.

function identity(val: string): string {
  return val;
}


let foo: 'foo' = 'foo';           // Works!
let bar: 'bar' = identity('bar'); // Error!

Generics allow you to hold onto the more specific type while adding a constraint. In this way types on generics act as "bounds".

function identity<T extends string>(val: T): T {
  return val;
}


let foo: 'foo' = 'foo';           // Works!
let bar: 'bar' = identity('bar'); // Works!

Note that when you have a value with a bound generic type, you can't use it as if it were a more specific type.

function identity<T extends string>(val: T): T {
  let str: string = val; // Works!
  let bar: 'bar'  = val; // Error!
  return val;
}


identity('bar');

Parameterized generics​

Generics sometimes allow you to pass types in like arguments to a function. These are known as parameterized generics (or parametric polymorphism).

For example, a type alias with a generic is parameterized. When you go to use it you will have to provide a type argument.

type Item<T> = {
  prop: T,
}


let item: Item<string> = {
  prop: "value"
};

You can think of this like passing arguments to a function, only the return value is a type that you can use.

Classes (when being used as a type), type aliases, and interfaces all require that you pass type arguments. Functions and function types do not have parameterized generics.

Classes

class Item<T> {
  prop: T;
  constructor(param: T) {
    this.prop = param;
  }
}


let item1: Item<number> = new Item(42); // Works!
let item2: Item = new Item(42); // Error!

Type Aliases

type Item<T> = {
  prop: T,
};


let item1: Item<number> = { prop: 42 }; // Works!
let item2: Item = { prop: 42 }; // Error!

Interfaces

interface HasProp<T> {
  prop: T,
}


class Item {
  prop: string;
}


Item.prototype as HasProp<string>; // Works!
Item.prototype as HasProp; // Error!

Adding defaults to parameterized generics​

You can also provide defaults for parameterized generics just like parameters of a function.

type Item<T extends number = 1> = {
  prop: T,
};


let foo: Item<> = { prop: 1 };
let bar: Item<2> = { prop: 2 };

You must always include the brackets <> when using the type (just like parentheses for a function call). Writing the bare name Item (without <>) errors with [missing-type-arg] even when every parameter has a default. The bare name is reserved for the unapplied type constructor — applying it always requires explicit type arguments or an empty <> to fall back on defaults.

Variance Keywords​

You can also specify the subtyping behavior of a generic via variance keywords. By default, generics behave invariantly, but you may add an out keyword before their declaration to make them behave covariantly, or an in keyword before their declaration to make them behave contravariantly. See our docs on variance for a more information on variance in Flow.

Variance keywords allow you to be more specific about how you intend to use your generics, giving Flow the power to do more precise type checking. For example, you may want this relationship to hold:

type GenericBox<out T> = T;


const x: GenericBox<number> = 3;
x as GenericBox<number| string>;

The example above could not be accomplished without the out variance keyword:

type GenericBoxError<T> = T;


const x: GenericBoxError<number> = 3;
x as GenericBoxError<number| string>; // Error: number | string is not compatible with number.

Note that if you annotate your generic with variance keywords then Flow will check to make sure those types only appear in positions that make sense for that variance keyword. For example, you cannot declare a generic type parameter to behave covariantly and use it in a contravariant position:

type NotActuallyCovariant<out T> = (T) => void; // Error

See Also​

• Type Aliases — defining named types, often combined with generics
• Functions — using generics in function signatures
• Variance — how covariance and contravariance interact with generics
• Conditional Types — types computed from generic type parameters
• Mapped Types — transforming object types using generics
• Modernizing Legacy Flow Syntax — migrating legacy <T: Bound> bounds and +T / -T variance sigils to modern keyword forms