RFC-0006: mapped types

summary

mapped types allow transforming the shape of record types at the type level, enabling patterns like making all fields optional, readonly, or transforming field types.

motivation

common patterns require creating variations of existing types:

type User = {
  id: u64,
  name: string,
  email: string,
};

// manual partial type
type PartialUser = {
  id: Option<u64>,
  name: Option<string>,
  email: Option<string>,
};

// manual readonly type
type ReadonlyUser = {
  const id: u64,
  const name: string,
  const email: string,
};

with mapped types:

type PartialUser = Partial<User>;
type ReadonlyUser = Readonly<User>;

detailed design

basic syntax

mapped types iterate over the fields of a type:

type Partial<T> = {
  [K in keyof T]: Option<T[K]>,
};

type Readonly<T> = {
  const [K in keyof T]: T[K],
};

keyof operator

keyof T produces a union of field name types:

type UserKeys = keyof User;  // "id" | "name" | "email"

field access

T[K] accesses the type of field K in T:

type IdType = User["id"];  // u64

modifiers

mapped types can add or remove modifiers:

// add const (readonly)
type Readonly<T> = {
  const [K in keyof T]: T[K],
};

// remove const (mutable)
type Mutable<T> = {
  var [K in keyof T]: T[K],
};

// add optional
type Partial<T> = {
  [K in keyof T]?: T[K],
};

// remove optional (required)
type Required<T> = {
  [K in keyof T]-?: T[K],
};

standard mapped types

built into the prelude:

type Partial<T>   // all fields optional
type Required<T>  // all fields required
type Readonly<T>  // all fields const
type Mutable<T>   // all fields var
type Pick<T, K>   // subset of fields
type Omit<T, K>   // exclude fields

pick and omit

select or exclude specific fields:

type UserCredentials = Pick<User, "email" | "password">;
// { email: string, password: string }

type PublicUser = Omit<User, "password" | "email">;
// { id: u64, name: string }

transforming field types

change the type of all fields:

type Nullable<T> = {
  [K in keyof T]: Option<T[K]>,
};

type Promisified<T> = {
  [K in keyof T]: Promise<T[K]>,
};

conditional mapping

combine with conditional types (RFC-0007):

type FunctionsOnly<T> = {
  [K in keyof T as T[K] extends function ? K : never]: T[K],
};

drawbacks

significant type system complexity
harder to understand error messages
potential for abuse creating complex type gymnastics

alternatives

comptime type generation

use comptime to generate types:

comptime type Partial<T> = generate_partial(T);

more flexible but less type-safe.

manual type definitions

just write out the types manually.

fine for one-offs but doesn't scale.

prior art

language	feature
typescript	mapped types, keyof, indexed access
flow	$Keys, $Values
scala 3	match types

typescript's mapped types are the direct inspiration.

unresolved questions

how do mapped types interact with generics?
should we allow computed field names?
how do we display mapped types in error messages?

future possibilities

key remapping (as clause)
recursive mapped types
template literal types for field names

On this page