Demystifying TypeScript's Extract Type

Many of the applications I develop at work utilize TypeScript for front-end development to provide a high level of type safety while also remaining familiar to developers who are used to JavaScript. In one such application, a need arose that required a more sophisticated type than I use on a daily basis. After some research, I discovered a solution using TypeScript’s Extract type. In this post, I’ll share a few examples of how you can use Extract in your own projects to improve type safety and developer experience.

Use case: Extracting subsets of a union

To get started with Extract, let’s look at a simple example. Suppose we have a type MixedArgs, which is a union type containing four different members of various types. Our goal is to “extract” the members of the union that are functions, which we can do using Extract. This would look something like this:

type MixedArgs = string | number | () => string | () => number
type FunctionArgs = Extract<MixedArgs, Function>

The first argument we pass to Extract is our union type and the second argument is the type that we will use when comparing each member of the union. If a member is assignable to our second argument, it will be included in the resulting type.

Pro tip: The second argument to Extract can also be a union!

Since string and number are not assignable to Function, they will not be included in the resulting type. This results in FunctionArgs evaluating to the following type:

type FunctionArgs = () => string | () => number

Use case: filterProducts

Now that we understand the basic concept, let’s look at a real world example. Say we have an array of products, each of which contains a key type which is a string to determine what type of object it is. This would look something like this:

type Product =
  | { type: "book"; author: string }
  | { type: "movie"; producer: string }
  | { type: "appliance"; manufacturer: string }

Let’s say we want to create a function that takes an array of Products and returns only those which match a specified type. This would look something like this:

function filterProducts(products: Product[], type: Product["type"]) {
  return products.filter((item) => item.type === type)
}

While this function will do exactly what we want at runtime, the type returned from calling filterProducts(products, 'book') will be Product[] even though we know that the resulting array won’t contain any movies or appliances. With the power of Extract, we can improve this:

function filterProducts<T extends Product, U extends T["type"]>(
  products: T[],
  type: U,
) {
  return products.filter(
    (item): item is Extract<T, Record<"type", U>> => item.type === type,
  )
}

There is a lot going on here, so let’s break it down. First, we’ve updated the function to accept two generic arguments: T and U. TypeScript will infer the value of these generic arguments since we have typed the function parameters using the generic variables. Since each generic argument has a corresponding generic constraint, the function will properly type check the arguments provided to the function like it did in the non-generic example we looked at before.

Now that we have our generic arguments, we can add a type predicate to our filter function. Our type predicate allows us to instruct TypeScript about the specific type of the argument passed to the function when the function returns true. Array.prototype.filter will use this to narrow the type of the resulting array.

The type that our type predicate is narrowing to is Extract<T, Record<"type", U>>, which probably looks a bit confusing at first. The type we are extracting from is T, which will be our array of Products that we passed to our function. Record<"type", U> will be an object that looks something like this: { type: 'book' }. Just like our previous example with FunctionArgs, Extract will return a type containing all members that { type: 'book' } is assignable to. Our final resulting type when calling our function will look like this:

type Result = { type: "book"; author: string }[]

Neat, right!

Bonus: Extract with template literal types

For a little bonus, let’s explore a complex but extremely powerful way of using Extract with template literal types to extract keys from an object with keys matching a specific pattern.

Let’s say we have a Person type that contains some details about the user.

type Person = {
  name: string
  email: string
  homePhone: number
  mobilePhone: number
  workPhone: number
}

Now, further suppose that we want to create a type from Person that contains only the phone number keys. This would be fairly simple with Pick as we could do this:

type PhoneInfo = Pick<Person, "homePhone" | "mobilePhone" | "workPhone">

However, if we have a large number of phone number keys, this will be difficult to maintain and prone to errors. With Extract and template literal types we can create the PhoneInfo type very easily with the following code:

type PhoneInfo = {
  [key in Extract<keyof Person, `${string}Phone`>]: Person[key]
}

The resulting type of PhoneInfo is the following:

type PhoneInfo = {
  homePhone: number
  mobilePhone: number
  workPhone: number
}

I don’t know about you, but the power of type constructs like this is one of the reasons I love TypeScript!