This overview page is a summary contains a shorter version of all the release notes for TypeScript.

TypeScript 3.9

Improvements in Inference and Promise.all

Recent versions of TypeScript (around 3.7) have had updates to the declarations of functions like Promise.all and Promise.race. Unfortunately, that introduced a few regressions, especially when mixing in values with null or undefined.

interface Lion { roar(): void; } interface Seal { singKissFromARose(): void; } async function visitZoo( lionExhibit: Promise<Lion>, sealExhibit: Promise<Seal | undefined> ) { let [lion, seal] = await Promise.all([lionExhibit, sealExhibit]); lion.roar(); // uh oh // ~~~~ // Object is possibly 'undefined'. }

This is strange behavior! The fact that sealExhibit contained an undefined somehow poisoned type of lion to include undefined.

Thanks to a pull request from Jack Bates, this has been fixed with improvements in our inference process in TypeScript 3.9. The above no longer errors. If you’ve been stuck on older versions of TypeScript due to issues around Promises, we encourage you to give 3.9 a shot!

What About the awaited Type?

If you’ve been following our issue tracker and design meeting notes, you might be aware of some work around a new type operator called awaited. This goal of this type operator is to accurately model the way that Promise unwrapping works in JavaScript.

We initially anticipated shipping awaited in TypeScript 3.9, but as we’ve run early TypeScript builds with existing codebases, we’ve realized that the feature needs more design work before we can roll it out to everyone smoothly. As a result, we’ve decided to pull the feature out of our main branch until we feel more confident. We’ll be experimenting more with the feature, but we won’t be shipping it as part of this release.

Speed Improvements

TypeScript 3.9 ships with many new speed improvements. Our team has been focusing on performance after observing extremely poor editing/compilation speed with packages like material-ui and styled-components. We’ve dived deep here, with a series of different pull requests that optimize certain pathological cases involving large unions, intersections, conditional types, and mapped types.

Each of these pull requests gains about a 5-10% reduction in compile times on certain codebases. In total, we believe we’ve achieved around a 40% reduction in material-ui’s compile time!

We also have some changes to file renaming functionality in editor scenarios. We heard from the Visual Studio Code team that when renaming a file, just figuring out which import statements needed to be updated could take between 5 to 10 seconds. TypeScript 3.9 addresses this issue by changing the internals of how the compiler and language service caches file lookups.

While there’s still room for improvement, we hope this work translates to a snappier experience for everyone!

// @ts-expect-error Comments

Imagine that we’re writing a library in TypeScript and we’re exporting some function called doStuff as part of our public API. The function’s types declare that it takes two strings so that other TypeScript users can get type-checking errors, but it also does a runtime check (maybe only in development builds) to give JavaScript users a helpful error.

function doStuff(abc: string, xyz: string) { assert(typeof abc === "string"); assert(typeof xyz === "string"); // do some stuff }

So TypeScript users will get a helpful red squiggle and an error message when they misuse this function, and JavaScript users will get an assertion error. We’d like to test this behavior, so we’ll write a unit test.

expect(() => { doStuff(123, 456); }).toThrow();

Unfortunately if our tests are written in TypeScript, TypeScript will give us an error!

doStuff(123, 456); // ~~~ // error: Type 'number' is not assignable to type 'string'.

That’s why TypeScript 3.9 brings a new feature: // @ts-expect-error comments. When a line is prefixed with a // @ts-expect-error comment, TypeScript will suppress that error from being reported; but if there’s no error, TypeScript will report that // @ts-expect-error wasn’t necessary.

As a quick example, the following code is okay

// @ts-expect-error console.log(47 * "octopus");

while the following code

// @ts-expect-error console.log(1 + 1);

results in the error

Unused '@ts-expect-error' directive.

We’d like to extend a big thanks to Josh Goldberg, the contributor who implemented this feature. For more information, you can take a look at the ts-expect-error pull request.

ts-ignore or ts-expect-error?

In some ways // @ts-expect-error can act as a suppression comment, similar to // @ts-ignore. The difference is that // @ts-ignore will do nothing if the following line is error-free.

You might be tempted to switch existing // @ts-ignore comments over to // @ts-expect-error, and you might be wondering which is appropriate for future code. While it’s entirely up to you and your team, we have some ideas of which to pick in certain situations.

Pick ts-expect-error if:

  • you’re writing test code where you actually want the type system to error on an operation
  • you expect a fix to be coming in fairly quickly and you just need a quick workaround
  • you’re in a reasonably-sized project with a proactive team that wants to remove suppression comments as soon affected code is valid again

Pick ts-ignore if:

  • you have an a larger project and and new errors have appeared in code with no clear owner
  • you are in the middle of an upgrade between two different versions of TypeScript, and a line of code errors in one version but not another.
  • you honestly don’t have the time to decide which of these options is better.

Uncalled Function Checks in Conditional Expressions

In TypeScript 3.7 we introduced uncalled function checks to report an error when you’ve forgotten to call a function.

function hasImportantPermissions(): boolean { // ... } // Oops! if (hasImportantPermissions) { // ~~~~~~~~~~~~~~~~~~~~~~~ // This condition will always return true since the function is always defined. // Did you mean to call it instead? deleteAllTheImportantFiles(); }

However, this error only applied to conditions in if statements. Thanks to a pull request from Alexander Tarasyuk, this feature is also now supported in ternary conditionals (i.e. the cond ? trueExpr : falseExpr syntax).

declare function listFilesOfDirectory(dirPath: string): string[]; declare function isDirectory(): boolean; function getAllFiles(startFileName: string) { const result: string[] = []; traverse(startFileName); return result; function traverse(currentPath: string) { return isDirectory ? // ~~~~~~~~~~~ // This condition will always return true // since the function is always defined. // Did you mean to call it instead? listFilesOfDirectory(currentPath).forEach(traverse) : result.push(currentPath); } }


Editor Improvements

The TypeScript compiler not only powers the TypeScript editing experience in most major editors, it also powers the JavaScript experience in the Visual Studio family of editors and more. Using new TypeScript/JavaScript functionality in your editor will differ depending on your editor, but

CommonJS Auto-Imports in JavaScript

One great new improvement is in auto-imports in JavaScript files using CommonJS modules.

In older versions, TypeScript always assumed that regardless of your file, you wanted an ECMAScript-style import like

import * as fs from "fs";

However, not everyone is targeting ECMAScript-style modules when writing JavaScript files. Plenty of users still use CommonJS-style require(...) imports like so

const fs = require("fs");

TypeScript now automatically detects the types of imports you’re using to keep your file’s style clean and consistent.

For more details on the change, see the corresponding pull request.

Code Actions Preserve Newlines

TypeScript’s refactorings and quick fixes often didn’t do a great job of preserving newlines. As a really basic example, take the following code.

const maxValue = 100; /*start*/ for (let i = 0; i <= maxValue; i++) { // First get the squared value. let square = i ** 2; // Now print the squared value. console.log(square); } /*end*/

If we highlighted the range from /*start*/ to /*end*/ in our editor to extract to a new function, we’d end up with code like the following.

const maxValue = 100; printSquares(); function printSquares() { for (let i = 0; i <= maxValue; i++) { // First get the squared value. let square = i ** 2; // Now print the squared value. console.log(square); } }

Extracting the for loop to a function in older versions of TypeScript. A newline is not preserved.

That’s not ideal - we had a blank line between each statement in our for loop, but the refactoring got rid of it! TypeScript 3.9 does a little more work to preserve what we write.

const maxValue = 100; printSquares(); function printSquares() { for (let i = 0; i <= maxValue; i++) { // First get the squared value. let square = i ** 2; // Now print the squared value. console.log(square); } }

Extracting the for loop to a function in TypeScript 3.9. A newline is preserved.

You can see more about the implementation in this pull request

Quick Fixes for Missing Return Expressions

There are occasions where we might forget to return the value of the last statement in a function, especially when adding curly braces to arrow functions.

// before let f1 = () => 42; // oops - not the same! let f2 = () => { 42; };

Thanks to a pull request from community member Wenlu Wang, TypeScript can provide a quick-fix to add missing return statements, remove curly braces, or add parentheses to arrow function bodies that look suspiciously like object literals.

TypeScript fixing an error where no expression is returned by adding a `return` statement or removing curly braces.

Support for “Solution Style” tsconfig.json Files

Editors need to figure out which configuration file a file belongs to so that it can apply the appropriate options and figure out which other files are included in the current “project”. By default, editors powered by TypeScript’s language server do this by walking up each parent directory to find a tsconfig.json.

One case where this slightly fell over is when a tsconfig.json simply existed to reference other tsconfig.json files.

// tsconfig.json { files: [], references: [ { path: "./tsconfig.shared.json" }, { path: "./tsconfig.frontend.json" }, { path: "./tsconfig.backend.json" }, ], }

This file that really does nothing but manage other project files is often called a “solution” in some environments. Here, none of these tsconfig.*.json files get picked up by the server, but we’d really like the language server to understand that the current .ts file probably belongs to one of the mentioned projects in this root tsconfig.json.

TypeScript 3.9 adds support to editing scenarios for this configuration. For more details, take a look at the pull request that added this functionality.

Breaking Changes

Parsing Differences in Optional Chaining and Non-Null Assertions

TypeScript recently implemented the optional chaining operator, but we’ve received user feedback that the behavior of optional chaining (?.) with the non-null assertion operator (!) is extremely counter-intuitive.

Specifically, in previous versions, the code


was interpreted to be equivalent to the following JavaScript.


In the above code the parentheses stop the “short-circuiting” behavior of optional chaining, so if foo is undefined, accessing baz will cause a runtime error.

The Babel team who pointed this behavior out, and most users who provided feedback to us, believe that this behavior is wrong. We do too! The thing we heard the most was that the ! operator should just “disappear” since the intent was to remove null and undefined from the type of bar.

In other words, most people felt that the original snippet should be interpreted as


which just evaluates to undefined when foo is undefined.

This is a breaking change, but we believe most code was written with the new interpretation in mind. Users who want to revert to the old behavior can add explicit parentheses around the left side of the ! operator.


} and > are Now Invalid JSX Text Characters

The JSX Specification forbids the use of the } and > characters in text positions. TypeScript and Babel have both decided to enforce this rule to be more comformant. The new way to insert these characters is to use an HTML escape code (e.g. <span> 2 &gt 1 </div>) or insert an expression with a string literal (e.g. <span> 2 {">"} 1 </div>).

Luckily, thanks to the pull request enforcing this from Brad Zacher, you’ll get an error message along the lines of

Unexpected token. Did you mean `{'>'}` or `&gt;`?
Unexpected token. Did you mean `{'}'}` or `&rbrace;`?

For example:

let directions = <span>Navigate to: Menu Bar > Tools > Options</div> // ~ ~ // Unexpected token. Did you mean `{'>'}` or `>`?

That error message came with a handy quick fix, and thanks to Alexander Tarasyuk, you can apply these changes in bulk if you have a lot of errors.

Stricter Checks on Intersections and Optional Properties

Generally, an intersection type like A & B is assignable to C if either A or B is assignable to C; however, sometimes that has problems with optional properties. For example, take the following:

interface A { a: number; // notice this is 'number' } interface B { b: string; } interface C { a?: boolean; // notice this is 'boolean' b: string; } declare let x: A & B; declare let y: C; y = x;

In previous versions of TypeScript, this was allowed because while A was totally incompatible with C, B was compatible with C.

In TypeScript 3.9, so long as every type in an intersection is a concrete object type, the type system will consider all of the properties at once. As a result, TypeScript will see that the a property of A & B is incompatible with that of C:

Type 'A & B' is not assignable to type 'C'.
  Types of property 'a' are incompatible.
    Type 'number' is not assignable to type 'boolean | undefined'.

For more information on this change, see the corresponding pull request.

Intersections Reduced By Discriminant Properties

There are a few cases where you might end up with types that describe values that just don’t exist. For example

declare function smushObjects<T, U>(x: T, y: U): T & U; interface Circle { kind: "circle"; radius: number; } interface Square { kind: "square"; sideLength: number; } declare let x: Circle; declare let y: Square; let z = smushObjects(x, y); console.log(z.kind);

This code is slightly weird because there’s really no way to create an intersection of a Circle and a Square - they have two incompatible kind fields. In previous versions of TypeScript, this code was allowed and the type of kind itself was never because "circle" & "square" described a set of values that could never exist.

In TypeScript 3.9, the type system is more aggressive here - it notices that it’s impossible to intersect Circle and Square because of their kind properties. So instead of collapsing the type of z.kind to never, it collapses the type of z itself (Circle & Square) to never. That means the above code now errors with:

Property 'kind' does not exist on type 'never'.

Most of the breaks we observed seem to correspond with slightly incorrect type declarations. For more details, see the original pull request.

Getters/Setters are No Longer Enumerable

In older versions of TypeScript, get and set accessors in classes were emitted in a way that made them enumerable; however, this wasn’t compliant with the ECMAScript specification which states that they must be non-enumerable. As a result, TypeScript code that targeted ES5 and ES2015 could differ in behavior.

Thanks to a pull request from GitHub user pathurs, TypeScript 3.9 now conforms more closely with ECMAScript in this regard.

Type Parameters That Extend any No Longer Act as any

In previous versions of TypeScript, a type parameter constrained to any could be treated as any.

function foo<T extends any>(arg: T) { arg.spfjgerijghoied; // no error! }

This was an oversight, so TypeScript 3.9 takes a more conservative approach and issues an error on these questionable operations.

function foo<T extends any>(arg: T) { arg.spfjgerijghoied; // ~~~~~~~~~~~~~~~ // Property 'spfjgerijghoied' does not exist on type 'T'. }

export * is Always Retained

In previous TypeScript versions, declarations like export * from "foo" would be dropped in our JavaScript output if foo didn’t export any values. This sort of emit is problematic because it’s type-directed and can’t be emulated by Babel. TypeScript 3.9 will always emit these export * declarations. In practice, we don’t expect this to break much existing code.

TypeScript 3.8

Type-Only Imports and Export

This feature is something most users may never have to think about; however, if you’ve hit issues under --isolatedModules, TypeScript’s transpileModule API, or Babel, this feature might be relevant.

TypeScript 3.8 adds a new syntax for type-only imports and exports.

import type { SomeThing } from "./some-module.js"; export type { SomeThing };

import type only imports declarations to be used for type annotations and declarations. It always gets fully erased, so there’s no remnant of it at runtime. Similarly, export type only provides an export that can be used for type contexts, and is also erased from TypeScript’s output.

It’s important to note that classes have a value at runtime and a type at design-time, and the use is context-sensitive. When using import type to import a class, you can’t do things like extend from it.

import type { Component } from "react"; interface ButtonProps { // ... } class Button extends Component<ButtonProps> { // ~~~~~~~~~ // error! 'Component' only refers to a type, but is being used as a value here. // ... }

If you’ve used Flow before, the syntax is fairly similar. One difference is that we’ve added a few restrictions to avoid code that might appear ambiguous.

// Is only 'Foo' a type? Or every declaration in the import? // We just give an error because it's not clear. import type Foo, { Bar, Baz } from "some-module"; // ~~~~~~~~~~~~~~~~~~~~~~ // error! A type-only import can specify a default import or named bindings, but not both.

In conjunction with import type, TypeScript 3.8 also adds a new compiler flag to control what happens with imports that won’t be utilized at runtime: importsNotUsedAsValues. This flag takes 3 different values:

  • remove: this is today’s behavior of dropping these imports. It’s going to continue to be the default, and is a non-breaking change.
  • preserve: this preserves all imports whose values are never used. This can cause imports/side-effects to be preserved.
  • error: this preserves all imports (the same as the preserve option), but will error when a value import is only used as a type. This might be useful if you want to ensure no values are being accidentally imported, but still make side-effect imports explicit.

For more information about the feature, you can take a look at the pull request, and relevant changes around broadening where imports from an import type declaration can be used.

ECMAScript Private Fields

TypeScript 3.8 brings support for ECMAScript’s private fields, part of the stage-3 class fields proposal.

class Person { #name: string; constructor(name: string) { this.#name = name; } greet() { console.log(`Hello, my name is ${this.#name}!`); } } let jeremy = new Person("Jeremy Bearimy"); jeremy.#name; // ~~~~~ // Property '#name' is not accessible outside class 'Person' // because it has a private identifier.

Unlike regular properties (even ones declared with the private modifier), private fields have a few rules to keep in mind. Some of them are:

  • Private fields start with a # character. Sometimes we call these private names.
  • Every private field name is uniquely scoped to its containing class.
  • TypeScript accessibility modifiers like public or private can’t be used on private fields.
  • Private fields can’t be accessed or even detected outside of the containing class - even by JS users! Sometimes we call this hard privacy.

Apart from “hard” privacy, another benefit of private fields is that uniqueness we just mentioned. For example, regular property declarations are prone to being overwritten in subclasses.

class C { foo = 10; cHelper() { return this.foo; } } class D extends C { foo = 20; dHelper() { return this.foo; } } let instance = new D(); // 'this.foo' refers to the same property on each instance. console.log(instance.cHelper()); // prints '20' console.log(instance.dHelper()); // prints '20'

With private fields, you’ll never have to worry about this, since each field name is unique to the containing class.

class C { #foo = 10; cHelper() { return this.#foo; } } class D extends C { #foo = 20; dHelper() { return this.#foo; } } let instance = new D(); // 'this.#foo' refers to a different field within each class. console.log(instance.cHelper()); // prints '10' console.log(instance.dHelper()); // prints '20'

Another thing worth noting is that accessing a private field on any other type will result in a TypeError!

class Square { #sideLength: number; constructor(sideLength: number) { this.#sideLength = sideLength; } equals(other: any) { return this.#sideLength === other.#sideLength; } } const a = new Square(100); const b = { sideLength: 100 }; // Boom! // TypeError: attempted to get private field on non-instance // This fails because 'b' is not an instance of 'Square'. console.log(a.equals(b));

Finally, for any plain .js file users, private fields always have to be declared before they’re assigned to.

class C { // No declaration for '#foo' // :( constructor(foo: number) { // SyntaxError! // '#foo' needs to be declared before writing to it. this.#foo = foo; } }

JavaScript has always allowed users to access undeclared properties, whereas TypeScript has always required declarations for class properties. With private fields, declarations are always needed regardless of whether we’re working in .js or .ts files.

class C { /** @type {number} */ #foo; constructor(foo: number) { // This works. this.#foo = foo; } }

For more information about the implementation, you can check out the original pull request

Which should I use?

We’ve already received many questions on which type of privates you should use as a TypeScript user: most commonly, “should I use the private keyword, or ECMAScript’s hash/pound (#) private fields?” It depends!

When it comes to properties, TypeScript’s private modifiers are fully erased - that means that at runtime, it acts entirely like a normal property and there’s no way to tell that it was declared with a private modifier. When using theprivate` keyword, privacy is only enforced at compile-time/design-time, and for JavaScript consumers it’s entirely intent-based.

class C { private foo = 10; } // This is an error at compile time, // but when TypeScript outputs .js files, // it'll run fine and print '10'. console.log(new C().foo); // prints '10' // ~~~ // error! Property 'foo' is private and only accessible within class 'C'. // TypeScript allows this at compile-time // as a "work-around" to avoid the error. console.log(new C()["foo"]); // prints '10'

The upside is that this sort of “soft privacy” can help your consumers temporarily work around not having access to some API, and also works in any runtime.

On the other hand, ECMAScript’s # privates are completely inaccessible outside of the class.

class C { #foo = 10; } console.log(new C().#foo); // SyntaxError // ~~~~ // TypeScript reports an error *and* // this won't work at runtime! console.log(new C()["#foo"]); // prints undefined // ~~~~~~~~~~~~~~~ // TypeScript reports an error under 'noImplicitAny', // and this prints 'undefined'.

This hard privacy is really useful for strictly ensuring that nobody can take use of any of your internals. If you’re a library author, removing or renaming a private field should never cause a breaking change.

As we mentioned, another benefit is that subclassing can be easier with ECMAScript’s # privates because they really are private. When using ECMAScript # private fields, no subclass ever has to worry about collisions in field naming. When it comes to TypeScript’s private property declarations, users still have to be careful not to trample over properties declared in superclasses.

One more thing to think about is where you intend for your code to run. TypeScript currently can’t support this feature unless targeting ECMAScript 2015 (ES6) targets or higher. This is because our downleveled implementation uses WeakMaps to enforce privacy, and WeakMaps can’t be polyfilled in a way that doesn’t cause memory leaks. In contrast, TypeScript’s private-declared properties work with all targets - even ECMAScript 3!

A final consideration might be speed: private properties are no different from any other property, so accessing them is as fast as any other property access no matter which runtime you target. In contrast, because # private fields are downleveled using WeakMaps, they may be slower to use. While some runtimes might optimize their actual implementations of # private fields, and even have speedy WeakMap implementations, that might not be the case in all runtimes.

export * as ns Syntax

It’s often common to have a single entry-point that exposes all the members of another module as a single member.

import * as utilities from "./utilities.js"; export { utilities };

This is so common that ECMAScript 2020 recently added a new syntax to support this pattern!

export * as utilities from "./utilities.js";

This is a nice quality-of-life improvement to JavaScript, and TypeScript 3.8 implements this syntax. When your module target is earlier than es2020, TypeScript will output something along the lines of the first code snippet.

Top-Level await

TypeScript 3.8 provides support for a handy upcoming ECMAScript feature called “top-level await“.

JavaScript users often introduce an async function in order to use await, and then immediately called the function after defining it.

async function main() { const response = await fetch("..."); const greeting = await response.text(); console.log(greeting); } main().catch((e) => console.error(e));

This is because previously in JavaScript (along with most other languages with a similar feature), await was only allowed within the body of an async function. However, with top-level await, we can use await at the top level of a module.

const response = await fetch("..."); const greeting = await response.text(); console.log(greeting); // Make sure we're a module export {};

Note there’s a subtlety: top-level await only works at the top level of a module, and files are only considered modules when TypeScript finds an import or an export. In some basic cases, you might need to write out export {} as some boilerplate to make sure of this.

Top level await may not work in all environments where you might expect at this point. Currently, you can only use top level await when the target compiler option is es2017 or above, and module is esnext or system. Support within several environments and bundlers may be limited or may require enabling experimental support.

For more information on our implementation, you can check out the original pull request.

es2020 for target and module

TypeScript 3.8 supports es2020 as an option for module and target. This will preserve newer ECMAScript 2020 features like optional chaining, nullish coalescing, export * as ns, and dynamic import(...) syntax. It also means bigint literals now have a stable target below esnext.

JSDoc Property Modifiers

TypeScript 3.8 supports JavaScript files by turning on the allowJs flag, and also supports type-checking those JavaScript files via the checkJs option or by adding a // @ts-check comment to the top of your .js files.

Because JavaScript files don’t have dedicated syntax for type-checking, TypeScript leverages JSDoc. TypeScript 3.8 understands a few new JSDoc tags for properties.

First are the accessibility modifiers: @public, @private, and @protected. These tags work exactly like public, private, and protected respectively work in TypeScript.

// @ts-check class Foo { constructor() { /** @private */ this.stuff = 100; } printStuff() { console.log(this.stuff); } } new Foo().stuff; // ~~~~~ // error! Property 'stuff' is private and only accessible within class 'Foo'.
  • @public is always implied and can be left off, but means that a property can be reached from anywhere.
  • @private means that a property can only be used within the containing class.
  • @protected means that a property can only be used within the containing class, and all derived subclasses, but not on dissimilar instances of the containing class.

Next, we’ve also added the @readonly modifier to ensure that a property is only ever written to during initialization.

// @ts-check class Foo { constructor() { /** @readonly */ this.stuff = 100; } writeToStuff() { this.stuff = 200; // ~~~~~ // Cannot assign to 'stuff' because it is a read-only property. } } new Foo().stuff++; // ~~~~~ // Cannot assign to 'stuff' because it is a read-only property.

Better Directory Watching on Linux and watchOptions

TypeScript 3.8 ships a new strategy for watching directories, which is crucial for efficiently picking up changes to node_modules.

For some context, on operating systems like Linux, TypeScript installs directory watchers (as opposed to file watchers) on node_modules and many of its subdirectories to detect changes in dependencies. This is because the number of available file watchers is often eclipsed by the of files in node_modules, whereas there are way fewer directories to track.

Older versions of TypeScript would immediately install directory watchers on folders, and at startup that would be fine; however, during an npm install, a lot of activity will take place within node_modules and that can overwhelm TypeScript, often slowing editor sessions to a crawl. To prevent this, TypeScript 3.8 waits slightly before installing directory watchers to give these highly volatile directories some time to stabilize.

Because every project might work better under different strategies, and this new approach might not work well for your workflows, TypeScript 3.8 introduces a new watchOptions field in tsconfig.json and jsconfig.json which allows users to tell the compiler/language service which watching strategies should be used to keep track of files and directories.

{ // Some typical compiler options compilerOptions: { target: "es2020", moduleResolution: "node", // ... }, // NEW: Options for file/directory watching watchOptions: { // Use native file system events for files and directories watchFile: "useFsEvents", watchDirectory: "useFsEvents", // Poll files for updates more frequently // when they're updated a lot. fallbackPolling: "dynamicPriority", }, }

watchOptions contains 4 new options that can be configured to handle how TypeScript keeps track of changes.

For more information on these changes, head over to GitHub to see the pull request to read more.

“Fast and Loose” Incremental Checking

TypeScript 3.8 introduces a new compiler option called assumeChangesOnlyAffectDirectDependencies. When this option is enabled, TypeScript will avoid rechecking/rebuilding all truly possibly-affected files, and only recheck/rebuild files that have changed as well as files that directly import them.

In a codebase like Visual Studio Code, this reduced rebuild times for changes in certain files from about 14 seconds to about 1 second. While we don’t necessarily recommend this option for all codebases, you might be interested if you have an extremely large codebase and are willing to defer full project errors until later (e.g. a dedicated build via a tsconfig.fullbuild.json or in CI).

For more details, you can see the original pull request.

TypeScript 3.7

Optional Chaining


Optional chaining is issue #16 on our issue tracker. For context, there have been over 23,000 issues on the TypeScript issue tracker since then.

At its core, optional chaining lets us write code where TypeScript can immediately stop running some expressions if we run into a null or undefined. The star of the show in optional chaining is the new ?. operator for optional property accesses. When we write code like

let x = foo?.bar.baz();

this is a way of saying that when foo is defined, foo.bar.baz() will be computed; but when foo is null or undefined, stop what we’re doing and just return undefined.”

More plainly, that code snippet is the same as writing the following.

let x = foo === null || foo === undefined ? undefined : foo.bar.baz();

Note that if bar is null or undefined, our code will still hit an error accessing baz. Likewise, if baz is null or undefined, we’ll hit an error at the call site. ?. only checks for whether the value on the left of it is null or undefined - not any of the subsequent properties.

You might find yourself using ?. to replace a lot of code that performs repetitive nullish checks using the && operator.

// Before if (foo && foo.bar && foo.bar.baz) { // ... } // After-ish if (foo?.bar?.baz) { // ... }

Keep in mind that ?. acts differently than those && operations since && will act specially on “falsy” values (e.g. the empty string, 0, NaN, and, well, false), but this is an intentional feature of the construct. It doesn’t short-circuit on valid data like 0 or empty strings.

Optional chaining also includes two other operations. First there’s the optional element access which acts similarly to optional property accesses, but allows us to access non-identifier properties (e.g. arbitrary strings, numbers, and symbols):

/** * Get the first element of the array if we have an array. * Otherwise return undefined. */ function tryGetFirstElement<T>(arr?: T[]) { return arr?.[0]; // equivalent to // return (arr === null || arr === undefined) ? // undefined : // arr[0]; }

There’s also optional call, which allows us to conditionally call expressions if they’re not null or undefined.

async function makeRequest(url: string, log?: (msg: string) => void) { log?.(`Request started at ${new Date().toISOString()}`); // roughly equivalent to // if (log != null) { // log(`Request started at ${new Date().toISOString()}`); // } const result = (await fetch(url)).json(); log?.(`Request finished at at ${new Date().toISOString()}`); return result; }

The “short-circuiting” behavior that optional chains have is limited property accesses, calls, element accesses - it doesn’t expand any further out from these expressions. In other words,

let result = foo?.bar / someComputation();

doesn’t stop the division or someComputation() call from occurring. It’s equivalent to

let temp = foo === null || foo === undefined ? undefined : foo.bar; let result = temp / someComputation();

That might result in dividing undefined, which is why in strictNullChecks, the following is an error.

function barPercentage(foo?: { bar: number }) { return foo?.bar / 100; // ~~~~~~~~ // Error: Object is possibly undefined. }

More more details, you can read up on the proposal and view the original pull request.

Nullish Coalescing


The nullish coalescing operator is another upcoming ECMAScript feature that goes hand-in-hand with optional chaining, and which our team has been involved with championing in TC39.

You can think of this feature - the ?? operator - as a way to “fall back” to a default value when dealing with null or undefined. When we write code like

let x = foo ?? bar();

this is a new way to say that the value foo will be used when it’s “present”; but when it’s null or undefined, calculate bar() in its place.

Again, the above code is equivalent to the following.

let x = foo !== null && foo !== undefined ? foo : bar();

The ?? operator can replace uses of || when trying to use a default value. For example, the following code snippet tries to fetch the volume that was last saved in localStorage (if it ever was); however, it has a bug because it uses ||.

function initializeAudio() { let volume = localStorage.volume || 0.5; // ... }

When localStorage.volume is set to 0, the page will set the volume to 0.5 which is unintended. ?? avoids some unintended behavior from 0, NaN and "" being treated as falsy values.

We owe a large thanks to community members Wenlu Wang and Titian Cernicova Dragomir for implementing this feature! For more details, check out their pull request and the nullish coalescing proposal repository.

Assertion Functions


There’s a specific set of functions that throw an error if something unexpected happened. They’re called “assertion” functions. As an example, Node.js has a dedicated function for this called assert.

assert(someValue === 42);

In this example if someValue isn’t equal to 42, then assert will throw an AssertionError.

Assertions in JavaScript are often used to guard against improper types being passed in. For example,

function multiply(x, y) { assert(typeof x === "number"); assert(typeof y === "number"); return x * y; }

Unfortunately in TypeScript these checks could never be properly encoded. For loosely-typed code this meant TypeScript was checking less, and for slightly conservative code it often forced users to use type assertions.

function yell(str) { assert(typeof str === "string"); return str.toUppercase(); // Oops! We misspelled 'toUpperCase'. // Would be great if TypeScript still caught this! }

The alternative was to instead rewrite the code so that the language could analyze it, but this isn’t convenient.

function yell(str) { if (typeof str !== "string") { throw new TypeError("str should have been a string."); } // Error caught! return str.toUppercase(); }

Ultimately the goal of TypeScript is to type existing JavaScript constructs in the least disruptive way. For that reason, TypeScript 3.7 introduces a new concept called “assertion signatures” which model these assertion functions.

The first type of assertion signature models the way that Node’s assert function works. It ensures that whatever condition is being checked must be true for the remainder of the containing scope.

function assert(condition: any, msg?: string): asserts condition { if (!condition) { throw new AssertionError(msg); } }

asserts condition says that whatever gets passed into the condition parameter must be true if the assert returns (because otherwise it would throw an error). That means that for the rest of the scope, that condition must be truthy. As an example, using this assertion function means we do catch our original yell example.

function yell(str) { assert(typeof str === "string"); return str.toUppercase(); // ~~~~~~~~~~~ // error: Property 'toUppercase' does not exist on type 'string'. // Did you mean 'toUpperCase'? } function assert(condition: any, msg?: string): asserts condition { if (!condition) { throw new AssertionError(msg); } }

The other type of assertion signature doesn’t check for a condition, but instead tells TypeScript that a specific variable or property has a different type.

function assertIsString(val: any): asserts val is string { if (typeof val !== "string") { throw new AssertionError("Not a string!"); } }

Here asserts val is string ensures that after any call to assertIsString, any variable passed in will be known to be a string.

function yell(str: any) { assertIsString(str); // Now TypeScript knows that 'str' is a 'string'. return str.toUppercase(); // ~~~~~~~~~~~ // error: Property 'toUppercase' does not exist on type 'string'. // Did you mean 'toUpperCase'? }

These assertion signatures are very similar to writing type predicate signatures:

function isString(val: any): val is string { return typeof val === "string"; } function yell(str: any) { if (isString(str)) { return str.toUppercase(); } throw "Oops!"; }

And just like type predicate signatures, these assertion signatures are incredibly expressive. We can express some fairly sophisticated ideas with these.

function assertIsDefined<T>(val: T): asserts val is NonNullable<T> { if (val === undefined || val === null) { throw new AssertionError( `Expected 'val' to be defined, but received ${val}` ); } }

To read up more about assertion signatures, check out the original pull request.

Better Support for never-Returning Functions

As part of the work for assertion signatures, TypeScript needed to encode more about where and which functions were being called. This gave us the opportunity to expand support for another class of functions: functions that return never.

The intent of any function that returns never is that it never returns. It indicates that an exception was thrown, a halting error condition occurred, or that the program exited. For example, process.exit(...) in @types/node is specified to return never.

In order to ensure that a function never potentially returned undefined or effectively returned from all code paths, TypeScript needed some syntactic signal - either a return or throw at the end of a function. So users found themselves return-ing their failure functions.

function dispatch(x: string | number): SomeType { if (typeof x === "string") { return doThingWithString(x); } else if (typeof x === "number") { return doThingWithNumber(x); } return process.exit(1); }

Now when these never-returning functions are called, TypeScript recognizes that they affect the control flow graph and accounts for them.

function dispatch(x: string | number): SomeType { if (typeof x === "string") { return doThingWithString(x); } else if (typeof x === "number") { return doThingWithNumber(x); } process.exit(1); }

As with assertion functions, you can read up more at the same pull request.

(More) Recursive Type Aliases


Type aliases have always had a limitation in how they could be “recursively” referenced. The reason is that any use of a type alias needs to be able to substitute itself with whatever it aliases. In some cases, that’s not possible, so the compiler rejects certain recursive aliases like the following:

type Foo = Foo;

This is a reasonable restriction because any use of Foo would need to be replaced with Foo which would need to be replaced with Foo which would need to be replaced with Foo which… well, hopefully you get the idea! In the end, there isn’t a type that makes sense in place of Foo.

This is fairly consistent with how other languages treat type aliases, but it does give rise to some slightly surprising scenarios for how users leverage the feature. For example, in TypeScript 3.6 and prior, the following causes an error.

type ValueOrArray<T> = T | Array<ValueOrArray<T>>; // ~~~~~~~~~~~~ // error: Type alias 'ValueOrArray' circularly references itself.

This is strange because there is technically nothing wrong with any use users could always write what was effectively the same code by introducing an interface.

type ValueOrArray<T> = T | ArrayOfValueOrArray<T>; interface ArrayOfValueOrArray<T> extends Array<ValueOrArray<T>> {}

Because interfaces (and other object types) introduce a level of indirection and their full structure doesn’t need to be eagerly built out, TypeScript has no problem working with this structure.

But workaround of introducing the interface wasn’t intuitive for users. And in principle there really wasn’t anything wrong with the original version of ValueOrArray that used Array directly. If the compiler was a little bit “lazier” and only calculated the type arguments to Array when necessary, then TypeScript could express these correctly.

That’s exactly what TypeScript 3.7 introduces. At the “top level” of a type alias, TypeScript will defer resolving type arguments to permit these patterns.

This means that code like the following that was trying to represent JSON…

type Json = string | number | boolean | null | JsonObject | JsonArray; interface JsonObject { [property: string]: Json; } interface JsonArray extends Array<Json> {}

can finally be rewritten without helper interfaces.

type Json = | string | number | boolean | null | { [property: string]: Json } | Json[];

This new relaxation also lets us recursively reference type aliases in tuples as well. The following code which used to error is now valid TypeScript code.

type VirtualNode = string | [string, { [key: string]: any }, ...VirtualNode[]]; const myNode: VirtualNode = [ "div", { id: "parent" }, ["div", { id: "first-child" }, "I'm the first child"], ["div", { id: "second-child" }, "I'm the second child"], ];

For more information, you can read up on the original pull request.

--declaration and --allowJs

The --declaration flag in TypeScript allows us to generate .d.ts files (declaration files) from TypeScript source files (i.e. .ts and .tsx files). These .d.ts files are important for a couple of reasons.

First of all, they’re important because they allow TypeScript to type-check against other projects without re-checking the original source code. They’re also important because they allow TypeScript to interoperate with existing JavaScript libraries that weren’t built with TypeScript in mind. Finally, a benefit that is often underappreciated: both TypeScript and JavaScript users can benefit from these files when using editors powered by TypeScript to get things like better auto-completion.

Unfortunately, --declaration didn’t work with the --allowJs flag which allows mixing TypeScript and JavaScript input files. This was a frustrating limitation because it meant users couldn’t use the --declaration flag when migrating codebases, even if they were JSDoc-annotated. TypeScript 3.7 changes that, and allows the two options to be used together!

The most impactful outcome of this feature might a bit subtle: with TypeScript 3.7, users can write libraries in JSDoc annotated JavaScript and support TypeScript users.

The way that this works is that when using allowJs, TypeScript has some best-effort analyses to understand common JavaScript patterns; however, the way that some patterns are expressed in JavaScript don’t necessarily look like their equivalents in TypeScript. When declaration emit is turned on, TypeScript figures out the best way to transform JSDoc comments and CommonJS exports into valid type declarations and the like in the output .d.ts files.

As an example, the following code snippet

const assert = require("assert"); module.exports.blurImage = blurImage; /** * Produces a blurred image from an input buffer. * * @param input {Uint8Array} * @param width {number} * @param height {number} */ function blurImage(input, width, height) { const numPixels = width * height * 4; assert(input.length === numPixels); const result = new Uint8Array(numPixels); // TODO return result; }

Will produce a .d.ts file like

/** * Produces a blurred image from an input buffer. * * @param input {Uint8Array} * @param width {number} * @param height {number} */ export function blurImage( input: Uint8Array, width: number, height: number ): Uint8Array;

This can go beyond basic functions with @param tags too, where the following example:

/** * @callback Job * @returns {void} */ /** Queues work */ export class Worker { constructor(maxDepth = 10) { this.started = false; this.depthLimit = maxDepth; /** * NOTE: queued jobs may add more items to queue * @type {Job[]} */ this.queue = []; } /** * Adds a work item to the queue * @param {Job} work */ push(work) { if (this.queue.length + 1 > this.depthLimit) throw new Error("Queue full!"); this.queue.push(work); } /** * Starts the queue if it has not yet started */ start() { if (this.started) return false; this.started = true; while (this.queue.length) { /** @type {Job} */ (this.queue.shift())(); } return true; } }

will be transformed into the following .d.ts file:

/** * @callback Job * @returns {void} */ /** Queues work */ export class Worker { constructor(maxDepth?: number); started: boolean; depthLimit: number; /** * NOTE: queued jobs may add more items to queue * @type {Job[]} */ queue: Job[]; /** * Adds a work item to the queue * @param {Job} work */ push(work: Job): void; /** * Starts the queue if it has not yet started */ start(): boolean; } export type Job = () => void;

Note that when using these flags together, TypeScript doesn’t necessarily have to downlevel .js files. If you simply want TypeScript to create .d.ts files, you can use the --emitDeclarationOnly compiler option.

For more details, you can check out the original pull request.

The useDefineForClassFields Flag and The declare Property Modifier

Back when TypeScript implemented public class fields, we assumed to the best of our abilities that the following code

class C { foo = 100; bar: string; }

would be equivalent to a similar assignment within a constructor body.

class C { constructor() { this.foo = 100; } }

Unfortunately, while this seemed to be the direction that the proposal moved towards in its earlier days, there is an extremely strong chance that public class fields will be standardized differently. Instead, the original code sample might need to de-sugar to something closer to the following:

class C { constructor() { Object.defineProperty(this, "foo", { enumerable: true, configurable: true, writable: true, value: 100, }); Object.defineProperty(this, "bar", { enumerable: true, configurable: true, writable: true, value: void 0, }); } }

While TypeScript 3.7 isn’t changing any existing emit by default, we’ve been rolling out changes incrementally to help users mitigate potential future breakage. We’ve provided a new flag called useDefineForClassFields to enable this emit mode with some new checking logic.

The two biggest changes are the following:

  • Declarations are initialized with Object.defineProperty.
  • Declarations are always initialized to undefined, even if they have no initializer.

This can cause quite a bit of fallout for existing code that use inheritance. First of all, set accessors from base classes won’t get triggered - they’ll be completely overwritten.

class Base { set data(value: string) { console.log("data changed to " + value); } } class Derived extends Base { // No longer triggers a 'console.log' // when using 'useDefineForClassFields'. data = 10; }

Secondly, using class fields to specialize properties from base classes also won’t work.

interface Animal { animalStuff: any; } interface Dog extends Animal { dogStuff: any; } class AnimalHouse { resident: Animal; constructor(animal: Animal) { this.resident = animal; } } class DogHouse extends AnimalHouse { // Initializes 'resident' to 'undefined' // after the call to 'super()' when // using 'useDefineForClassFields'! resident: Dog; constructor(dog: Dog) { super(dog); } }

What these two boil down to is that mixing properties with accessors is going to cause issues, and so will re-declaring properties with no initializers.

To detect the issue around accessors, TypeScript 3.7 will now emit get/set accessors in .d.ts files so that in TypeScript can check for overridden accessors.

Code that’s impacted by the class fields change can get around the issue by converting field initializers to assignments in constructor bodies.

class Base { set data(value: string) { console.log("data changed to " + value); } } class Derived extends Base { constructor() { data = 10; } }

To help mitigate the second issue, you can either add an explicit initializer or add a declare modifier to indicate that a property should have no emit.

interface Animal { animalStuff: any; } interface Dog extends Animal { dogStuff: any; } class AnimalHouse { resident: Animal; constructor(animal: Animal) { this.resident = animal; } } class DogHouse extends AnimalHouse { declare resident: Dog; // ^^^^^^^ // 'resident' now has a 'declare' modifier, // and won't produce any output code. constructor(dog: Dog) { super(dog); } }

Currently useDefineForClassFields is only available when targeting ES5 and upwards, since Object.defineProperty doesn’t exist in ES3. To achieve similar checking for issues, you can create a seperate project that targets ES5 and uses --noEmit to avoid a full build.

For more information, you can take a look at the original pull request for these changes.

We strongly encourage users to try the useDefineForClassFields flag and report back on our issue tracker or in the comments below. This includes feedback on difficulty of adopting the flag so we can understand how we can make migration easier.

Build-Free Editing with Project References

TypeScript’s project references provide us with an easy way to break codebases up to give us faster compiles. Unfortunately, editing a project whose dependencies hadn’t been built (or whose output was out of date) meant that the editing experience wouldn’t work well.

In TypeScript 3.7, when opening a project with dependencies, TypeScript will automatically use the source .ts/.tsx files instead. This means projects using project references will now see an improved editing experience where semantic operations are up-to-date and “just work”. You can disable this behavior with the compiler option disableSourceOfProjectReferenceRedirect which may be appropriate when working in very large projects where this change may impact editing performance.

You can read up more about this change by reading up on its pull request.

Uncalled Function Checks

A common and dangerous error is to forget to invoke a function, especially if the function has zero arguments or is named in a way that implies it might be a property rather than a function.

interface User { isAdministrator(): boolean; notify(): void; doNotDisturb?(): boolean; } // later... // Broken code, do not use! function doAdminThing(user: User) { // oops! if (user.isAdministrator) { sudo(); editTheConfiguration(); } else { throw new AccessDeniedError("User is not an admin"); } }

Here, we forgot to call isAdministrator, and the code incorrectly allows non-adminstrator users to edit the configuration!

In TypeScript 3.7, this is identified as a likely error:

function doAdminThing(user: User) { if (user.isAdministrator) { // ~~~~~~~~~~~~~~~~~~~~ // error! This condition will always return true since the function is always defined. // Did you mean to call it instead?

This check is a breaking change, but for that reason the checks are very conservative. This error is only issued in if conditions, and it is not issued on optional properties, if strictNullChecks is off, or if the function is later called within the body of the if:

interface User { isAdministrator(): boolean; notify(): void; doNotDisturb?(): boolean; } function issueNotification(user: User) { if (user.doNotDisturb) { // OK, property is optional } if (user.notify) { // OK, called the function user.notify(); } }

If you intended to test the function without calling it, you can correct the definition of it to include undefined/null, or use !! to write something like if (!!user.isAdministrator) to indicate that the coercion is intentional.

We owe a big thanks to GitHub user @jwbay who took the initiative to create a proof-of-concept and iterated to provide us with with the current version.

// @ts-nocheck in TypeScript Files

TypeScript 3.7 allows us to add // @ts-nocheck comments to the top of TypeScript files to disable semantic checks. Historically this comment was only respected in JavaScript source files in the presence of checkJs, but we’ve expanded support to TypeScript files to make migrations easier for all users.

Semicolon Formatter Option

TypeScript’s built-in formatter now supports semicolon insertion and removal at locations where a trailing semicolon is optional due to JavaScript’s automatic semicolon insertion (ASI) rules. The setting is available now in Visual Studio Code Insiders, and will be available in Visual Studio 16.4 Preview 2 in the Tools Options menu.

New semicolon formatter option in VS Code

Choosing a value of “insert” or “remove” also affects the format of auto-imports, extracted types, and other generated code provided by TypeScript services. Leaving the setting on its default value of “ignore” makes generated code match the semicolon preference detected in the current file.

3.7 Breaking Changes

DOM Changes

Types in lib.dom.d.ts have been updated. These changes are largely correctness changes related to nullability, but impact will ultimately depend on your codebase.

Class Field Mitigations

As mentioned above, TypeScript 3.7 emits get/set accessors in .d.ts files which can cause breaking changes for consumers on older versions of TypeScript like 3.5 and prior. TypeScript 3.6 users will not be impacted, since that version was future-proofed for this feature.

While not a breakage per se, opting in to the useDefineForClassFields flag can cause breakage when:

  • overriding an accessor in a derived class with a property declaration
  • re-declaring a property declaration with no initializer

To understand the full impact, read the section above on the useDefineForClassFields flag.

Function Truthy Checks

As mentioned above, TypeScript now errors when functions appear to be uncalled within if statement conditions. An error is issued when a function type is checked in if conditions unless any of the following apply:

  • the checked value comes from an optional property
  • strictNullChecks is disabled
  • the function is later called within the body of the if

Local and Imported Type Declarations Now Conflict

Due to a bug, the following construct was previously allowed in TypeScript:

// ./someOtherModule.ts interface SomeType { y: string; } // ./myModule.ts import { SomeType } from "./someOtherModule"; export interface SomeType { x: number; } function fn(arg: SomeType) { console.log(arg.x); // Error! 'x' doesn't exist on 'SomeType' }

Here, SomeType appears to originate in both the import declaration and the local interface declaration. Perhaps surprisingly, inside the module, SomeType refers exclusively to the imported definition, and the local declaration SomeType is only usable when imported from another file. This is very confusing and our review of the very small number of cases of code like this in the wild showed that developers usually thought something different was happening.

In TypeScript 3.7, this is now correctly identified as a duplicate identifier error. The correct fix depends on the original intent of the author and should be addressed on a case-by-case basis. Usually, the naming conflict is unintentional and the best fix is to rename the imported type. If the intent was to augment the imported type, a proper module augmentation should be written instead.

3.7 API Changes

To enable the recursive type alias patterns described above, the typeArguments property has been removed from the TypeReference interface. Users should instead use the getTypeArguments function on TypeChecker instances.

TypeScript 3.6

Stricter Generators

TypeScript 3.6 introduces stricter checking for iterators and generator functions. In earlier versions, users of generators had no way to differentiate whether a value was yielded or returned from a generator.

function* foo() { if (Math.random() < 0.5) yield 100; return "Finished!"; } let iter = foo(); let curr = iter.next(); if (curr.done) { // TypeScript 3.5 and prior thought this was a 'string | number'. // It should know it's 'string' since 'done' was 'true'! curr.value; }

Additionally, generators just assumed the type of yield was always any.

function* bar() { let x: { hello(): void } = yield; x.hello(); } let iter = bar(); iter.next(); iter.next(123); // oops! runtime error!

In TypeScript 3.6, the checker now knows that the correct type for curr.value should be string in our first example, and will correctly error on our call to next() in our last example. This is thanks to some changes in the Iterator and IteratorResult type declarations to include a few new type parameters, and to a new type that TypeScript uses to represent generators called the Generator type.

The Iterator type now allows users to specify the yielded type, the returned type, and the type that next can accept.

interface Iterator<T, TReturn = any, TNext = undefined> { // Takes either 0 or 1 arguments - doesn't accept 'undefined' next(...args: [] | [TNext]): IteratorResult<T, TReturn>; return?(value?: TReturn): IteratorResult<T, TReturn>; throw?(e?: any): IteratorResult<T, TReturn>; }

Building on that work, the new Generator type is an Iterator that always has both the return and throw methods present, and is also iterable.

interface Generator<T = unknown, TReturn = any, TNext = unknown> extends Iterator<T, TReturn, TNext> { next(...args: [] | [TNext]): IteratorResult<T, TReturn>; return(value: TReturn): IteratorResult<T, TReturn>; throw(e: any): IteratorResult<T, TReturn>; [Symbol.iterator](): Generator<T, TReturn, TNext>; }

To allow differentiation between returned values and yielded values, TypeScript 3.6 converts the IteratorResult type to a discriminated union type:

type IteratorResult<T, TReturn = any> = | IteratorYieldResult<T> | IteratorReturnResult<TReturn>; interface IteratorYieldResult<TYield> { done?: false; value: TYield; } interface IteratorReturnResult<TReturn> { done: true; value: TReturn; }

In short, what this means is that you’ll be able to appropriately narrow down values from iterators when dealing with them directly.

To correctly represent the types that can be passed in to a generator from calls to next(), TypeScript 3.6 also infers certain uses of yield within the body of a generator function.

function* foo() { let x: string = yield; console.log(x.toUpperCase()); } let x = foo(); x.next(); // first call to 'next' is always ignored x.next(42); // error! 'number' is not assignable to 'string'

If you’d prefer to be explicit, you can also enforce the type of values that can be returned, yielded, and evaluated from yield expressions using an explicit return type. Below, next() can only be called with booleans, and depending on the value of done, value is either a string or a number.

/** * - yields numbers * - returns strings * - can be passed in booleans */ function* counter(): Generator<number, string, boolean> { let i = 0; while (true) { if (yield i++) { break; } } return "done!"; } var iter = counter(); var curr = iter.next(); while (!curr.done) { console.log(curr.value); curr = iter.next(curr.value === 5); } console.log(curr.value.toUpperCase()); // prints: // // 0 // 1 // 2 // 3 // 4 // 5 // DONE!

For more details on the change, see the pull request here.

More Accurate Array Spread

In pre-ES2015 targets, the most faithful emit for constructs like for/of loops and array spreads can be a bit heavy. For this reason, TypeScript uses a simpler emit by default that only supports array types, and supports iterating on other types using the --downlevelIteration flag. The looser default without --downlevelIteration works fairly well; however, there were some common cases where the transformation of array spreads had observable differences. For example, the following array containing a spread


can be rewritten as the following array literal

[undefined, undefined, undefined, undefined, undefined];

However, TypeScript would instead transform the original code into this code:


which is slightly different. Array(5) produces an array with a length of 5, but with no defined property slots.

TypeScript 3.6 introduces a new __spreadArrays helper to accurately model what happens in ECMAScript 2015 in older targets outside of --downlevelIteration. __spreadArrays is also available in tslib.

For more information, see the relevant pull request.

Improved UX Around Promises

TypeScript 3.6 introduces some improvements for when Promises are mis-handled.

For example, it’s often very common to forget to .then() or await the contents of a Promise before passing it to another function. TypeScript’s error messages are now specialized, and inform the user that perhaps they should consider using the await keyword.

interface User { name: string; age: number; location: string; } declare function getUserData(): Promise<User>; declare function displayUser(user: User): void; async function f() { displayUser(getUserData()); // ~~~~~~~~~~~~~ // Argument of type 'Promise<User>' is not assignable to parameter of type 'User'. // ... // Did you forget to use 'await'? }

It’s also common to try to access a method before await-ing or .then()-ing a Promise. This is another example, among many others, where we’re able to do better.

async function getCuteAnimals() { fetch("https://reddit.com/r/aww.json").json(); // ~~~~ // Property 'json' does not exist on type 'Promise<Response>'. // // Did you forget to use 'await'? }

For more details, see the originating issue, as well as the pull requests that link back to it.

Better Unicode Support for Identifiers

TypeScript 3.6 contains better support for Unicode characters in identifiers when emitting to ES2015 and later targets.

const 𝓱𝓮𝓵𝓵𝓸 = "world"; // previously disallowed, now allowed in '--target es2015'

import.meta Support in SystemJS

TypeScript 3.6 supports transforming import.meta to context.meta when your module target is set to system.

// This module: console.log(import.meta.url); // gets turned into the following: System.register([], function (exports, context) { return { setters: [], execute: function () { console.log(context.meta.url); }, }; });

get and set Accessors Are Allowed in Ambient Contexts

In previous versions of TypeScript, the language didn’t allow get and set accessors in ambient contexts (like in declare-d classes, or in .d.ts files in general). The rationale was that accessors weren’t distinct from properties as far as writing and reading to these properties; however, because ECMAScript’s class fields proposal may have differing behavior from in existing versions of TypeScript, we realized we needed a way to communicate this different behavior to provide appropriate errors in subclasses.

As a result, users can write getters and setters in ambient contexts in TypeScript 3.6.

declare class Foo { // Allowed in 3.6+. get x(): number; set x(val: number); }

In TypeScript 3.7, the compiler itself will take advantage of this feature so that generated .d.ts files will also emit get/set accessors.

Ambient Classes and Functions Can Merge

In previous versions of TypeScript, it was an error to merge classes and functions under any circumstances. Now, ambient classes and functions (classes/functions with the declare modifier, or in .d.ts files) can merge. This means that now you can write the following:

export declare function Point2D(x: number, y: number): Point2D; export declare class Point2D { x: number; y: number; constructor(x: number, y: number); }

instead of needing to use

export interface Point2D { x: number; y: number; } export declare var Point2D: { (x: number, y: number): Point2D; new (x: number, y: number): Point2D; };

One advantage of this is that the callable constructor pattern can be easily expressed while also allowing namespaces to merge with these declarations (since var declarations can’t merge with namespaces).

In TypeScript 3.7, the compiler will take advantage of this feature so that .d.ts files generated from .js files can appropriately capture both the callability and constructability of a class-like function.

For more details, see the original PR on GitHub.

APIs to Support --build and --incremental

TypeScript 3.0 introduced support for referencing other and building them incrementally using the --build flag. Additionally, TypeScript 3.4 introduced the --incremental flag for saving information about previous compilations to only rebuild certain files. These flags were incredibly useful for structuring projects more flexibly and speeding builds up. Unfortunately, using these flags didn’t work with 3rd party build tools like Gulp and Webpack. TypeScript 3.6 now exposes two sets of APIs to operate on project references and incremental program building.

For creating --incremental builds, users can leverage the createIncrementalProgram and createIncrementalCompilerHost APIs. Users can also re-hydrate old program instances from .tsbuildinfo files generated by this API using the newly exposed readBuilderProgram function, which is only meant to be used as for creating new programs (i.e. you can’t modify the returned instance - it’s only meant to be used for the oldProgram parameter in other create*Program functions).

For leveraging project references, a new createSolutionBuilder function has been exposed, which returns an instance of the new type SolutionBuilder.

For more details on these APIs, you can see the original pull request.

Semicolon-Aware Code Edits

Editors like Visual Studio and Visual Studio Code can automatically apply quick fixes, refactorings, and other transformations like automatically importing values from other modules. These transformations are powered by TypeScript, and older versions of TypeScript unconditionally added semicolons to the end of every statement; unfortunately, this disagreed with many users’ style guidelines, and many users were displeased with the editor inserting semicolons.

TypeScript is now smart enough to detect whether your file uses semicolons when applying these sorts of edits. If your file generally lacks semicolons, TypeScript won’t add one.

For more details, see the corresponding pull request.

Smarter Auto-Import Syntax

JavaScript has a lot of different module syntaxes or conventions: the one in the ECMAScript standard, the one Node already supports (CommonJS), AMD, System.js, and more! For the most part, TypeScript would default to auto-importing using ECMAScript module syntax, which was often inappropriate in certain TypeScript projects with different compiler settings, or in Node projects with plain JavaScript and require calls.

TypeScript 3.6 is now a bit smarter about looking at your existing imports before deciding on how to auto-import other modules. You can see more details in the original pull request here.

await Completions on Promises

New TypeScript Playground

The TypeScript playground has received a much-needed refresh with handy new functionality! The new playground is largely a fork of Artem Tyurin’s TypeScript playground which community members have been using more and more. We owe Artem a big thanks for helping out here!

The new playground now supports many new options including:

  • The target option (allowing users to switch out of es5 to es3, es2015, esnext, etc.)
  • All the strictness flags (including just strict)
  • Support for plain JavaScript files (using allowJS and optionally checkJs)

These options also persist when sharing links to playground samples, allowing users to more reliably share examples without having to tell the recipient “oh, don’t forget to turn on the noImplicitAny option!“.

In the near future, we’re going to be refreshing the playground samples, adding JSX support, and polishing automatic type acquisition, meaning that you’ll be able to see the same experience on the playground as you’d get in your personal editor.

As we improve the playground and the website, we welcome feedback and pull requests on GitHub!

TypeScript 3.5

Speed improvements

TypeScript 3.5 introduces several optimizations around type-checking and incremental builds.

Type-checking speed-ups

TypeScript 3.5 contains certain optimizations over TypeScript 3.4 for type-checking more efficiently. These improvements are significantly more pronounced in editor scenarios where type-checking drives operations like code completion lists.

--incremental improvements

TypeScript 3.5 improves on 3.4’s --incremental build mode, by saving information about how the state of the world was calculated - compiler settings, why files were looked up, where files were found, etc. In scenarios involving hundreds of projects using TypeScript’s project references in --build mode, we’ve found that the amount of time rebuilding can be reduced by as much as 68% compared to TypeScript 3.4!

For more details, you can see the pull requests to

The Omit helper type

TypeScript 3.5 introduces the new Omit helper type, which creates a new type with some properties dropped from the original.

type Person = { name: string; age: number; location: string; }; type QuantumPerson = Omit<Person, "location">; // equivalent to type QuantumPerson = { name: string; age: number; };

Here we were able to copy over all the properties of Person except for location using the Omit helper.

For more details, see the pull request on GitHub to add Omit, as well as the change to use Omit for object rest.

Improved excess property checks in union types

In TypeScript 3.4 and earlier, certain excess properties were allowed in situations where they really shouldn’t have been. For instance, TypeScript 3.4 permitted the incorrect name property in the object literal even though its types don’t match between Point and Label.

type Point = { x: number; y: number; }; type Label = { name: string; }; const thing: Point | Label = { x: 0, y: 0, name: true, // uh-oh! };

Previously, a non-disciminated union wouldn’t have any excess property checking done on its members, and as a result, the incorrectly typed name property slipped by.

In TypeScript 3.5, the type-checker at least verifies that all the provided properties belong to some union member and have the appropriate type, meaning that the sample above correctly issues an error.

Note that partial overlap is still permitted as long as the property types are valid.

const pl: Point | Label = { x: 0, y: 0, name: "origin", // okay };

The --allowUmdGlobalAccess flag

In TypeScript 3.5, you can now reference UMD global declarations like

export as namespace foo;

from anywhere - even modules - using the new --allowUmdGlobalAccess flag.

This mode adds flexibility for mixing and matching the way 3rd party libraries, where globals that libraries declare can always be consumed, even from within modules.

For more details, see the pull request on GitHub.

Smarter union type checking

In TypeScript 3.4 and prior, the following example would fail:

type S = { done: boolean; value: number }; type T = { done: false; value: number } | { done: true; value: number }; declare let source: S; declare let target: T; target = source;

That’s because S isn’t assignable to { done: false, value: number } nor { done: true, value: number }. Why? Because the done property in S isn’t specific enough - it’s boolean whereas each constituent of T has a done property that’s specifically true or false. That’s what we meant by each constituent type being checked in isolation: TypeScript doesn’t just union each property together and see if S is assignable to that. If it did, some bad code could get through like the following:

interface Foo { kind: "foo"; value: string; } interface Bar { kind: "bar"; value: number; } function doSomething(x: Foo | Bar) { if (x.kind === "foo") { x.value.toLowerCase(); } } // uh-oh - luckily TypeScript errors here! doSomething({ kind: "foo", value: 123, });

However, this was a bit overly strict for the original example. If you figure out the precise type of any possible value of S, you can actually see that it matches the types in T exactly.

In TypeScript 3.5, when assigning to types with discriminant properties like in T, the language actually will go further and decompose types like S into a union of every possible inhabitant type. In this case, since boolean is a union of true and false, S will be viewed as a union of { done: false, value: number } and { done: true, value: number }.

For more details, you can see the original pull request on GitHub.

Higher order type inference from generic constructors

In TypeScript 3.4, we improved inference for when generic functions that return functions like so:

function compose<T, U, V>(f: (x: T) => U, g: (y: U) => V): (x: T) => V { return (x) => g(f(x)); }

took other generic functions as arguments, like so:

function arrayify<T>(x: T): T[] { return [x]; } type Box<U> = { value: U }; function boxify<U>(y: U): Box<U> { return { value: y }; } let newFn = compose(arrayify, boxify);

Instead of a relatively useless type like (x: {}) => Box<{}[]>, which older versions of the language would infer, TypeScript 3.4’s inference allows newFn to be generic. Its new type is <T>(x: T) => Box<T[]>.

TypeScript 3.5 generalizes this behavior to work on constructor functions as well.

class Box<T> { kind: "box"; value: T; constructor(value: T) { this.value = value; } } class Bag<U> { kind: "bag"; value: U; constructor(value: U) { this.value = value; } } function composeCtor<T, U, V>( F: new (x: T) => U, G: new (y: U) => V ): (x: T) => V { return (x) => new G(new F(x)); } let f = composeCtor(Box, Bag); // has type '<T>(x: T) => Bag<Box<T>>' let a = f(1024); // has type 'Bag<Box<number>>'

In addition to compositional patterns like the above, this new inference on generic constructors means that functions that operate on class components in certain UI libraries like React can more correctly operate on generic class components.

type ComponentClass<P> = new (props: P) => Component<P>; declare class Component<P> { props: P; constructor(props: P); } declare function myHoc<P>(C: ComponentClass<P>): ComponentClass<P>; type NestedProps<T> = { foo: number; stuff: T }; declare class GenericComponent<T> extends Component<NestedProps<T>> {} // type is 'new <T>(props: NestedProps<T>) => Component<NestedProps<T>>' const GenericComponent2 = myHoc(GenericComponent);

To learn more, check out the original pull request on GitHub.

TypeScript 3.4

Faster subsequent builds with the --incremental flag

TypeScript 3.4 introduces a new flag called --incremental which tells TypeScript to save information about the project graph from the last compilation. The next time TypeScript is invoked with --incremental, it will use that information to detect the least costly way to type-check and emit changes to your project.

// tsconfig.json { compilerOptions: { incremental: true, outDir: "./lib", }, include: ["./src"], }

By default with these settings, when we run tsc, TypeScript will look for a file called .tsbuildinfo in the output directory (./lib). If ./lib/.tsbuildinfo doesn’t exist, it’ll be generated. But if it does, tsc will try to use that file to incrementally type-check and update our output files.

These .tsbuildinfo files can be safely deleted and don’t have any impact on our code at runtime - they’re purely used to make compilations faster. We can also name them anything that we want, and place them anywhere we want using the --tsBuildInfoFile flag.

// front-end.tsconfig.json { compilerOptions: { incremental: true, tsBuildInfoFile: "./buildcache/front-end", outDir: "./lib", }, include: ["./src"], }

Composite projects

Part of the intent with composite projects (tsconfig.jsons with composite set to true) is that references between different projects can be built incrementally. As such, composite projects will always produce .tsbuildinfo files.


When outFile is used, the build information file’s name will be based on the output file’s name. As an example, if our output JavaScript file is ./output/foo.js, then under the --incremental flag, TypeScript will generate the file ./output/foo.tsbuildinfo. As above, this can be controlled with the --tsBuildInfoFile flag.

Higher order type inference from generic functions

TypeScript 3.4 can now produce generic function types when inference from other generic functions produces free type variables for inferences. This means many function composition patterns now work better in 3.4.

To get more specific, let’s build up some motivation and consider the following compose function:

function compose<A, B, C>(f: (arg: A) => B, g: (arg: B) => C): (arg: A) => C { return (x) => g(f(x)); }

compose takes two other functions:

  • f which takes some argument (of type A) and returns a value of type B
  • g which takes an argument of type B (the type f returned), and returns a value of type C

compose then returns a function which feeds its argument through f and then g.

When calling this function, TypeScript will try to figure out the types of A, B, and C through a process called type argument inference. This inference process usually works pretty well:

interface Person { name: string; age: number; } function getDisplayName(p: Person) { return p.name.toLowerCase(); } function getLength(s: string) { return s.length; } // has type '(p: Person) => number' const getDisplayNameLength = compose(getDisplayName, getLength); // works and returns the type 'number' getDisplayNameLength({ name: "Person McPersonface", age: 42 });

The inference process is fairly straightforward here because getDisplayName and getLength use types that can easily be referenced. However, in TypeScript 3.3 and earlier, generic functions like compose didn’t work so well when passed other generic functions.

interface Box<T> { value: T; } function makeArray<T>(x: T): T[] { return [x]; } function makeBox<U>(value: U): Box<U> { return { value }; } // has type '(arg: {}) => Box<{}[]>' const makeBoxedArray = compose(makeArray, makeBox); makeBoxedArray("hello!").value[0].toUpperCase(); // ~~~~~~~~~~~ // error: Property 'toUpperCase' does not exist on type '{}'.

In older versions, TypeScript would infer the empty object type ({}) when inferring from other type variables like T and U.

During type argument inference in TypeScript 3.4, for a call to a generic function that returns a function type, TypeScript will, as appropriate, propagate type parameters from generic function arguments onto the resulting function type.

In other words, instead of producing the type

(arg: {}) => Box<{}[]>

TypeScript 3.4 produces the type

<T>(arg: T) => Box<T[]>

Notice that T has been propagated from makeArray into the resulting type’s type parameter list. This means that genericity from compose’s arguments has been preserved and our makeBoxedArray sample will just work!

interface Box<T> { value: T; } function makeArray<T>(x: T): T[] { return [x]; } function makeBox<U>(value: U): Box<U> { return { value }; } // has type '<T>(arg: T) => Box<T[]>' const makeBoxedArray = compose(makeArray, makeBox); // works with no problem! makeBoxedArray("hello!").value[0].toUpperCase();

For more details, you can read more at the original change.

Improvements for ReadonlyArray and readonly tuples

TypeScript 3.4 makes it a little bit easier to use read-only array-like types.

A new syntax for ReadonlyArray

The ReadonlyArray type describes Arrays that can only be read from. Any variable with a reference to a ReadonlyArray can’t add, remove, or replace any elements of the array.

function foo(arr: ReadonlyArray<string>) { arr.slice(); // okay arr.push("hello!"); // error! }

While it’s good practice to use ReadonlyArray over Array when no mutation is intended, it’s often been a pain given that arrays have a nicer syntax. Specifically, number[] is a shorthand version of Array<number>, just as Date[] is a shorthand for Array<Date>.

TypeScript 3.4 introduces a new syntax for ReadonlyArray using a new readonly modifier for array types.

function foo(arr: readonly string[]) { arr.slice(); // okay arr.push("hello!"); // error! }

readonly tuples

TypeScript 3.4 also introduces new support for readonly tuples. We can prefix any tuple type with the readonly keyword to make it a readonly tuple, much like we now can with array shorthand syntax. As you might expect, unlike ordinary tuples whose slots could be written to, readonly tuples only permit reading from those positions.

function foo(pair: readonly [string, string]) { console.log(pair[0]); // okay pair[1] = "hello!"; // error }

The same way that ordinary tuples are types that extend from Array - a tuple with elements of type T1, T2, … Tn extends from Array< T1 | T2 | … Tn > - readonly tuples are types that extend from ReadonlyArray. So a readonly tuple with elements T1, T2, … Tn extends from ReadonlyArray< T1 | T2 | … Tn >.

readonly mapped type modifiers and readonly arrays

In earlier versions of TypeScript, we generalized mapped types to operate differently on array-like types. This meant that a mapped type like Boxify could work on arrays and tuples alike.

interface Box<T> { value: T; } type Boxify<T> = { [K in keyof T]: Box<T[K]>; }; // { a: Box<string>, b: Box<number> } type A = Boxify<{ a: string; b: number }>; // Array<Box<number>> type B = Boxify<number[]>; // [Box<string>, Box<number>] type C = Boxify<[string, boolean]>;

Unfortunately, mapped types like the Readonly utility type were effectively no-ops on array and tuple types.

// lib.d.ts type Readonly<T> = { readonly [K in keyof T]: T[K]; }; // How code acted *before* TypeScript 3.4 // { readonly a: string, readonly b: number } type A = Readonly<{ a: string; b: number }>; // number[] type B = Readonly<number[]>; // [string, boolean] type C = Readonly<[string, boolean]>;

In TypeScript 3.4, the readonly modifier in a mapped type will automatically convert array-like types to their corresponding readonly counterparts.

// How code acts now *with* TypeScript 3.4 // { readonly a: string, readonly b: number } type A = Readonly<{ a: string; b: number }>; // readonly number[] type B = Readonly<number[]>; // readonly [string, boolean] type C = Readonly<[string, boolean]>;

Similarly, you could write a utility type like Writable mapped type that strips away readonly-ness, and that would convert readonly array containers back to their mutable equivalents.

type Writable<T> = { -readonly [K in keyof T]: T[K]; }; // { a: string, b: number } type A = Writable<{ readonly a: string; readonly b: number; }>; // number[] type B = Writable<readonly number[]>; // [string, boolean] type C = Writable<readonly [string, boolean]>;


Despite its appearance, the readonly type modifier can only be used for syntax on array types and tuple types. It is not a general-purpose type operator.

let err1: readonly Set<number>; // error! let err2: readonly Array<boolean>; // error! let okay: readonly boolean[]; // works fine

You can see more details in the pull request.

const assertions

TypeScript 3.4 introduces a new construct for literal values called const assertions. Its syntax is a type assertion with const in place of the type name (e.g. 123 as const). When we construct new literal expressions with const assertions, we can signal to the language that

  • no literal types in that expression should be widened (e.g. no going from "hello" to string)
  • object literals get readonly properties
  • array literals become readonly tuples
// Type '"hello"' let x = "hello" as const; // Type 'readonly [10, 20]' let y = [10, 20] as const; // Type '{ readonly text: "hello" }' let z = { text: "hello" } as const;

Outside of .tsx files, the angle bracket assertion syntax can also be used.

// Type '"hello"' let x = <const>"hello"; // Type 'readonly [10, 20]' let y = <const>[10, 20]; // Type '{ readonly text: "hello" }' let z = <const>{ text: "hello" };

This feature means that types that would otherwise be used just to hint immutability to the compiler can often be omitted.

// Works with no types referenced or declared. // We only needed a single const assertion. function getShapes() { let result = [ { kind: "circle", radius: 100 }, { kind: "square", sideLength: 50 }, ] as const; return result; } for (const shape of getShapes()) { // Narrows perfectly! if (shape.kind === "circle") { console.log("Circle radius", shape.radius); } else { console.log("Square side length", shape.sideLength); } }

Notice the above needed no type annotations. The const assertion allowed TypeScript to take the most specific type of the expression.

This can even be used to enable enum-like patterns in plain JavaScript code if you choose not to use TypeScript’s enum construct.

export const Colors = { red: "RED", blue: "BLUE", green: "GREEN", } as const; // or use an 'export default' export default { red: "RED", blue: "BLUE", green: "GREEN", } as const;


One thing to note is that const assertions can only be applied immediately on simple literal expressions.

// Error! A 'const' assertion can only be applied to a // to a string, number, boolean, array, or object literal. let a = (Math.random() < 0.5 ? 0 : 1) as const; // Works! let b = Math.random() < 0.5 ? (0 as const) : (1 as const);

Another thing to keep in mind is that const contexts don’t immediately convert an expression to be fully immutable.

let arr = [1, 2, 3, 4]; let foo = { name: "foo", contents: arr, } as const; foo.name = "bar"; // error! foo.contents = []; // error! foo.contents.push(5); // ...works!

For more details, you can check out the respective pull request.

Type-checking for globalThis

TypeScript 3.4 introduces support for type-checking ECMAScript’s new globalThis - a global variable that, well, refers to the global scope. Unlike the above solutions, globalThis provides a standard way for accessing the global scope which can be used across different environments.

// in a global file: var abc = 100; // Refers to 'abc' from above. globalThis.abc = 200;

Note that global variables declared with let and const don’t show up on globalThis.

let answer = 42; // error! Property 'answer' does not exist on 'typeof globalThis'. globalThis.answer = 333333;

It’s also important to note that TypeScript doesn’t transform references to globalThis when compiling to older versions of ECMAScript. As such, unless you’re targeting evergreen browsers (which already support globalThis), you may want to use an appropriate polyfill instead.

For more details on the implementation, see the feature’s pull request.

TypeScript 3.3

Improved behavior for calling union types

In prior versions of TypeScript, unions of callable types could only be invoked if they had identical parameter lists.

type Fruit = "apple" | "orange"; type Color = "red" | "orange"; type FruitEater = (fruit: Fruit) => number; // eats and ranks the fruit type ColorConsumer = (color: Color) => string; // consumes and describes the colors declare let f: FruitEater | ColorConsumer; // Cannot invoke an expression whose type lacks a call signature. // Type 'FruitEater | ColorConsumer' has no compatible call signatures.ts(2349) f("orange");

However, in the above example, both FruitEaters and ColorConsumers should be able to take the string "orange", and return either a number or a string.

In TypeScript 3.3, this is no longer an error.

type Fruit = "apple" | "orange"; type Color = "red" | "orange"; type FruitEater = (fruit: Fruit) => number; // eats and ranks the fruit type ColorConsumer = (color: Color) => string; // consumes and describes the colors declare let f: FruitEater | ColorConsumer; f("orange"); // It works! Returns a 'number | string'. f("apple"); // error - Argument of type '"red"' is not assignable to parameter of type '"orange"'. f("red"); // error - Argument of type '"red"' is not assignable to parameter of type '"orange"'.

In TypeScript 3.3, the parameters of these signatures are intersected together to create a new signature.

In the example above, the parameters fruit and color are intersected together to a new parameter of type Fruit & Color. Fruit & Color is really the same as ("apple" | "orange") & ("red" | "orange") which is equivalent to ("apple" & "red") | ("apple" & "orange") | ("orange" & "red") | ("orange" & "orange"). Each of those impossible intersections reduces to never, and we’re left with "orange" & "orange" which is just "orange".


This new behavior only kicks in when at most one type in the union has multiple overloads, and at most one type in the union has a generic signature. That means methods on number[] | string[] like map (which is generic) still won’t be callable.

On the other hand, methods like forEach will now be callable, but under noImplicitAny there may be some issues.

interface Dog { kind: "dog"; dogProp: any; } interface Cat { kind: "cat"; catProp: any; } const catOrDogArray: Dog[] | Cat[] = []; catOrDogArray.forEach((animal) => { // ~~~~~~ error! // Parameter 'animal' implicitly has an 'any' type. });

This is still strictly more capable in TypeScript 3.3, and adding an explicit type annotation will work.

interface Dog { kind: "dog"; dogProp: any; } interface Cat { kind: "cat"; catProp: any; } const catOrDogArray: Dog[] | Cat[] = []; catOrDogArray.forEach((animal: Dog | Cat) => { if (animal.kind === "dog") { animal.dogProp; // ... } else if (animal.kind === "cat") { animal.catProp; // ... } });

Incremental file watching for composite projects in --build --watch

TypeScript 3.0 introduced a new feature for structuring builds called “composite projects”. Part of the goal here was to ensure users could break up large projects into smaller parts that build quickly and preserve project structure, without compromising the existing TypeScript experience. Thanks to composite projects, TypeScript can use --build mode to recompile only the set of projects and dependencies. You can think of this as optimizing inter-project builds.

TypeScript 2.7 also introduced --watch mode builds via a new incremental “builder” API. In a similar vein, the entire idea is that this mode only re-checks and re-emits changed files or files whose dependencies might impact type-checking. You can think of this as optimizing intra-project builds.

Prior to 3.3, building composite projects using --build --watch actually didn’t use this incremental file watching infrastructure. An update in one project under --build --watch mode would force a full build of that project, rather than determining which files within that project were affected.

In TypeScript 3.3, --build mode’s --watch flag does leverage incremental file watching as well. That can mean signficantly faster builds under --build --watch. In our testing, this functionality has resulted in a reduction of 50% to 75% in build times of the original --build --watch times. You can read more on the original pull request for the change to see specific numbers, but we believe most composite project users will see significant wins here.

TypeScript 3.2


TypeScript 3.2 introduces a new --strictBindCallApply compiler option (in the --strict family of options) with which the bind, call, and apply methods on function objects are strongly typed and strictly checked.

function foo(a: number, b: string): string { return a + b; } let a = foo.apply(undefined, [10]); // error: too few argumnts let b = foo.apply(undefined, [10, 20]); // error: 2nd argument is a number let c = foo.apply(undefined, [10, "hello", 30]); // error: too many arguments let d = foo.apply(undefined, [10, "hello"]); // okay! returns a string

This is achieved by introducing two new types, CallableFunction and NewableFunction, in lib.d.ts. These types contain specialized generic method declarations for bind, call, and apply for regular functions and constructor functions, respectively. The declarations use generic rest parameters (see #24897) to capture and reflect parameter lists in a strongly typed manner. In --strictBindCallApply mode these declarations are used in place of the (very permissive) declarations provided by type Function.


Since the stricter checks may uncover previously unreported errors, this is a breaking change in --strict mode.

Additionally, another caveat of this new functionality is that due to certain limitations, bind, call, and apply can’t yet fully model generic functions or functions that have overloads. When using these methods on a generic function, type parameters will be substituted with the empty object type ({}), and when used on a function with overloads, only the last overload will ever be modeled.

Generic spread expressions in object literals

In TypeScript 3.2, object literals now allow generic spread expressions which now produce intersection types, similar to the Object.assign function and JSX literals. For example:

function taggedObject<T, U extends string>(obj: T, tag: U) { return { ...obj, tag }; // T & { tag: U } } let x = taggedObject({ x: 10, y: 20 }, "point"); // { x: number, y: number } & { tag: "point" }

Property assignments and non-generic spread expressions are merged to the greatest extent possible on either side of a generic spread expression. For example:

function foo1<T>(t: T, obj1: { a: string }, obj2: { b: string }) { return { ...obj1, x: 1, ...t, ...obj2, y: 2 }; // { a: string, x: number } & T & { b: string, y: number } }

Non-generic spread expressions continue to be processed as before: Call and construct signatures are stripped, only non-method properties are preserved, and for properties with the same name, the type of the rightmost property is used. This contrasts with intersection types which concatenate call and construct signatures, preserve all properties, and intersect the types of properties with the same name. Thus, spreads of the same types may produce different results when they are created through instantiation of generic types:

function spread<T, U>(t: T, u: U) { return { ...t, ...u }; // T & U } declare let x: { a: string; b: number }; declare let y: { b: string; c: boolean }; let s1 = { ...x, ...y }; // { a: string, b: string, c: boolean } let s2 = spread(x, y); // { a: string, b: number } & { b: string, c: boolean } let b1 = s1.b; // string let b2 = s2.b; // number & string

Generic object rest variables and parameters

TypeScript 3.2 also allows destructuring a rest binding from a generic variable. This is achieved by using the predefined Pick and Exclude helper types from lib.d.ts, and using the generic type in question as well as the names of the other bindings in the destructuring pattern.

function excludeTag<T extends { tag: string }>(obj: T) { let { tag, ...rest } = obj; return rest; // Pick<T, Exclude<keyof T, "tag">> } const taggedPoint = { x: 10, y: 20, tag: "point" }; const point = excludeTag(taggedPoint); // { x: number, y: number }


BigInts are part of an upcoming proposal in ECMAScript that allow us to model theoretically arbitrarily large integers. TypeScript 3.2 brings type-checking for BigInts, as well as support for emitting BigInt literals when targeting esnext.

BigInt support in TypeScript introduces a new primitive type called the bigint (all lowercase). You can get a bigint by calling the BigInt() function or by writing out a BigInt literal by adding an n to the end of any integer numeric literal:

let foo: bigint = BigInt(100); // the BigInt function let bar: bigint = 100n; // a BigInt literal // *Slaps roof of fibonacci function* // This bad boy returns ints that can get *so* big! function fibonacci(n: bigint) { let result = 1n; for (let last = 0n, i = 0n; i < n; i++) { const current = result; result += last; last = current; } return result; } fibonacci(10000n);

While you might imagine close interaction between number and bigint, the two are separate domains.

declare let foo: number; declare let bar: bigint; foo = bar; // error: Type 'bigint' is not assignable to type 'number'. bar = foo; // error: Type 'number' is not assignable to type 'bigint'.

As specified in ECMAScript, mixing numbers and bigints in arithmetic operations is an error. You’ll have to explicitly convert values to BigInts.

console.log(3.141592 * 10000n); // error console.log(3145 * 10n); // error console.log(BigInt(3145) * 10n); // okay!

Also important to note is that bigints produce a new string when using the typeof operator: the string "bigint". Thus, TypeScript correctly narrows using typeof as you’d expect.

function whatKindOfNumberIsIt(x: number | bigint) { if (typeof x === "bigint") { console.log("'x' is a bigint!"); } else { console.log("'x' is a floating-point number"); } }

We’d like to extend a huge thanks to Caleb Sander for all the work on this feature. We’re grateful for the contribution, and we’re sure our users are too!


As we mentioned, BigInt support is only available for the esnext target. It may not be obvious, but because BigInts have different behavior for mathematical operators like +, -, *, etc., providing functionality for older targets where the feature doesn’t exist (like es2017 and below) would involve rewriting each of these operations. TypeScript would need to dispatch to the correct behavior depending on the type, and so every addition, string concatenation, multiplication, etc. would involve a function call.

For that reason, we have no immediate plans to provide downleveling support. On the bright side, Node 11 and newer versions of Chrome already support this feature, so you’ll be able to use BigInts there when targeting esnext.

Certain targets may include a polyfill or BigInt-like runtime object. For those purposes you may want to add esnext.bigint to the lib setting in your compiler options.

Non-unit types as union discriminants

TypeScript 3.2 makes narrowing easier by relaxing rules for what it considers a discriminant property. Common properties of unions are now considered discriminants as long as they contain some singleton type (e.g. a string literal, null, or undefined), and they contain no generics.

As a result, TypeScript 3.2 considers the error property in the following example to be a discriminant, whereas before it wouldn’t since Error isn’t a singleton type. Thanks to this, narrowing works correctly in the body of the unwrap function.

type Result<T> = { error: Error; data: null } | { error: null; data: T }; function unwrap<T>(result: Result<T>) { if (result.error) { // Here 'error' is non-null throw result.error; } // Now 'data' is non-null return result.data; }

tsconfig.json inheritance via Node.js packages

TypeScript 3.2 now resolves tsconfig.jsons from node_modules. When using a bare path for the "extends" field in tsconfig.json, TypeScript will dive into node_modules packages for us.

{ "extends": "@my-team/tsconfig-base", "include": ["./**/*"] "compilerOptions": { // Override certain options on a project-by-project basis. "strictBindCallApply": false, } }

Here, TypeScript will climb up node_modules folders looking for a @my-team/tsconfig-base package. For each of those packages, TypeScript will first check whether package.json contains a "tsconfig" field, and if it does, TypeScript will try to load a configuration file from that field. If neither exists, TypeScript will try to read from a tsconfig.json at the root. This is similar to the lookup process for .js files in packages that Node uses, and the .d.ts lookup process that TypeScript already uses.

This feature can be extremely useful for bigger organizations, or projects with lots of distributed dependencies.

The new --showConfig flag

tsc, the TypeScript compiler, supports a new flag called --showConfig. When running tsc --showConfig, TypeScript will calculate the effective tsconfig.json (after calculating options inherited from the extends field) and print that out. This can be useful for diagnosing configuration issues in general.

Object.defineProperty declarations in JavaScript

When writing in JavaScript files (using allowJs), TypeScript now recognizes declarations that use Object.defineProperty. This means you’ll get better completions, and stronger type-checking when enabling type-checking in JavaScript files (by turning on the checkJs option or adding a // @ts-check comment to the top of your file).

// @ts-check let obj = {}; Object.defineProperty(obj, "x", { value: "hello", writable: false }); obj.x.toLowercase(); // ~~~~~~~~~~~ // error: // Property 'toLowercase' does not exist on type 'string'. // Did you mean 'toLowerCase'? obj.x = "world"; // ~ // error: // Cannot assign to 'x' because it is a read-only property.

TypeScript 3.1

Mapped types on tuples and arrays

In TypeScript 3.1, mapped object types[1] over tuples and arrays now produce new tuples/arrays, rather than creating a new type where members like push(), pop(), and length are converted. For example:

type MapToPromise<T> = { [K in keyof T]: Promise<T[K]> }; type Coordinate = [number, number]; type PromiseCoordinate = MapToPromise<Coordinate>; // [Promise<number>, Promise<number>]

MapToPromise takes a type T, and when that type is a tuple like Coordinate, only the numeric properties are converted. In [number, number], there are two numerically named properties: 0 and 1. When given a tuple like that, MapToPromise will create a new tuple where the 0 and 1 properties are Promises of the original type. So the resulting type PromiseCoordinate ends up with the type [Promise<number>, Promise<number>].

Properties declarations on functions

TypeScript 3.1 brings the ability to define properties on function declarations and const-declared functons, simply by assigning to properties on these functions in the same scope. This allows us to write canonical JavaScript code without resorting to namespace hacks. For example:

function readImage(path: string, callback: (err: any, image: Image) => void) { // ... } readImage.sync = (path: string) => { const contents = fs.readFileSync(path); return decodeImageSync(contents); };

Here, we have a function readImage which reads an image in a non-blocking asynchronous way. In addition to readImage, we’ve provided a convenience function on readImage itself called readImage.sync.

While ECMAScript exports are often a better way of providing this functionality, this new support allows code written in this style to “just work” TypeScript. Additionaly, this approach for property declarations allows us to express common patterns like defaultProps and propTypes on React stateless function components (SFCs).

export const FooComponent = ({ name }) => <div>Hello! I am {name}</div>; FooComponent.defaultProps = { name: "(anonymous)", };

[1] More specifically, homomorphic mapped types like in the above form.

Version selection with typesVersions

Feedback from our community, as well as our own experience, has shown us that leveraging the newest TypeScript features while also accomodating users on the older versions are difficult. TypeScript introduces a new feature called typesVersions to help accomodate these scenarios.

When using Node module resolution in TypeScript 3.1, when TypeScript cracks open a package.json file to figure out which files it needs to read, it first looks at a new field called typesVersions. A package.json with a typesVersions field might look like this:

{ "name": "package-name", "version": "1.0", "types": "./index.d.ts", "typesVersions": { ">=3.1": { "*": ["ts3.1/*"] } } }

This package.json tells TypeScript to check whether the current version of TypeScript is running. If it’s 3.1 or later, it figures out the path you’ve imported relative to the package, and reads from the package’s ts3.1 folder. That’s what that { "*": ["ts3.1/*"] } means - if you’re familiar with path mapping today, it works exactly like that.

So in the above example, if we’re importing from "package-name", we’ll try to resolve from [...]/node_modules/package-name/ts3.1/index.d.ts (and other relevant paths) when running in TypeScript 3.1. If we import from package-name/foo, we’ll try to look for [...]/node_modules/package-name/ts3.1/foo.d.ts and [...]/node_modules/package-name/ts3.1/foo/index.d.ts.

What if we’re not running in TypeScript 3.1 in this example? Well, if none of the fields in typesVersions get matched, TypeScript falls back to the types field, so here TypeScript 3.0 and earlier will be redirected to [...]/node_modules/package-name/index.d.ts.

Matching behavior

The way that TypeScript decides on whether a version of the compiler & language matches is by using Node’s semver ranges.

Multiple fields

typesVersions can support multiple fields where each field name is specified by the range to match on.

{ "name": "package-name", "version": "1.0", "types": "./index.d.ts", "typesVersions": { ">=3.2": { "*": ["ts3.2/*"] }, ">=3.1": { "*": ["ts3.1/*"] } } }

Since ranges have the potential to overlap, determining which redirect applies is order-specific. That means in the above example, even though both the >=3.2 and the >=3.1 matchers support TypeScript 3.2 and above, reversing the order could have different behavior, so the above sample would not be equivalent to the following.

{ name: "package-name", version: "1.0", types: "./index.d.ts", typesVersions: { // NOTE: this doesn't work! ">=3.1": { "*": ["ts3.1/*"] }, ">=3.2": { "*": ["ts3.2/*"] }, }, }

TypeScript 3.0

Tuples in rest parameters and spread expressions

TypeScript 3.0 adds support to multiple new capabilities to interact with function parameter lists as tuple types. TypeScript 3.0 adds support for:

With these features it becomes possible to strongly type a number of higher-order functions that transform functions and their parameter lists.

Rest parameters with tuple types

When a rest parameter has a tuple type, the tuple type is expanded into a sequence of discrete parameters. For example the following two declarations are equivalent:

declare function foo(...args: [number, string, boolean]): void;
declare function foo(args_0: number, args_1: string, args_2: boolean): void;

Spread expressions with tuple types

When a function call includes a spread expression of a tuple type as the last argument, the spread expression corresponds to a sequence of discrete arguments of the tuple element types.

Thus, the following calls are equivalent:

const args: [number, string, boolean] = [42, "hello", true]; foo(42, "hello", true); foo(args[0], args[1], args[2]); foo(...args);

Generic rest parameters

A rest parameter is permitted to have a generic type that is constrained to an array type, and type inference can infer tuple types for such generic rest parameters. This enables higher-order capturing and spreading of partial parameter lists:

declare function bind<T, U extends any[], V>( f: (x: T, ...args: U) => V, x: T ): (...args: U) => V; declare function f3(x: number, y: string, z: boolean): void; const f2 = bind(f3, 42); // (y: string, z: boolean) => void const f1 = bind(f2, "hello"); // (z: boolean) => void const f0 = bind(f1, true); // () => void f3(42, "hello", true); f2("hello", true); f1(true); f0();

In the declaration of f2 above, type inference infers types number, [string, boolean] and void for T, U and V respectively.

Note that when a tuple type is inferred from a sequence of parameters and later expanded into a parameter list, as is the case for U, the original parameter names are used in the expansion (however, the names have no semantic meaning and are not otherwise observable).

Optional elements in tuple types

Tuple types now permit a ? postfix on element types to indicate that the element is optional:

let t: [number, string?, boolean?]; t = [42, "hello", true]; t = [42, "hello"]; t = [42];

In --strictNullChecks mode, a ? modifier automatically includes undefined in the element type, similar to optional parameters.

A tuple type permits an element to be omitted if it has a postfix ? modifier on its type and all elements to the right of it also have ? modifiers.

When tuple types are inferred for rest parameters, optional parameters in the source become optional tuple elements in the inferred type.

The length property of a tuple type with optional elements is a union of numeric literal types representing the possible lengths. For example, the type of the length property in the tuple type [number, string?, boolean?] is 1 | 2 | 3.

Rest elements in tuple types

The last element of a tuple type can be a rest element of the form ...X, where X is an array type. A rest element indicates that the tuple type is open-ended and may have zero or more additional elements of the array element type. For example, [number, ...string[]] means tuples with a number element followed by any number of string elements.

function tuple<T extends any[]>(...args: T): T { return args; } const numbers: number[] = getArrayOfNumbers(); const t1 = tuple("foo", 1, true); // [string, number, boolean] const t2 = tuple("bar", ...numbers); // [string, ...number[]]

The type of the length property of a tuple type with a rest element is number.

New unknown top type

TypeScript 3.0 introduces a new top type unknown. unknown is the type-safe counterpart of any. Anything is assignable to unknown, but unknown isn’t assignable to anything but itself and any without a type assertion or a control flow based narrowing. Likewise, no operations are permitted on an unknown without first asserting or narrowing to a more specific type.

// In an intersection everything absorbs unknown type T00 = unknown & null; // null type T01 = unknown & undefined; // undefined type T02 = unknown & null & undefined; // null & undefined (which becomes never) type T03 = unknown & string; // string type T04 = unknown & string[]; // string[] type T05 = unknown & unknown; // unknown type T06 = unknown & any; // any // In a union an unknown absorbs everything type T10 = unknown | null; // unknown type T11 = unknown | undefined; // unknown type T12 = unknown | null | undefined; // unknown type T13 = unknown | string; // unknown type T14 = unknown | string[]; // unknown type T15 = unknown | unknown; // unknown type T16 = unknown | any; // any // Type variable and unknown in union and intersection type T20<T> = T & {}; // T & {} type T21<T> = T | {}; // T | {} type T22<T> = T & unknown; // T type T23<T> = T | unknown; // unknown // unknown in conditional types type T30<T> = unknown extends T ? true : false; // Deferred type T31<T> = T extends unknown ? true : false; // Deferred (so it distributes) type T32<T> = never extends T ? true : false; // true type T33<T> = T extends never ? true : false; // Deferred // keyof unknown type T40 = keyof any; // string | number | symbol type T41 = keyof unknown; // never // Only equality operators are allowed with unknown function f10(x: unknown) { x == 5; x !== 10; x >= 0; // Error x + 1; // Error x * 2; // Error -x; // Error +x; // Error } // No property accesses, element accesses, or function calls function f11(x: unknown) { x.foo; // Error x[5]; // Error x(); // Error new x(); // Error } // typeof, instanceof, and user defined type predicates declare function isFunction(x: unknown): x is Function; function f20(x: unknown) { if (typeof x === "string" || typeof x === "number") { x; // string | number } if (x instanceof Error) { x; // Error } if (isFunction(x)) { x; // Function } } // Homomorphic mapped type over unknown type T50<T> = { [P in keyof T]: number }; type T51 = T50<any>; // { [x: string]: number } type T52 = T50<unknown>; // {} // Anything is assignable to unknown function f21<T>(pAny: any, pNever: never, pT: T) { let x: unknown; x = 123; x = "hello"; x = [1, 2, 3]; x = new Error(); x = x; x = pAny; x = pNever; x = pT; } // unknown assignable only to itself and any function f22(x: unknown) { let v1: any = x; let v2: unknown = x; let v3: object = x; // Error let v4: string = x; // Error let v5: string[] = x; // Error let v6: {} = x; // Error let v7: {} | null | undefined = x; // Error } // Type parameter 'T extends unknown' not related to object function f23<T extends unknown>(x: T) { let y: object = x; // Error } // Anything but primitive assignable to { [x: string]: unknown } function f24(x: { [x: string]: unknown }) { x = {}; x = { a: 5 }; x = [1, 2, 3]; x = 123; // Error } // Locals of type unknown always considered initialized function f25() { let x: unknown; let y = x; } // Spread of unknown causes result to be unknown function f26(x: {}, y: unknown, z: any) { let o1 = { a: 42, ...x }; // { a: number } let o2 = { a: 42, ...x, ...y }; // unknown let o3 = { a: 42, ...x, ...y, ...z }; // any } // Functions with unknown return type don't need return expressions function f27(): unknown {} // Rest type cannot be created from unknown function f28(x: unknown) { let { ...a } = x; // Error } // Class properties of type unknown don't need definite assignment class C1 { a: string; // Error b: unknown; c: any; }

Support for defaultProps in JSX

TypeScript 2.9 and earlier didn’t leverage React defaultProps declarations inside JSX components. Users would often have to declare properties optional and use non-null assertions inside of render, or they’d use type-assertions to fix up the type of the component before exporting it.

TypeScript 3.0 adds supports a new type alias in the JSX namespace called LibraryManagedAttributes. This helper type defines a transformation on the component’s Props type, before using to check a JSX expression targeting it; thus allowing customization like: how conflicts between provided props and inferred props are handled, how inferences are mapped, how optionality is handled, and how inferences from differing places should be combined.

In short using this general type, we can model React’s specific behavior for things like defaultProps and, to some extent, propTypes.

export interface Props { name: string; } export class Greet extends React.Component<Props> { render() { const { name } = this.props; return <div>Hello ${name.toUpperCase()}!</div>; } static defaultProps = { name: "world" }; } // Type-checks! No type assertions needed! let el = <Greet />;


Explicit types on defaultProps

The default-ed properties are inferred from the defaultProps property type. If an explicit type annotation is added, e.g. static defaultProps: Partial<Props>; the compiler will not be able to identify which properties have defaults (since the type of defaultProps include all properties of Props).

Use static defaultProps: Pick<Props, "name">; as an explicit type annotation instead, or do not add a type annotation as done in the example above.

For stateless function components (SFCs) use ES2015 default initializers for SFCs:

function Greet({ name = "world" }: Props) { return <div>Hello ${name.toUpperCase()}!</div>; }
Changes to @types/React

Corresponding changes to add LibraryManagedAttributes definition to the JSX namespace in @types/React are still needed. Keep in mind that there are some limitations.

/// <reference lib="..." /> reference directives

TypeScript adds a new triple-slash-reference directive (/// <reference lib="name" />), allowing a file to explicitly include an existing built-in lib file.

Built-in lib files are referenced in the same fashion as the "lib" compiler option in tsconfig.json (e.g. use lib="es2015" and not lib="lib.es2015.d.ts", etc.).

For declaration file authors who rely on built-in types, e.g. DOM APIs or built-in JS run-time constructors like Symbol or Iterable, triple-slash-reference lib directives are the recommended. Previously these .d.ts files had to add forward/duplicate declarations of such types.


Using /// <reference lib="es2017.string" /> to one of the files in a compilation is equivalent to compiling with --lib es2017.string.

/// <reference lib="es2017.string" /> "foo".padStart(4);

TypeScript 2.9

Support number and symbol named properties with keyof and mapped types

TypeScript 2.9 adds support for number and symbol named properties in index types and mapped types. Previously, the keyof operator and mapped types only supported string named properties.

Changes include:

  • An index type keyof T for some type T is a subtype of string | number | symbol.
  • A mapped type { [P in K]: XXX } permits any K assignable to string | number | symbol.
  • In a for...in statement for an object of a generic type T, the inferred type of the iteration variable was previously keyof T but is now Extract<keyof T, string>. (In other words, the subset of keyof T that includes only string-like values.)

Given an object type X, keyof X is resolved as follows:

  • If X contains a string index signature, keyof X is a union of string, number, and the literal types representing symbol-like properties, otherwise
  • If X contains a numeric index signature, keyof X is a union of number and the literal types representing string-like and symbol-like properties, otherwise
  • keyof X is a union of the literal types representing string-like, number-like, and symbol-like properties.


  • String-like properties of an object type are those declared using an identifier, a string literal, or a computed property name of a string literal type.
  • Number-like properties of an object type are those declared using a numeric literal or computed property name of a numeric literal type.
  • Symbol-like properties of an object type are those declared using a computed property name of a unique symbol type.

In a mapped type { [P in K]: XXX }, each string literal type in K introduces a property with a string name, each numeric literal type in K introduces a property with a numeric name, and each unique symbol type in K introduces a property with a unique symbol name. Furthermore, if K includes type string, a string index signature is introduced, and if K includes type number, a numeric index signature is introduced.


const c = "c"; const d = 10; const e = Symbol(); const enum E1 { A, B, C, } const enum E2 { A = "A", B = "B", C = "C", } type Foo = { a: string; // String-like name 5: string; // Number-like name [c]: string; // String-like name [d]: string; // Number-like name [e]: string; // Symbol-like name [E1.A]: string; // Number-like name [E2.A]: string; // String-like name }; type K1 = keyof Foo; // "a" | 5 | "c" | 10 | typeof e | E1.A | E2.A type K2 = Extract<keyof Foo, string>; // "a" | "c" | E2.A type K3 = Extract<keyof Foo, number>; // 5 | 10 | E1.A type K4 = Extract<keyof Foo, symbol>; // typeof e

Since keyof now reflects the presence of a numeric index signature by including type number in the key type, mapped types such as Partial<T> and Readonly<T> work correctly when applied to object types with numeric index signatures:

type Arrayish<T> = { length: number; [x: number]: T; }; type ReadonlyArrayish<T> = Readonly<Arrayish<T>>; declare const map: ReadonlyArrayish<string>; let n = map.length; let x = map[123]; // Previously of type any (or an error with --noImplicitAny)

Furthermore, with the keyof operator’s support for number and symbol named keys, it is now possible to abstract over access to properties of objects that are indexed by numeric literals (such as numeric enum types) and unique symbols.

const enum Enum { A, B, C, } const enumToStringMap = { [Enum.A]: "Name A", [Enum.B]: "Name B", [Enum.C]: "Name C", }; const sym1 = Symbol(); const sym2 = Symbol(); const sym3 = Symbol(); const symbolToNumberMap = { [sym1]: 1, [sym2]: 2, [sym3]: 3, }; type KE = keyof typeof enumToStringMap; // Enum (i.e. Enum.A | Enum.B | Enum.C) type KS = keyof typeof symbolToNumberMap; // typeof sym1 | typeof sym2 | typeof sym3 function getValue<T, K extends keyof T>(obj: T, key: K): T[K] { return obj[key]; } let x1 = getValue(enumToStringMap, Enum.C); // Returns "Name C" let x2 = getValue(symbolToNumberMap, sym3); // Returns 3

This is a breaking change; previously, the keyof operator and mapped types only supported string named properties. Code that assumed values typed with keyof T were always strings, will now be flagged as error.


function useKey<T, K extends keyof T>(o: T, k: K) { var name: string = k; // Error: keyof T is not assignable to string }


  • If your functions are only able to handle string named property keys, use Extract<keyof T, string> in the declaration:

    function useKey<T, K extends Extract<keyof T, string>>(o: T, k: K) { var name: string = k; // OK }
  • If your functions are open to handling all property keys, then the changes should be done down-stream:

    function useKey<T, K extends keyof T>(o: T, k: K) { var name: string | number | symbol = k; }
  • Otherwise use --keyofStringsOnly compiler option to disable the new behavior.

Generic type arguments in JSX elements

JSX elements now allow passing type arguments to generic components.


class GenericComponent<P> extends React.Component<P> { internalProp: P; } type Props = { a: number; b: string }; const x = <GenericComponent<Props> a={10} b="hi" />; // OK const y = <GenericComponent<Props> a={10} b={20} />; // Error

Generic type arguments in generic tagged templates

Tagged templates are a form of invocation introduced in ECMAScript 2015. Like call expressions, generic functions may be used in a tagged template and TypeScript will infer the type arguments utilized.

TypeScript 2.9 allows passing generic type arguments to tagged template strings.


declare function styledComponent<Props>( strs: TemplateStringsArray ): Component<Props>; interface MyProps { name: string; age: number; } styledComponent<MyProps>` font-size: 1.5em; text-align: center; color: palevioletred; `; declare function tag<T>(strs: TemplateStringsArray, ...args: T[]): T; // inference fails because 'number' and 'string' are both candidates that conflict let a = tag<string | number>`${100} ${"hello"}`;

import types

Modules can import types declared in other modules. But non-module global scripts cannot access types declared in modules. Enter import types.

Using import("mod") in a type annotation allows for reaching in a module and accessing its exported declaration without importing it.


Given a declaration of a class Pet in a module file:

// module.d.ts export declare class Pet { name: string; }

Can be used in a non-module file global-script.ts:

// global-script.ts function adopt(p: import("./module").Pet) { console.log(`Adopting ${p.name}...`); }

This also works in JSDoc comments to refer to types from other modules in .js:

// a.js /** * @param p { import("./module").Pet } */ function walk(p) { console.log(`Walking ${p.name}...`); }

Relaxing declaration emit visiblity rules

With import types available, many of the visibility errors reported during declaration file generation can be handled by the compiler without the need to change the input.

For instance:

import { createHash } from "crypto"; export const hash = createHash("sha256"); // ^^^^ // Exported variable 'hash' has or is using name 'Hash' from external module "crypto" but cannot be named.

With TypeScript 2.9, no errors are reported, and now the generated file looks like:

export declare const hash: import("crypto").Hash;

Support for import.meta

TypeScript 2.9 introduces support for import.meta, a new meta-property as described by the current TC39 proposal.

The type of import.meta is the global ImportMeta type which is defined in lib.es5.d.ts. This interface is extremely limited. Adding well-known properties for Node or browsers requires interface merging and possibly a global augmentation depending on the context.


Assuming that __dirname is always available on import.meta, the declaration would be done through reopening ImportMeta interface:

// node.d.ts interface ImportMeta { __dirname: string; }

And usage would be:

import.meta.__dirname; // Has type 'string'

import.meta is only allowed when targeting ESNext modules and ECMAScript targets.

New --resolveJsonModule

Often in Node.js applications a .json is needed. With TypeScript 2.9, --resolveJsonModule allows for importing, extracting types from and generating .json files.


// settings.json { "repo": "TypeScript", "dry": false, "debug": false }
// a.ts import settings from "./settings.json"; settings.debug === true; // OK settings.dry === 2; // Error: Operator '===' cannot be applied boolean and number
// tsconfig.json { "compilerOptions": { "module": "commonjs", "resolveJsonModule": true, "esModuleInterop": true } }

--pretty output by default

Starting TypeScript 2.9 errors are displayed under --pretty by default if the output device is applicable for colorful text. TypeScript will check if the output steam has isTty property set.

Use --pretty false on the command line or set "pretty": false in your tsconfig.json to disable --pretty output.

New --declarationMap

Enabling --declarationMap alongside --declaration causes the compiler to emit .d.ts.map files alongside the output .d.ts files. Language Services can also now understand these map files, and uses them to map declaration-file based definition locations to their original source, when available.

In other words, hitting go-to-definition on a declaration from a .d.ts file generated with --declarationMap will take you to the source file (.ts) location where that declaration was defined, and not to the .d.ts.

TypeScript 2.8

Conditional Types

TypeScript 2.8 introduces conditional types which add the ability to express non-uniform type mappings. A conditional type selects one of two possible types based on a condition expressed as a type relationship test:

T extends U ? X : Y

The type above means when T is assignable to U the type is X, otherwise the type is Y.

A conditional type T extends U ? X : Y is either resolved to X or Y, or deferred because the condition depends on one or more type variables. Whether to resolve or defer is determined as follows:

  • First, given types T' and U' that are instantiations of T and U where all occurrences of type parameters are replaced with any, if T' is not assignable to U', the conditional type is resolved to Y. Intuitively, if the most permissive instantiation of T is not assignable to the most permissive instantiation of U, we know that no instantiation will be and we can just resolve to Y.
  • Next, for each type variable introduced by an infer (more later) declaration within U collect a set of candidate types by inferring from T to U (using the same inference algorithm as type inference for generic functions). For a given infer type variable V, if any candidates were inferred from co-variant positions, the type inferred for V is a union of those candidates. Otherwise, if any candidates were inferred from contra-variant positions, the type inferred for V is an intersection of those candidates. Otherwise, the type inferred for V is never.
  • Then, given a type T'' that is an instantiation of T where all infer type variables are replaced with the types inferred in the previous step, if T'' is definitely assignable to U, the conditional type is resolved to X. The definitely assignable relation is the same as the regular assignable relation, except that type variable constraints are not considered. Intuitively, when a type is definitely assignable to another type, we know that it will be assignable for all instantiations of those types.
  • Otherwise, the condition depends on one or more type variables and the conditional type is deferred.


type TypeName<T> = T extends string ? "string" : T extends number ? "number" : T extends boolean ? "boolean" : T extends undefined ? "undefined" : T extends Function ? "function" : "object"; type T0 = TypeName<string>; // "string" type T1 = TypeName<"a">; // "string" type T2 = TypeName<true>; // "boolean" type T3 = TypeName<() => void>; // "function" type T4 = TypeName<string[]>; // "object"

Distributive conditional types

Conditional types in which the checked type is a naked type parameter are called distributive conditional types. Distributive conditional types are automatically distributed over union types during instantiation. For example, an instantiation of T extends U ? X : Y with the type argument A | B | C for T is resolved as (A extends U ? X : Y) | (B extends U ? X : Y) | (C extends U ? X : Y).


type T10 = TypeName<string | (() => void)>; // "string" | "function" type T12 = TypeName<string | string[] | undefined>; // "string" | "object" | "undefined" type T11 = TypeName<string[] | number[]>; // "object"

In instantiations of a distributive conditional type T extends U ? X : Y, references to T within the conditional type are resolved to individual constituents of the union type (i.e. T refers to the individual constituents after the conditional type is distributed over the union type). Furthermore, references to T within X have an additional type parameter constraint U (i.e. T is considered assignable to U within X).


type BoxedValue<T> = { value: T }; type BoxedArray<T> = { array: T[] }; type Boxed<T> = T extends any[] ? BoxedArray<T[number]> : BoxedValue<T>; type T20 = Boxed<string>; // BoxedValue<string>; type T21 = Boxed<number[]>; // BoxedArray<number>; type T22 = Boxed<string | number[]>; // BoxedValue<string> | BoxedArray<number>;

Notice that T has the additional constraint any[] within the true branch of Boxed<T> and it is therefore possible to refer to the element type of the array as T[number]. Also, notice how the conditional type is distributed over the union type in the last example.

The distributive property of conditional types can conveniently be used to filter union types:

type Diff<T, U> = T extends U ? never : T; // Remove types from T that are assignable to U type Filter<T, U> = T extends U ? T : never; // Remove types from T that are not assignable to U type T30 = Diff<"a" | "b" | "c" | "d", "a" | "c" | "f">; // "b" | "d" type T31 = Filter<"a" | "b" | "c" | "d", "a" | "c" | "f">; // "a" | "c" type T32 = Diff<string | number | (() => void), Function>; // string | number type T33 = Filter<string | number | (() => void), Function>; // () => void type NonNullable<T> = Diff<T, null | undefined>; // Remove null and undefined from T type T34 = NonNullable<string | number | undefined>; // string | number type T35 = NonNullable<string | string[] | null | undefined>; // string | string[] function f1<T>(x: T, y: NonNullable<T>) { x = y; // Ok y = x; // Error } function f2<T extends string | undefined>(x: T, y: NonNullable<T>) { x = y; // Ok y = x; // Error let s1: string = x; // Error let s2: string = y; // Ok }

Conditional types are particularly useful when combined with mapped types:

type FunctionPropertyNames<T> = { [K in keyof T]: T[K] extends Function ? K : never; }[keyof T]; type FunctionProperties<T> = Pick<T, FunctionPropertyNames<T>>; type NonFunctionPropertyNames<T> = { [K in keyof T]: T[K] extends Function ? never : K; }[keyof T]; type NonFunctionProperties<T> = Pick<T, NonFunctionPropertyNames<T>>; interface Part { id: number; name: string; subparts: Part[]; updatePart(newName: string): void; } type T40 = FunctionPropertyNames<Part>; // "updatePart" type T41 = NonFunctionPropertyNames<Part>; // "id" | "name" | "subparts" type T42 = FunctionProperties<Part>; // { updatePart(newName: string): void } type T43 = NonFunctionProperties<Part>; // { id: number, name: string, subparts: Part[] }

Similar to union and intersection types, conditional types are not permitted to reference themselves recursively. For example the following is an error.


type ElementType<T> = T extends any[] ? ElementType<T[number]> : T; // Error

Type inference in conditional types

Within the extends clause of a conditional type, it is now possible to have infer declarations that introduce a type variable to be inferred. Such inferred type variables may be referenced in the true branch of the conditional type. It is possible to have multiple infer locations for the same type variable.

For example, the following extracts the return type of a function type:

type ReturnType<T> = T extends (...args: any[]) => infer R ? R : any;

Conditional types can be nested to form a sequence of pattern matches that are evaluated in order:

type Unpacked<T> = T extends (infer U)[] ? U : T extends (...args: any[]) => infer U ? U : T extends Promise<infer U> ? U : T; type T0 = Unpacked<string>; // string type T1 = Unpacked<string[]>; // string type T2 = Unpacked<() => string>; // string type T3 = Unpacked<Promise<string>>; // string type T4 = Unpacked<Promise<string>[]>; // Promise<string> type T5 = Unpacked<Unpacked<Promise<string>[]>>; // string

The following example demonstrates how multiple candidates for the same type variable in co-variant positions causes a union type to be inferred:

type Foo<T> = T extends { a: infer U; b: infer U } ? U : never; type T10 = Foo<{ a: string; b: string }>; // string type T11 = Foo<{ a: string; b: number }>; // string | number

Likewise, multiple candidates for the same type variable in contra-variant positions causes an intersection type to be inferred:

type Bar<T> = T extends { a: (x: infer U) => void; b: (x: infer U) => void } ? U : never; type T20 = Bar<{ a: (x: string) => void; b: (x: string) => void }>; // string type T21 = Bar<{ a: (x: string) => void; b: (x: number) => void }>; // string & number

When inferring from a type with multiple call signatures (such as the type of an overloaded function), inferences are made from the last signature (which, presumably, is the most permissive catch-all case). It is not possible to perform overload resolution based on a list of argument types.

declare function foo(x: string): number; declare function foo(x: number): string; declare function foo(x: string | number): string | number; type T30 = ReturnType<typeof foo>; // string | number

It is not possible to use infer declarations in constraint clauses for regular type parameters:

type ReturnType<T extends (...args: any[]) => infer R> = R; // Error, not supported

However, much the same effect can be obtained by erasing the type variables in the constraint and instead specifying a conditional type:

type AnyFunction = (...args: any[]) => any; type ReturnType<T extends AnyFunction> = T extends (...args: any[]) => infer R ? R : any;

Predefined conditional types

TypeScript 2.8 adds several predefined conditional types to lib.d.ts:

  • Exclude<T, U> — Exclude from T those types that are assignable to U.
  • Extract<T, U> — Extract from T those types that are assignable to U.
  • NonNullable<T> — Exclude null and undefined from T.
  • ReturnType<T> — Obtain the return type of a function type.
  • InstanceType<T> — Obtain the instance type of a constructor function type.


type T00 = Exclude<"a" | "b" | "c" | "d", "a" | "c" | "f">; // "b" | "d" type T01 = Extract<"a" | "b" | "c" | "d", "a" | "c" | "f">; // "a" | "c" type T02 = Exclude<string | number | (() => void), Function>; // string | number type T03 = Extract<string | number | (() => void), Function>; // () => void type T04 = NonNullable<string | number | undefined>; // string | number type T05 = NonNullable<(() => string) | string[] | null | undefined>; // (() => string) | string[] function f1(s: string) { return { a: 1, b: s }; } class C { x = 0; y = 0; } type T10 = ReturnType<() => string>; // string type T11 = ReturnType<(s: string) => void>; // void type T12 = ReturnType<<T>() => T>; // {} type T13 = ReturnType<<T extends U, U extends number[]>() => T>; // number[] type T14 = ReturnType<typeof f1>; // { a: number, b: string } type T15 = ReturnType<any>; // any type T16 = ReturnType<never>; // any type T17 = ReturnType<string>; // Error type T18 = ReturnType<Function>; // Error type T20 = InstanceType<typeof C>; // C type T21 = InstanceType<any>; // any type T22 = InstanceType<never>; // any type T23 = InstanceType<string>; // Error type T24 = InstanceType<Function>; // Error

Note: The Exclude type is a proper implementation of the Diff type suggested here. We’ve used the name Exclude to avoid breaking existing code that defines a Diff, plus we feel that name better conveys the semantics of the type. We did not include the Omit<T, K> type because it is trivially written as Pick<T, Exclude<keyof T, K>>.

Improved control over mapped type modifiers

Mapped types support adding a readonly or ? modifier to a mapped property, but they did not provide support the ability to remove modifiers. This matters in homomorphic mapped types which by default preserve the modifiers of the underlying type.

TypeScript 2.8 adds the ability for a mapped type to either add or remove a particular modifier. Specifically, a readonly or ? property modifier in a mapped type can now be prefixed with either + or - to indicate that the modifier should be added or removed.


type MutableRequired<T> = { -readonly [P in keyof T]-?: T[P] }; // Remove readonly and ? type ReadonlyPartial<T> = { +readonly [P in keyof T]+?: T[P] }; // Add readonly and ?

A modifier with no + or - prefix is the same as a modifier with a + prefix. So, the ReadonlyPartial<T> type above corresponds to

type ReadonlyPartial<T> = { readonly [P in keyof T]?: T[P] }; // Add readonly and ?

Using this ability, lib.d.ts now has a new Required<T> type. This type strips ? modifiers from all properties of T, thus making all properties required.


type Required<T> = { [P in keyof T]-?: T[P] };

Note that in --strictNullChecks mode, when a homomorphic mapped type removes a ? modifier from a property in the underlying type it also removes undefined from the type of that property:


type Foo = { a?: string }; // Same as { a?: string | undefined } type Bar = Required<Foo>; // Same as { a: string }

Improved keyof with intersection types

With TypeScript 2.8 keyof applied to an intersection type is transformed to a union of keyof applied to each intersection constituent. In other words, types of the form keyof (A & B) are transformed to be keyof A | keyof B. This change should address inconsistencies with inference from keyof expressions.


type A = { a: string }; type B = { b: string }; type T1 = keyof (A & B); // "a" | "b" type T2<T> = keyof (T & B); // keyof T | "b" type T3<U> = keyof (A & U); // "a" | keyof U type T4<T, U> = keyof (T & U); // keyof T | keyof U type T5 = T2<A>; // "a" | "b" type T6 = T3<B>; // "a" | "b" type T7 = T4<A, B>; // "a" | "b"

Better handling for namespace patterns in .js files

TypeScript 2.8 adds support for understanding more namespace patterns in .js files. Empty object literals declarations on top level, just like functions and classes, are now recognized as as namespace declarations in JavaScript.

var ns = {}; // recognized as a declaration for a namespace `ns` ns.constant = 1; // recognized as a declaration for var `constant`

Assignments at the top-level should behave the same way; in other words, a var or const declaration is not required.

app = {}; // does NOT need to be `var app = {}` app.C = class {}; app.f = function () {}; app.prop = 1;

IIFEs as namespace declarations

An IIFE returning a function, class or empty object literal, is also recognized as a namespace:

var C = (function () { function C(n) { this.p = n; } return C; })(); C.staticProperty = 1;

Defaulted declarations

“Defaulted declarations” allow initializers that reference the declared name in the left side of a logical or:

my = window.my || {}; my.app = my.app || {};

Prototype assignment

You can assign an object literal directly to the prototype property. Individual prototype assignments still work too:

var C = function (p) { this.p = p; }; C.prototype = { m() { console.log(this.p); }, }; C.prototype.q = function (r) { return this.p === r; };

Nested and merged declarations

Nesting works to any level now, and merges correctly across files. Previously neither was the case.

var app = window.app || {}; app.C = class {};

Per-file JSX factories

TypeScript 2.8 adds support for a per-file configurable JSX factory name using @jsx dom paragma. JSX factory can be configured for a compilation using --jsxFactory (default is React.createElement). With TypeScript 2.8 you can override this on a per-file-basis by adding a comment to the beginning of the file.


/** @jsx dom */ import { dom } from "./renderer"; <h></h>;


var renderer_1 = require("./renderer"); renderer_1.dom("h", null);

Locally scoped JSX namespaces

JSX type checking is driven by definitions in a JSX namespace, for instance JSX.Element for the type of a JSX element, and JSX.IntrinsicElements for built-in elements. Before TypeScript 2.8 the JSX namespace was expected to be in the global namespace, and thus only allowing one to be defined in a project. Starting with TypeScript 2.8 the JSX namespace will be looked under the jsxNamespace (e.g. React) allowing for multiple jsx factories in one compilation. For backward compatibility the global JSX namespace is used as a fallback if none was defined on the factory function. Combined with the per-file @jsx pragma, each file can have a different JSX factory.

New --emitDeclarationsOnly

--emitDeclarationsOnly allows for only generating declaration files; .js/.jsx output generation will be skipped with this flag. The flag is useful when the .js output generation is handled by a different transpiler like Babel.

TypeScript 2.7

Constant-named properties

TypeScript 2.7 adds support for declaring const-named properties on types including ECMAScript symbols.


// Lib export const SERIALIZE = Symbol("serialize-method-key"); export interface Serializable { [SERIALIZE](obj: {}): string; }
// consumer import { SERIALIZE, Serializable } from "lib"; class JSONSerializableItem implements Serializable { [SERIALIZE](obj: {}) { return JSON.stringify(obj); } }

This also applies to numeric and string literals.


const Foo = "Foo"; const Bar = "Bar"; let x = { [Foo]: 100, [Bar]: "hello", }; let a = x[Foo]; // has type 'number' let b = x[Bar]; // has type 'string'

unique symbol

To enable treating symbols as unique literals a new type unique symbol is available. unique symbol is are subtype of symbol, and are produced only from calling Symbol() or Symbol.for(), or from explicit type annotations. The new type is only allowed on const declarations and readonly static properties, and in order to reference a specific unique symbol, you’ll have to use the typeof operator. Each reference to a unique symbol implies a completely unique identity that’s tied to a given declaration.


// Works declare const Foo: unique symbol; // Error! 'Bar' isn't a constant. let Bar: unique symbol = Symbol(); // Works - refers to a unique symbol, but its identity is tied to 'Foo'. let Baz: typeof Foo = Foo; // Also works. class C { static readonly StaticSymbol: unique symbol = Symbol(); }

Because each unique symbol has a completely separate identity, no two unique symbol types are assignable or comparable to each other.


const Foo = Symbol(); const Bar = Symbol(); // Error: can't compare two unique symbols. if (Foo === Bar) { // ... }

Strict Class Initialization

TypeScript 2.7 introduces a new flag called --strictPropertyInitialization. This flag performs checks to ensure that each instance property of a class gets initialized in the constructor body, or by a property initializer. For example

class C { foo: number; bar = "hello"; baz: boolean; // ~~~ // Error! Property 'baz' has no initializer and is not definitely assigned in the // constructor. constructor() { this.foo = 42; } }

In the above, if we truly meant for baz to potentially be undefined, we should have declared it with the type boolean | undefined.

There are certain scenarios where properties can be initialized indirectly (perhaps by a helper method or dependency injection library), in which case you can use the new definite assignment assertion modifiers for your properties (discussed below).

class C { foo!: number; // ^ // Notice this '!' modifier. // This is the "definite assignment assertion" constructor() { this.initialize(); } initialize() { this.foo = 0; } }

Keep in mind that --strictPropertyInitialization will be turned on along with other --strict mode flags, which can impact your project. You can set the strictPropertyInitialization setting to false in your tsconfig.json’s compilerOptions, or --strictPropertyInitialization false on the command line to turn off this checking.

Definite Assignment Assertions

The definite assignment assertion is a feature that allows a ! to be placed after instance property and variable declarations to relay to TypeScript that a variable is indeed assigned for all intents and purposes, even if TypeScript’s analyses cannot detect so.

For example:

let x: number; initialize(); console.log(x + x); // ~ ~ // Error! Variable 'x' is used before being assigned. function initialize() { x = 10; }

With definite assignment assertions, we can assert that x is really assigned by appending an ! to its declaration:

// Notice the '!' let x!: number; initialize(); // No error! console.log(x + x); function initialize() { x = 10; }

In a sense, the definite assignment assertion operator is the dual of the non-null assertion operator (in which expressions are post-fixed with a !), which we could also have used in the example.

let x: number; initialize(); // No error! console.log(x! + x!); function initialize() { x = 10;

In our example, we knew that all uses of x would be initialized so it makes more sense to use definite assignment assertions than non-null assertions.

Fixed Length Tuples

In TypeScript 2.6 and earlier, [number, string, string] was considered a subtype of [number, string]. This was motivated by TypeScript’s structural nature; the first and second elements of a [number, string, string] are respectively subtypes of the first and second elements of [number, string]. However, after examining real world usage of tuples, we noticed that most situations in which this was permitted was typically undesirable.

In TypeScript 2.7, tuples of different arities are no longer assignable to each other. Thanks to a pull request from Tycho Grouwstra, tuple types now encode their arity into the type of their respective length property. This is accomplished by leveraging numeric literal types, which now allow tuples to be distinct from tuples of different arities.

Conceptually, you might consider the type [number, string] to be equivalent to the following declaration of NumStrTuple:

interface NumStrTuple extends Array<number | string> { 0: number; 1: string; length: 2; // using the numeric literal type '2' }

Note that this is a breaking change for some code. If you need to resort to the original behavior in which tuples only enforce a minimum length, you can use a similar declaration that does not explicitly define a length property, falling back to number.

interface MinimumNumStrTuple extends Array<number | string> { 0: number; 1: string; }

Note that this does not imply tuples represent immutable arrays, but it is an implied convention.

Improved type inference for object literals

TypeScript 2.7 improves type inference for multiple object literals occurring in the same context. When multiple object literal types contribute to a union type, we now normalize the object literal types such that all properties are present in each constituent of the union type.


const obj = test ? { text: "hello" } : {}; // { text: string } | { text?: undefined } const s = obj.text; // string | undefined

Previously type {} was inferred for obj and the second line subsequently caused an error because obj would appear to have no properties. That obviously wasn’t ideal.


// let obj: { a: number, b: number } | // { a: string, b?: undefined } | // { a?: undefined, b?: undefined } let obj = [{ a: 1, b: 2 }, { a: "abc" }, {}][0]; obj.a; // string | number | undefined obj.b; // number | undefined

Multiple object literal type inferences for the same type parameter are similarly collapsed into a single normalized union type:

declare function f<T>(...items: T[]): T; // let obj: { a: number, b: number } | // { a: string, b?: undefined } | // { a?: undefined, b?: undefined } let obj = f({ a: 1, b: 2 }, { a: "abc" }, {}); obj.a; // string | number | undefined obj.b; // number | undefined

Improved handling of structurally identical classes and instanceof expressions

TypeScript 2.7 improves the handling of structurally identical classes in union types and instanceof expressions:

  • Structurally identical, but distinct, class types are now preserved in union types (instead of eliminating all but one).
  • Union type subtype reduction only removes a class type if it is a subclass of and derives from another class type in the union.
  • Type checking of the instanceof operator is now based on whether the type of the left operand derives from the type indicated by the right operand (as opposed to a structural subtype check).

This means that union types and instanceof properly distinguish between structurally identical classes.


class A {} class B extends A {} class C extends A {} class D extends A { c: string; } class E extends D {} let x1 = !true ? new A() : new B(); // A let x2 = !true ? new B() : new C(); // B | C (previously B) let x3 = !true ? new C() : new D(); // C | D (previously C) let a1 = [new A(), new B(), new C(), new D(), new E()]; // A[] let a2 = [new B(), new C(), new D(), new E()]; // (B | C | D)[] (previously B[]) function f1(x: B | C | D) { if (x instanceof B) { x; // B (previously B | D) } else if (x instanceof C) { x; // C } else { x; // D (previously never) } }

Type guards inferred from in operator

The in operator now acts as a narrowing expression for types.

For a n in x expression, where n is a string literal or string literal type and x is a union type, the “true” branch narrows to types which have an optional or required property n, and the “false” branch narrows to types which have an optional or missing property n.


interface A { a: number; } interface B { b: string; } function foo(x: A | B) { if ("a" in x) { return x.a; } return x.b; }

Support for import d from "cjs" form CommonJS modules with --esModuleInterop

TypeScript 2.7 updates CommonJS/AMD/UMD module emit to synthesize namespace records based on the presence of an __esModule indicator under --esModuleInterop. The change brings the generated output from TypeScript closer to that generated by Babel.

Previously CommonJS/AMD/UMD modules were treated in the same way as ES6 modules, resulting in a couple of problems. Namely:

  • TypeScript treats a namespace import (i.e. import * as foo from "foo") for a CommonJS/AMD/UMD module as equivalent to const foo = require("foo"). Things are simple here, but they don’t work out if the primary object being imported is a primitive or a class or a function. ECMAScript spec stipulates that a namespace record is a plain object, and that a namespace import (foo in the example above) is not callable, though allowed by TypeScript
  • Similarly a default import (i.e. import d from "foo") for a CommonJS/AMD/UMD module as equivalent to const d = require("foo").default. Most of the CommonJS/AMD/UMD modules available today do not have a default export, making this import pattern practically unusable to import non-ES modules (i.e. CommonJS/AMD/UMD). For instance import fs from "fs" or import express from "express" are not allowed.

Under the new --esModuleInterop these two issues should be addressed:

  • A namespace import (i.e. import * as foo from "foo") is now correctly flagged as uncallabale. Calling it will result in an error.
  • Default imports to CommonJS/AMD/UMD are now allowed (e.g. import fs from "fs"), and should work as expected.

Note: The new behavior is added under a flag to avoid unwarranted breaks to existing code bases. We highly recommend applying it both to new and existing projects. For existing projects, namespace imports (import * as express from "express"; express();) will need to be converted to default imports (import express from "express"; express();).


With --esModuleInterop two new helpers are generated __importStar and __importDefault for import * and import default respectively. For instance input like:

import * as foo from "foo"; import b from "bar";

Will generate:

"use strict"; var __importStar = (this && this.__importStar) || function (mod) { if (mod && mod.__esModule) return mod; var result = {}; if (mod != null) for (var k in mod) if (Object.hasOwnProperty.call(mod, k)) result[k] = mod[k]; result["default"] = mod; return result; }; var __importDefault = (this && this.__importDefault) || function (mod) { return mod && mod.__esModule ? mod : { default: mod }; }; exports.__esModule = true; var foo = __importStar(require("foo")); var bar_1 = __importDefault(require("bar"));

Numeric separators

TypeScript 2.7 brings support for ES Numeric Separators. Numeric literals can now be separated into segments using _.

const milion = 1_000_000; const phone = 555_734_2231; const bytes = 0xff_0c_00_ff; const word = 0b1100_0011_1101_0001;

Cleaner output in --watch mode

TypeScript’s --watch mode now clears the screen after a re-compilation is requested.

Prettier --pretty output

TypeScript’s --pretty flag can make error messages easier to read and manage. --pretty now uses colors for file names, diagnostic codes, and line numbers. File names and positions are now also formatted to allow navigation in common terminals (e.g. Visual Studio Code terminal).

TypeScript 2.6

Strict function types

TypeScript 2.6 introduces a new strict checking flag, --strictFunctionTypes. The --strictFunctionTypes switch is part of the --strict family of switches, meaning that it defaults to on in --strict mode. You can opt-out by setting --strictFunctionTypes false on your command line or in your tsconfig.json.

Under --strictFunctionTypes function type parameter positions are checked contravariantly instead of bivariantly. For some background on what variance means for function types check out What are covariance and contravariance?.

The stricter checking applies to all function types, except those originating in method or constructor declarations. Methods are excluded specifically to ensure generic classes and interfaces (such as Array<T>) continue to mostly relate covariantly.

Consider the following example in which Animal is the supertype of Dog and Cat:

declare let f1: (x: Animal) => void; declare let f2: (x: Dog) => void; declare let f3: (x: Cat) => void; f1 = f2; // Error with --strictFunctionTypes f2 = f1; // Ok f2 = f3; // Error

The first assignment is permitted in default type checking mode, but flagged as an error in strict function types mode. Intuitively, the default mode permits the assignment because it is possibly sound, whereas strict function types mode makes it an error because it isn’t provably sound. In either mode the third assignment is an error because it is never sound.

Another way to describe the example is that the type (x: T) => void is bivariant (i.e. covariant or contravariant) for T in default type checking mode, but contravariant for T in strict function types mode.


interface Comparer<T> { compare: (a: T, b: T) => number; } declare let animalComparer: Comparer<Animal>; declare let dogComparer: Comparer<Dog>; animalComparer = dogComparer; // Error dogComparer = animalComparer; // Ok

The first assignment is now an error. Effectively, T is contravariant in Comparer<T> because it is used only in function type parameter positions.

By the way, note that whereas some languages (e.g. C# and Scala) require variance annotations (out/in or +/-), variance emerges naturally from the actual use of a type parameter within a generic type due to TypeScript’s structural type system.


Under --strictFunctionTypes the first assignment is still permitted if compare was declared as a method. Effectively, T is bivariant in Comparer<T> because it is used only in method parameter positions.

interface Comparer<T> { compare(a: T, b: T): number; } declare let animalComparer: Comparer<Animal>; declare let dogComparer: Comparer<Dog>; animalComparer = dogComparer; // Ok because of bivariance dogComparer = animalComparer; // Ok

TypeScript 2.6 also improves type inference involving contravariant positions:

function combine<T>(...funcs: ((x: T) => void)[]): (x: T) => void { return (x) => { for (const f of funcs) f(x); }; } function animalFunc(x: Animal) {} function dogFunc(x: Dog) {} let combined = combine(animalFunc, dogFunc); // (x: Dog) => void

Above, all inferences for T originate in contravariant positions, and we therefore infer the best common subtype for T. This contrasts with inferences from covariant positions, where we infer the best common supertype.

Support for JSX Fragment Syntax

TypeScript 2.6.2 adds support for the new <>...</> syntax for fragments in JSX. It is frequently desirable to return multiple children from a component. However, this is invalid, so the usual approach has been to wrap the text in an extra element, such as a <div> or <span> as shown below.

render() { return ( <div> Some text. <h2>A heading</h2> More text. </div> ); }

To address this pattern, React introduced the React.Fragment component, which provides a dedicated way to wrap such elements without adding an element to the DOM. Correspondingly, the <>...</> syntax was added to JSX to facilitate this new construct. Therefore, the above scenario becomes:

render() { return ( <> Some text. <h2>A heading</h2> More text. </> ); }

Under --jsx preserve, the new syntax is left untouched for TypeScript emit. Otherwise, for --jsx react, <>...</> is compiled to React.createElement(React.Fragment, null, ...), where React.createElement respects --jsxFactory. Note that it is an error to use <>...</> when --jsx react and --jsxFactory are both enabled.

Please refer to the React blog for more details on fragments and the new syntax.

Cache tagged template objects in modules

TypeScript 2.6 fixes the tagged string template emit to align better with the ECMAScript spec. As per the ECMAScript spec, every time a template tag is evaluated, the same template strings object (the same TemplateStringsArray) should be passed as the first argument. Before TypeScript 2.6, the generated output was a completely new template object each time. Though the string contents are the same, this emit affects libraries that use the identity of the string for cache invalidation purposes, e.g. lit-html.


export function id(x: TemplateStringsArray) { return x; } export function templateObjectFactory() { return id`hello world`; } let result = templateObjectFactory() === templateObjectFactory(); // true in TS 2.6

Results in the following generated code:

"use strict"; var __makeTemplateObject = (this && this.__makeTemplateObject) || function (cooked, raw) { if (Object.defineProperty) { Object.defineProperty(cooked, "raw", { value: raw }); } else { cooked.raw = raw; } return cooked; }; function id(x) { return x; } var _a; function templateObjectFactory() { return id( _a || (_a = __makeTemplateObject(["hello world"], ["hello world"])) ); } var result = templateObjectFactory() === templateObjectFactory();

Note: This change brings a new emit helper, __makeTemplateObject; if you are using --importHelpers with tslib, an updated to version 1.8 or later.

Localized diagnostics on the command line

TypeScript 2.6 npm package ships with localized versions of diagnostic messages for 13 languages. The localized messages are available when using --locale flag on the command line.


Error messages in Russian:

c:\ts>tsc --v Version 2.6.0-dev.20171003 c:\ts>tsc --locale ru --pretty c:\test\a.ts ../test/a.ts(1,5): error TS2322: Тип ""string"" не может быть назначен для типа "number". 1 var x: number = "string"; ~

And help in Japanese:

PS C:\ts> tsc --v Version 2.6.0-dev.20171003 PS C:\ts> tsc --locale ja-jp バージョン 2.6.0-dev.20171003 構文: tsc [オプション] [ファイル ...] 例: tsc hello.ts tsc --outFile file.js file.ts tsc @args.txt オプション: -h, --help このメッセージを表示します。 --all コンパイラ オプションをすべて表示します。 -v, --version コンパイラのバージョンを表示します。 --init TypeScript プロジェクトを初期化して、tsconfig.json ファイルを作成します。 -p ファイルまたはディレクトリ, --project ファイルまたはディレクトリ 構成ファイルか、'tsconfig.json' を含むフォルダーにパスが指定されたプロジェクトをコ ンパイルします。 --pretty 色とコンテキストを使用してエラーとメッセージにスタイルを適用します (試験的)。 -w, --watch 入力ファイルを監視します。 -t バージョン, --target バージョン ECMAScript のターゲット バージョンを指定します: 'ES3' (既定)、'ES5''ES2015''ES2016''ES2017''ES NEXT' -m 種類, --module 種類 モジュール コード生成を指定します: 'none''commonjs''amd''system''umd''es2015''ESNext' --lib コンパイルに含めるライブラリ ファイルを指定します: 'es5' 'es6' 'es2015' 'es7' 'es2016' 'es2017' 'esnext' 'dom' 'dom.iterable' 'webworker' 'scripthost' 'es201 5.core' 'es2015.collection' 'es2015.generator' 'es2015.iterable' 'es2015.promise' 'es2015.proxy' 'es2015.reflect' 'es2015.symbol' 'es2015.symbol.wellkno wn' 'es2016.array.include' 'es2017.object' 'es2017.sharedmemory' 'es2017.string' 'es2017.intl' 'esnext.asynciterable' --allowJs javascript ファイルのコンパイルを許可します。 --jsx 種類 JSX コード生成を指定します: 'preserve''react-native''react' -d, --declaration 対応する '.d.ts' ファイルを生成します。 --sourceMap 対応する '.map' ファイルを生成します。 --outFile ファイル 出力を連結して 1 つのファイルを生成します。 --outDir ディレクトリ ディレクトリへ出力構造をリダイレクトします。 --removeComments コメントを出力しないでください。 --noEmit 出力しないでください。 --strict strict 型チェックのオプションをすべて有効にします。 --noImplicitAny 暗黙的な 'any' 型を含む式と宣言に関するエラーを発生させます。 --strictNullChecks 厳格な null チェックを有効にします。 --noImplicitThis 暗黙的な 'any' 型を持つ 'this' 式でエラーが発生します。 --alwaysStrict 厳格モードで解析してソース ファイルごとに "use strict" を生成します。 --noUnusedLocals 使用されていないローカルに関するエラーを報告します。 --noUnusedParameters 使用されていないパラメーターに関するエラーを報告します。 --noImplicitReturns 関数の一部のコード パスが値を返さない場合にエラーを報告します。 --noFallthroughCasesInSwitch switch ステートメントに case のフォールスルーがある場合にエラーを報告します。 --types コンパイルに含む型宣言ファイル。 @<ファイル>

Suppress errors in .ts files using ’// @ts-ignore’ comments

TypeScript 2.6 support suppressing errors in .js files using // @ts-ignore comments placed above the offending lines.


if (false) { // @ts-ignore: Unreachable code error console.log("hello"); }

A // @ts-ignore comment suppresses all errors that originate on the following line. It is recommended practice to have the remainder of the comment following @ts-ignore explain which error is being suppressed.

Please note that this comment only suppresses the error reporting, and we recommend you use this comments very sparingly.

Faster tsc --watch

TypeScript 2.6 brings a faster --watch implementation. The new version optimizes code generation and checking for code bases using ES modules. Changes detected in a module file will result in only regenerating the changed module, and files that depend on it, instead of the whole project. Projects with large number of files should reap the most benefit from this change.

The new implementation also brings performance enhancements to watching in tsserver. The watcher logic has been completely rewritten to respond faster to change events.

Write-only references now flagged as unused

TypeScript 2.6 adds revised implementation the --noUnusedLocals and --noUnusedParameters compiler options. Declarations are only written to but never read from are now flagged as unused.


Bellow both n and m will be marked as unused, because their values are never read. Previously TypeScript would only check whether their values were referenced.

function f(n: number) { n = 0; } class C { private m: number; constructor() { this.m = 0; } }

Also functions that are only called within their own bodies are considered unused.


function f() { f(); // Error: 'f' is declared but its value is never read }

TypeScript 2.5

Optional catch clause variables

Thanks to work done by @tinganho, TypeScript 2.5 implements a new ECMAScript feature that allows users to omit the variable in catch clauses. For example, when using JSON.parse you may need to wrap calls to the function with a try/catch, but you may not end up using the SyntaxError that gets thrown when input is erroneous.

let input = "..."; try { JSON.parse(input); } catch { // ^ Notice that our `catch` clause doesn't declare a variable. console.log("Invalid JSON given\n\n" + input); }

Type assertion/cast syntax in checkJs/@ts-check mode

TypeScript 2.5 introduces the ability to assert the type of expressions when using plain JavaScript in your projects. The syntax is an /** @type {...} */ annotation comment followed by a parenthesized expression whose type needs to be re-evaluated. For example:

var x = /** @type {SomeType} */ AnyParenthesizedExpression;

Deduplicated and redirected packages

When importing using the Node module resolution strategy in TypeScript 2.5, the compiler will now check whether files originate from “identical” packages. If a file originates from a package with a package.json containing the same name and version fields as a previously encountered package, then TypeScript will redirect itself to the top-most package. This helps resolve problems where two packages might contain identical declarations of classes, but which contain private members that cause them to be structurally incompatible.

As a nice bonus, this can also reduce the memory and runtime footprint of the compiler and language service by avoiding loading .d.ts files from duplicate packages.

TypeScript 2.5 brings the preserveSymlinks flag, which parallels the behavior of the --preserve-symlinks flag in Node.js. This flag also exhibits the opposite behavior to Webpack’s resolve.symlinks option (i.e. setting TypeScript’s preserveSymlinks to true parallels setting Webpack’s resolve.symlinks to false, and vice-versa).

In this mode, references to modules and packages (e.g. imports and /// <reference type="..." /> directives) are all resolved relative to the location of the symbolic link file, rather than relative to the path that the symbolic link resolves to. For a more concrete example, we’ll defer to the documentation on the Node.js website.

TypeScript 2.4

Dynamic Import Expressions

Dynamic import expressions are a new feature and part of ECMAScript that allows users to asynchronously request a module at any arbitrary point in your program.

This means that you can conditionally and lazily import other modules and libraries. For example, here’s an async function that only imports a utility library when it’s needed:

async function getZipFile(name: string, files: File[]): Promise<File> { const zipUtil = await import("./utils/create-zip-file"); const zipContents = await zipUtil.getContentAsBlob(files); return new File(zipContents, name); }

Many bundlers have support for automatically splitting output bundles based on these import expressions, so consider using this new feature with the esnext module target.

String Enums

TypeScript 2.4 now allows enum members to contain string initializers.

enum Colors { Red = "RED", Green = "GREEN", Blue = "BLUE", }

The caveat is that string-initialized enums can’t be reverse-mapped to get the original enum member name. In other words, you can’t write Colors["RED"] to get the string "Red".

Improved inference for generics

TypeScript 2.4 introduces a few wonderful changes around the way generics are inferred.

Return types as inference targets

For one, TypeScript can now make inferences for the return type of a call. This can improve your experience and catch errors. Something that now works:

function arrayMap<T, U>(f: (x: T) => U): (a: T[]) => U[] { return (a) => a.map(f); } const lengths: (a: string[]) => number[] = arrayMap((s) => s.length);

As an example of new errors you might spot as a result:

let x: Promise<string> = new Promise((resolve) => { resolve(10); // ~~ Error! });

Type parameter inference from contextual types

Prior to TypeScript 2.4, in the following example

let f: <T>(x: T) => T = (y) => y;

y would have the type any. This meant the program would type-check, but you could technically do anything with y, such as the following:

let f: <T>(x: T) => T = (y) => y() + y.foo.bar;

That last example isn’t actually type-safe.

In TypeScript 2.4, the function on the right side implicitly gains type parameters, and y is inferred to have the type of that type-parameter.

If you use y in a way that the type parameter’s constraint doesn’t support, you’ll correctly get an error. In this case, the constraint of T was (implicitly) {}, so the last example will appropriately fail.

Stricter checking for generic functions

TypeScript now tries to unify type parameters when comparing two single-signature types. As a result, you’ll get stricter checks when relating two generic signatures, and may catch some bugs.

type A = <T, U>(x: T, y: U) => [T, U]; type B = <S>(x: S, y: S) => [S, S]; function f(a: A, b: B) { a = b; // Error b = a; // Ok }

Strict contravariance for callback parameters

TypeScript has always compared parameters in a bivariant way. There are a number of reasons for this, but by-and-large this was not been a huge issue for our users until we saw some of the adverse effects it had with Promises and Observables.

TypeScript 2.4 introduces tightens this up when relating two callback types. For example:

interface Mappable<T> { map<U>(f: (x: T) => U): Mappable<U>; } declare let a: Mappable<number>; declare let b: Mappable<string | number>; a = b; b = a;

Prior to TypeScript 2.4, this example would succeed. When relating the types of map, TypeScript would bidirectionally relate their parameters (i.e. the type of f). When relating each f, TypeScript would also bidirectionally relate the type of those parameters.

When relating the type of map in TS 2.4, the language will check whether each parameter is a callback type, and if so, it will ensure that those parameters are checked in a contravariant manner with respect to the current relation.

In other words, TypeScript now catches the above bug, which may be a breaking change for some users, but will largely be helpful.

Weak Type Detection

TypeScript 2.4 introduces the concept of “weak types”. Any type that contains nothing but a set of all-optional properties is considered to be weak. For example, this Options type is a weak type:

interface Options { data?: string; timeout?: number; maxRetries?: number; }

In TypeScript 2.4, it’s now an error to assign anything to a weak type when there’s no overlap in properties. For example:

function sendMessage(options: Options) { // ... } const opts = { payload: "hello world!", retryOnFail: true, }; // Error! sendMessage(opts); // No overlap between the type of 'opts' and 'Options' itself. // Maybe we meant to use 'data'/'maxRetries' instead of 'payload'/'retryOnFail'.

You can think of this as TypeScript “toughening up” the weak guarantees of these types to catch what would otherwise be silent bugs.

Since this is a breaking change, you may need to know about the workarounds which are the same as those for strict object literal checks:

  1. Declare the properties if they really do exist.
  2. Add an index signature to the weak type (i.e. [propName: string]: {}).
  3. Use a type assertion (i.e. opts as Options).

TypeScript 2.3

Generators and Iteration for ES5/ES3

First some ES2016 terminology:


ES2015 introduced Iterator, which is an object that exposes three methods, next, return, and throw, as per the following interface:

interface Iterator<T> { next(value?: any): IteratorResult<T>; return?(value?: any): IteratorResult<T>; throw?(e?: any): IteratorResult<T>; }

This kind of iterator is useful for iterating over synchronously available values, such as the elements of an Array or the keys of a Map. An object that supports iteration is said to be “iterable” if it has a Symbol.iterator method that returns an Iterator object.

The Iterator protocol also defines the target of some of the ES2015 features like for..of and spread operator and the array rest in destructuring assignmnets.


ES2015 also introduced “Generators”, which are functions that can be used to yield partial computation results via the Iterator interface and the yield keyword. Generators can also internally delegate calls to another iterable through yield *. For example:

function* f() { yield 1; yield* [2, 3]; }

New --downlevelIteration

Previously generators were only supported if the target is ES6/ES2015 or later. Moreover, constructs that operate on the Iterator protocol, e.g. for..of were only supported if they operate on arrays for targets below ES6/ES2015.

TypeScript 2.3 adds full support for generators and the Iterator protocol for ES3 and ES5 targets with --downlevelIteration flag.

With --downlevelIteration, the compiler uses new type check and emit behavior that attempts to call a [Symbol.iterator]() method on the iterated object if it is found, and creates a synthetic array iterator over the object if it is not.

Please note that this requires a native Symbol.iterator or Symbol.iterator shim at runtime for any non-array values.

for..of statements, Array Destructuring, and Spread elements in Array, Call, and New expressions support Symbol.iterator in ES5/E3 if available when using --downlevelIteration, but can be used on an Array even if it does not define Symbol.iterator at run time or design time.

Async Iteration

TypeScript 2.3 adds support for the async iterators and generators as described by the current TC39 proposal.

Async iterators

The Async Iteration introduces an AsyncIterator, which is similar to Iterator. The difference lies in the fact that the next, return, and throw methods of an AsyncIterator return a Promise for the iteration result, rather than the result itself. This allows the caller to enlist in an asynchronous notification for the time at which the AsyncIterator has advanced to the point of yielding a value. An AsyncIterator has the following shape:

interface AsyncIterator<T> { next(value?: any): Promise<IteratorResult<T>>; return?(value?: any): Promise<IteratorResult<T>>; throw?(e?: any): Promise<IteratorResult<T>>; }

An object that supports async iteration is said to be “iterable” if it has a Symbol.asyncIterator method that returns an AsyncIterator object.

Async Generators

The Async Iteration proposal introduces “Async Generators”, which are async functions that also can be used to yield partial computation results. Async Generators can also delegate calls via yield* to either an iterable or async iterable:

async function* g() { yield 1; await sleep(100); yield* [2, 3]; yield* (async function* () { await sleep(100); yield 4; })(); }

As with Generators, Async Generators can only be function declarations, function expressions, or methods of classes or object literals. Arrow functions cannot be Async Generators. Async Generators require a valid, global Promise implementation (either native or an ES2015-compatible polyfill), in addition to a valid Symbol.asyncIterator reference (either a native symbol or a shim).

The for-await-of Statement

Finally, ES2015 introduced the for..of statement as a means of iterating over an iterable. Similarly, the Async Iteration proposal introduces the for..await..of statement to iterate over an async iterable:

async function f() { for await (const x of g()) { console.log(x); } }

The for..await..of statement is only legal within an Async Function or Async Generator.


  • Keep in mind that our support for async iterators relies on support for Symbol.asyncIterator to exist at runtime. You may need to polyfill Symbol.asyncIterator, which for simple purposes can be as simple as: (Symbol as any).asyncIterator = Symbol.asyncIterator || Symbol.from("Symbol.asyncIterator");
  • You also need to include esnext in your --lib option, to get the AsyncIterator declaration if you do not already have it.
  • Finally, if your target is ES5 or ES3, you’ll also need to set the --downlevelIterators flag.

Generic parameter defaults

TypeScript 2.3 adds support for declaring defaults for generic type parameters.


Consider a function that creates a new HTMLElement, calling it with no arguments generates a Div; you can optionally pass a list of children as well. Previously you would have to define it as:

declare function create(): Container<HTMLDivElement, HTMLDivElement[]>; declare function create<T extends HTMLElement>(element: T): Container<T, T[]>; declare function create<T extends HTMLElement, U extends HTMLElement>( element: T, children: U[] ): Container<T, U[]>;

With generic parameter defaults we can reduce it to:

declare function create<T extends HTMLElement = HTMLDivElement, U = T[]>( element?: T, children?: U ): Container<T, U>;

A generic parameter default follows the following rules:

  • A type parameter is deemed optional if it has a default.
  • Required type parameters must not follow optional type parameters.
  • Default types for a type parameter must satisfy the constraint for the type parameter, if it exists.
  • When specifying type arguments, you are only required to specify type arguments for the required type parameters. Unspecified type parameters will resolve to their default types.
  • If a default type is specified and inference cannot chose a candidate, the default type is inferred.
  • A class or interface declaration that merges with an existing class or interface declaration may introduce a default for an existing type parameter.
  • A class or interface declaration that merges with an existing class or interface declaration may introduce a new type parameter as long as it specifies a default.

New --strict master option

New checks added to TypeScript are often off by default to avoid breaking existing projects. While avoiding breakage is a good thing, this strategy has the drawback of making it increasingly complex to choose the highest level of type safety, and doing so requires explicit opt-in action on every TypeScript release. With the --strict option it becomes possible to choose maximum type safety with the understanding that additional errors might be reported by newer versions of the compiler as improved type checking features are added.

The new --strict compiler option represents the recommended setting of a number of type checking options. Specifically, specifying --strict corresponds to specifying all of the following options (and may in the future include more options):

  • --strictNullChecks
  • --noImplicitAny
  • --noImplicitThis
  • --alwaysStrict

In exact terms, the --strict option sets the default value for the compiler options listed above. This means it is still possible to individually control the options. For example,

--strict --noImplicitThis false

has the effect of turning on all strict options except the --noImplicitThis option. Using this scheme it is possible to express configurations consisting of all strict options except some explicitly listed options. In other words, it is now possible to default to the highest level of type safety but opt out of certain checks.

Starting with TypeScript 2.3, the default tsconfig.json generated by tsc --init includes a "strict": true setting in the "compilerOptions" section. Thus, new projects started with tsc --init will by default have the highest level of type safety enabled.

Enhanced --init output

Along with setting --strict on by default, tsc --init has an enhanced output. Default tsconfig.json files generated by tsc --init now include a set of the common compiler options along with their descriptions commented out. Just un-comment the configuration you like to set to get the desired behavior; we hope the new output simplifies the setting up new projects and keeps configuration files readable as projects grow.

Errors in .js files with --checkJs

By default the TypeScript compiler does not report any errors in .js files including using --allowJs. With TypeScript 2.3 type-checking errors can also be reported in .js files with --checkJs.

You can skip checking some files by adding // @ts-nocheck comment to them; conversely you can choose to check only a few .js files by adding // @ts-check comment to them without setting --checkJs. You can also ignore errors on specific lines by adding // @ts-ignore on the preceding line.

.js files are still checked to ensure that they only include standard ECMAScript features; type annotations are only allowed in .ts files and are flagged as errors in .js files. JSDoc comments can be used to add some type information to your JavaScript code, see JSDoc Support documentation for more details about the supported JSDoc constructs.

See Type checking JavaScript Files documentation for more details.

TypeScript 2.2

Support for Mix-in classes

TypeScript 2.2 adds support for the ECMAScript 2015 mixin class pattern (see MDN Mixin description and “Real” Mixins with JavaScript Classes for more details) as well as rules for combining mixin construct signatures with regular construct signatures in intersection types.

First some terminology:
  • A mixin constructor type refers to a type that has a single construct signature with a single rest argument of type any[] and an object-like return type. For example, given an object-like type X, new (...args: any[]) => X is a mixin constructor type with an instance type X.
  • A mixin class is a class declaration or expression that extends an expression of a type parameter type. The following rules apply to mixin class declarations:
  • The type parameter type of the extends expression must be constrained to a mixin constructor type.
  • The constructor of a mixin class (if any) must have a single rest parameter of type any[] and must use the spread operator to pass those parameters as arguments in a super(...args) call.

Given an expression Base of a parametric type T with a constraint X, a mixin class class C extends Base {...} is processed as if Base had type X and the resulting type is the intersection typeof C & T. In other words, a mixin class is represented as an intersection between the mixin class constructor type and the parametric base class constructor type.

When obtaining the construct signatures of an intersection type that contains mixin constructor types, the mixin construct signatures are discarded and their instance types are mixed into the return types of the other construct signatures in the intersection type. For example, the intersection type { new(...args: any[]) => A } & { new(s: string) => B } has a single construct signature new(s: string) => A & B.

Putting all of the above rules together in an example:
class Point { constructor(public x: number, public y: number) {} } class Person { constructor(public name: string) {} } type Constructor<T> = new (...args: any[]) => T; function Tagged<T extends Constructor<{}>>(Base: T) { return class extends Base { _tag: string; constructor(...args: any[]) { super(...args); this._tag = ""; } }; } const TaggedPoint = Tagged(Point); let point = new TaggedPoint(10, 20); point._tag = "hello"; class Customer extends Tagged(Person) { accountBalance: number; } let customer = new Customer("Joe"); customer._tag = "test"; customer.accountBalance = 0;

Mixin classes can constrain the types of classes they can mix into by specifying a construct signature return type in the constraint for the type parameter. For example, the following WithLocation function implements a subclass factory that adds a getLocation method to any class that satisfies the Point interface (i.e. that has x and y properties of type number).

interface Point { x: number; y: number; } const WithLocation = <T extends Constructor<Point>>(Base: T) => class extends Base { getLocation(): [number, number] { return [this.x, this.y]; } };

object type

TypeScript did not have a type that represents the non-primitive type, i.e. any thing that is not number | string | boolean | symbol | null | undefined. Enter the new object type.

With object type, APIs like Object.create can be better represented. For example:

declare function create(o: object | null): void; create({ prop: 0 }); // OK create(null); // OK create(42); // Error create("string"); // Error create(false); // Error create(undefined); // Error

Support for new.target

The new.target meta-property is new syntax introduced in ES2015. When an instance of a constructor is created via new, the value of new.target is set to be a reference to the constructor function initially used to allocate the instance. If a function is called rather than constructed via new, new.target is set to undefined.

new.target comes in handy when Object.setPrototypeOf or __proto__ needs to be set in a class constructor. One such use case is inheriting from Error in NodeJS v4 and higher.


class CustomError extends Error { constructor(message?: string) { super(message); // 'Error' breaks prototype chain here Object.setPrototypeOf(this, new.target.prototype); // restore prototype chain } }

This results in the generated JS

var CustomError = (function (_super) { __extends(CustomError, _super); function CustomError() { var _newTarget = this.constructor; var _this = _super.apply(this, arguments); // 'Error' breaks prototype chain here _this.__proto__ = _newTarget.prototype; // restore prototype chain return _this; } return CustomError; })(Error);

new.target also comes in handy for writing constructable functions, for example:

function f() { if (new.target) { /* called via 'new' */ } }

Which translates to:

function f() { var _newTarget = this && this instanceof f ? this.constructor : void 0; if (_newTarget) { /* called via 'new' */ } }

Better checking for null/undefined in operands of expressions

TypeScript 2.2 improves checking of nullable operands in expressions. Specifically, these are now flagged as errors:

  • If either operand of a + operator is nullable, and neither operand is of type any or string.
  • If either operand of a -, *, **, /, %, <<, >>, >>>, &, |, or ^ operator is nullable.
  • If either operand of a <, >, <=, >=, or in operator is nullable.
  • If the right operand of an instanceof operator is nullable.
  • If the operand of a +, -, ~, ++, or -- unary operator is nullable.

An operand is considered nullable if the type of the operand is null or undefined or a union type that includes null or undefined. Note that the union type case only only occurs in --strictNullChecks mode because null and undefined disappear from unions in classic type checking mode.

Dotted property for types with string index signatures

Types with a string index signature can be indexed using the [] notation, but were not allowed to use the .. Starting with TypeScript 2.2 using either should be allowed.

interface StringMap<T> { [x: string]: T; } const map: StringMap<number>; map["prop1"] = 1; map.prop2 = 2;

This only apply to types with an explicit string index signature. It is still an error to access unknown properties on a type using . notation.

Support for spread operator on JSX element children

TypeScript 2.2 adds support for using spread on a JSX element children. Please see facebook/jsx#57 for more details.


function Todo(prop: { key: number; todo: string }) { return <div>{prop.key.toString() + prop.todo}</div>; } function TodoList({ todos }: TodoListProps) { return ( <div>{...todos.map((todo) => <Todo key={todo.id} todo={todo.todo} />)}</div> ); } let x: TodoListProps; <TodoList {...x} />;

New jsx: react-native

React-native build pipeline expects all files to have a .js extensions even if the file contains JSX syntax. The new --jsx value react-native will persevere the JSX syntax in the output file, but give it a .js extension.

TypeScript 2.1

keyof and Lookup Types

In JavaScript it is fairly common to have APIs that expect property names as parameters, but so far it hasn’t been possible to express the type relationships that occur in those APIs.

Enter Index Type Query or keyof; An indexed type query keyof T yields the type of permitted property names for T. A keyof T type is considered a subtype of string.

interface Person { name: string; age: number; location: string; } type K1 = keyof Person; // "name" | "age" | "location" type K2 = keyof Person[]; // "length" | "push" | "pop" | "concat" | ... type K3 = keyof { [x: string]: Person }; // string

The dual of this is indexed access types, also called lookup types. Syntactically, they look exactly like an element access, but are written as types:

type P1 = Person["name"]; // string type P2 = Person["name" | "age"]; // string | number type P3 = string["charAt"]; // (pos: number) => string type P4 = string[]["push"]; // (...items: string[]) => number type P5 = string[][0]; // string

You can use this pattern with other parts of the type system to get type-safe lookups.

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; } let x = { foo: 10, bar: "hello!" }; let foo = getProperty(x, "foo"); // number let bar = getProperty(x, "bar"); // string let oops = getProperty(x, "wargarbl"); // Error! "wargarbl" is not "foo" | "bar" setProperty(x, "foo", "string"); // Error!, string expected number

Mapped Types

One common task is to take an existing type and make each of its properties entirely optional. Let’s say we have a `Person:

interface Person { name: string; age: number; location: string; }

A partial version of it would be:

interface PartialPerson { name?: string; age?: number; location?: string; }

with Mapped types, PartialPerson can be written as a generalized transformation on the type Person as:

type Partial<T> = { [P in keyof T]?: T[P]; }; type PartialPerson = Partial<Person>;

Mapped types are produced by taking a union of literal types, and computing a set of properties for a new object type. They’re like list comprehensions in Python, but instead of producing new elements in a list, they produce new properties in a type.

In addition to Partial, Mapped Types can express many useful transformations on types:

// Keep types the same, but make each property to be read-only. type Readonly<T> = { readonly [P in keyof T]: T[P]; }; // Same property names, but make the value a promise instead of a concrete one type Deferred<T> = { [P in keyof T]: Promise<T[P]>; }; // Wrap proxies around properties of T type Proxify<T> = { [P in keyof T]: { get(): T[P]; set(v: T[P]): void }; };

Partial, Readonly, Record, and Pick

Partial and Readonly, as described earlier, are very useful constructs. You can use them to describe some common JS routines like:

function assign<T>(obj: T, props: Partial<T>): void; function freeze<T>(obj: T): Readonly<T>;

Because of that, they are now included by default in the standard library.

We’re also including two other utility types as well: Record and Pick.

// From T pick a set of properties K declare function pick<T, K extends keyof T>(obj: T, ...keys: K[]): Pick<T, K>; const nameAndAgeOnly = pick(person, "name", "age"); // { name: string, age: number }
// For every properties K of type T, transform it to U function mapObject<K extends string, T, U>( obj: Record<K, T>, f: (x: T) => U ): Record<K, U>; const names = { foo: "hello", bar: "world", baz: "bye" }; const lengths = mapObject(names, (s) => s.length); // { foo: number, bar: number, baz: number }

Object Spread and Rest

TypeScript 2.1 brings support for ES2017 Spread and Rest.

Similar to array spread, spreading an object can be handy to get a shallow copy:

let copy = { ...original };

Similarly, you can merge several different objects. In the following example, merged will have properties from foo, bar, and baz.

let merged = { ...foo, ...bar, ...baz };

You can also override existing properties and add new ones:

let obj = { x: 1, y: "string" }; var newObj = { ...obj, z: 3, y: 4 }; // { x: number, y: number, z: number }

The order of specifying spread operations determines what properties end up in the resulting object; properties in later spreads “win out” over previously created properties.

Object rests are the dual of object spreads, in that they can extract any extra properties that don’t get picked up when destructuring an element:

let obj = { x: 1, y: 1, z: 1 }; let { z, ...obj1 } = obj; obj1; // {x: number, y: number};

Downlevel Async Functions

This feature was supported before TypeScript 2.1, but only when targeting ES6/ES2015. TypeScript 2.1 brings the capability to ES3 and ES5 run-times, meaning you’ll be free to take advantage of it no matter what environment you’re using.

Note: first, we need to make sure our run-time has an ECMAScript-compliant Promise available globally. That might involve grabbing a polyfill for Promise, or relying on one that you might have in the run-time that you’re targeting. We also need to make sure that TypeScript knows Promise exists by setting your lib flag to something like "dom", "es2015" or "dom", "es2015.promise", "es5"

{ "compilerOptions": { "lib": ["dom", "es2015.promise", "es5"] } }
function delay(milliseconds: number) { return new Promise<void>((resolve) => { setTimeout(resolve, milliseconds); }); } async function dramaticWelcome() { console.log("Hello"); for (let i = 0; i < 3; i++) { await delay(500); console.log("."); } console.log("World!"); } dramaticWelcome();

Compiling and running the output should result in the correct behavior on an ES3/ES5 engine.

Support for external helpers library (tslib)

TypeScript injects a handful of helper functions such as __extends for inheritance, __assign for spread operator in object literals and JSX elements, and __awaiter for async functions.

Previously there were two options:

  1. inject helpers in every file that needs them, or
  2. no helpers at all with --noEmitHelpers.

The two options left more to be desired; bundling the helpers in every file was a pain point for customers trying to keep their package size small. And not including helpers, meant customers had to maintain their own helpers library.

TypeScript 2.1 allows for including these files in your project once in a separate module, and the compiler will emit imports to them as needed.

First, install the tslib utility library:

npm install tslib

Second, compile your files using --importHelpers:

tsc --module commonjs --importHelpers a.ts

So given the following input, the resulting .js file will include an import to tslib and use the __assign helper from it instead of inlining it.

export const o = { a: 1, name: "o" }; export const copy = { ...o };
"use strict"; var tslib_1 = require("tslib"); exports.o = { a: 1, name: "o" }; exports.copy = tslib_1.__assign({}, exports.o);

Untyped imports

TypeScript has traditionally been overly strict about how you can import modules. This was to avoid typos and prevent users from using modules incorrectly.

However, a lot of the time, you might just want to import an existing module that may not have its own .d.ts file. Previously this was an error. Starting with TypeScript 2.1 this is now much easier.

With TypeScript 2.1, you can import a JavaScript module without needing a type declaration. A type declaration (such as declare module "foo" { ... } or node_modules/@types/foo) still takes priority if it exists.

An import to a module with no declaration file will still be flagged as an error under --noImplicitAny.

// Succeeds if `node_modules/asdf/index.js` exists, or if `node_modules/asdf/package.json` defines a valid "main" entry point import { x } from "asdf";

Support for --target ES2016, --target ES2017 and --target ESNext

TypeScript 2.1 supports three new target values --target ES2016, --target ES2017 and --target ESNext.

Using target --target ES2016 will instruct the compiler not to transform ES2016-specific features, e.g. ** operator.

Similarly, --target ES2017 will instruct the compiler not to transform ES2017-specific features like async/await.

--target ESNext targets latest supported ES proposed features.

Improved any Inference

Previously, if TypeScript couldn’t figure out the type of a variable, it would choose the any type.

let x; // implicitly 'any' let y = []; // implicitly 'any[]' let z: any; // explicitly 'any'.

With TypeScript 2.1, instead of just choosing any, TypeScript will infer types based on what you end up assigning later on.

This is only enabled if --noImplicitAny is set.

let x; // You can still assign anything you want to 'x'. x = () => 42; // After that last assignment, TypeScript 2.1 knows that 'x' has type '() => number'. let y = x(); // Thanks to that, it will now tell you that you can't add a number to a function! console.log(x + y); // ~~~~~ // Error! Operator '+' cannot be applied to types '() => number' and 'number'. // TypeScript still allows you to assign anything you want to 'x'. x = "Hello world!"; // But now it also knows that 'x' is a 'string'! x.toLowerCase();

The same sort of tracking is now also done for empty arrays.

A variable declared with no type annotation and an initial value of [] is considered an implicit any[] variable. However, each subsequent x.push(value), x.unshift(value) or x[n] = value operation evolves the type of the variable in accordance with what elements are added to it.

function f1() { let x = []; x.push(5); x[1] = "hello"; x.unshift(true); return x; // (string | number | boolean)[] } function f2() { let x = null; if (cond()) { x = []; while (cond()) { x.push("hello"); } } return x; // string[] | null }

Implicit any errors

One great benefit of this is that you’ll see way fewer implicit any errors when running with --noImplicitAny. Implicit any errors are only reported when the compiler is unable to know the type of a variable without a type annotation.

function f3() { let x = []; // Error: Variable 'x' implicitly has type 'any[]' in some locations where its type cannot be determined. x.push(5); function g() { x; // Error: Variable 'x' implicitly has an 'any[]' type. } }

Better inference for literal types

String, numeric and boolean literal types (e.g. "abc", 1, and true) were previously inferred only in the presence of an explicit type annotation. Starting with TypeScript 2.1, literal types are always inferred for const variables and readonly properties.

The type inferred for a const variable or readonly property without a type annotation is the type of the literal initializer. The type inferred for a let variable, var variable, parameter, or non-readonly property with an initializer and no type annotation is the widened literal type of the initializer. Where the widened type for a string literal type is string, number for numeric literal types, boolean for true or false and the containing enum for enum literal types.

const c1 = 1; // Type 1 const c2 = c1; // Type 1 const c3 = "abc"; // Type "abc" const c4 = true; // Type true const c5 = cond ? 1 : "abc"; // Type 1 | "abc" let v1 = 1; // Type number let v2 = c2; // Type number let v3 = c3; // Type string let v4 = c4; // Type boolean let v5 = c5; // Type number | string

Literal type widening can be controlled through explicit type annotations. Specifically, when an expression of a literal type is inferred for a const location without a type annotation, that const variable gets a widening literal type inferred. But when a const location has an explicit literal type annotation, the const variable gets a non-widening literal type.

const c1 = "hello"; // Widening type "hello" let v1 = c1; // Type string const c2: "hello" = "hello"; // Type "hello" let v2 = c2; // Type "hello"

Use returned values from super calls as ‘this’

In ES2015, constructors which return an object implicitly substitute the value of this for any callers of super(). As a result, it is necessary to capture any potential return value of super() and replace it with this. This change enables working with Custom Elements, which takes advantage of this to initialize browser-allocated elements with user-written constructors.

class Base { x: number; constructor() { // return a new object other than `this` return { x: 1, }; } } class Derived extends Base { constructor() { super(); this.x = 2; } }


var Derived = (function (_super) { __extends(Derived, _super); function Derived() { var _this = _super.call(this) || this; _this.x = 2; return _this; } return Derived; })(Base);

This change entails a break in the behavior of extending built-in classes like Error, Array, Map, etc.. Please see the extending built-ins breaking change documentation for more details.

Configuration inheritance

Often a project has multiple output targets, e.g. ES5 and ES2015, debug and production or CommonJS and System; Just a few configuration options change between these two targets, and maintaining multiple tsconfig.json files can be a hassle.

TypeScript 2.1 supports inheriting configuration using extends, where:

  • extends is a new top-level property in tsconfig.json (alongside compilerOptions, files, include, and exclude).
  • The value of extends must be a string containing a path to another configuration file to inherit from.
  • The configuration from the base file are loaded first, then overridden by those in the inheriting config file.
  • Circularity between configuration files is not allowed.
  • files, include and exclude from the inheriting config file overwrite those from the base config file.
  • All relative paths found in the configuration file will be resolved relative to the configuration file they originated in.


{ "compilerOptions": { "noImplicitAny": true, "strictNullChecks": true } }


{ "extends": "./configs/base", "files": ["main.ts", "supplemental.ts"] }


{ "extends": "./tsconfig", "compilerOptions": { "strictNullChecks": false } }

New --alwaysStrict

Invoking the compiler with --alwaysStrict causes:

  1. Parses all the code in strict mode.
  2. Writes "use strict"; directive atop every generated file.

Modules are parsed automatically in strict mode. The new flag is recommended for non-module code.

TypeScript 2.0

Null- and undefined-aware types

TypeScript has two special types, Null and Undefined, that have the values null and undefined respectively. Previously it was not possible to explicitly name these types, but null and undefined may now be used as type names regardless of type checking mode.

The type checker previously considered null and undefined assignable to anything. Effectively, null and undefined were valid values of every type and it wasn’t possible to specifically exclude them (and therefore not possible to detect erroneous use of them).


--strictNullChecks switches to a new strict null checking mode.

In strict null checking mode, the null and undefined values are not in the domain of every type and are only assignable to themselves and any (the one exception being that undefined is also assignable to void). So, whereas T and T | undefined are considered synonymous in regular type checking mode (because undefined is considered a subtype of any T), they are different types in strict type checking mode, and only T | undefined permits undefined values. The same is true for the relationship of T to T | null.


// Compiled with --strictNullChecks let x: number; let y: number | undefined; let z: number | null | undefined; x = 1; // Ok y = 1; // Ok z = 1; // Ok x = undefined; // Error y = undefined; // Ok z = undefined; // Ok x = null; // Error y = null; // Error z = null; // Ok x = y; // Error x = z; // Error y = x; // Ok y = z; // Error z = x; // Ok z = y; // Ok

Assigned-before-use checking

In strict null checking mode the compiler requires every reference to a local variable of a type that doesn’t include undefined to be preceded by an assignment to that variable in every possible preceding code path.


// Compiled with --strictNullChecks let x: number; let y: number | null; let z: number | undefined; x; // Error, reference not preceded by assignment y; // Error, reference not preceded by assignment z; // Ok x = 1; y = null; x; // Ok y; // Ok

The compiler checks that variables are definitely assigned by performing control flow based type analysis. See later for further details on this topic.

Optional parameters and properties

Optional parameters and properties automatically have undefined added to their types, even when their type annotations don’t specifically include undefined. For example, the following two types are identical:

// Compiled with --strictNullChecks type T1 = (x?: number) => string; // x has type number | undefined type T2 = (x?: number | undefined) => string; // x has type number | undefined

Non-null and non-undefined type guards

A property access or a function call produces a compile-time error if the object or function is of a type that includes null or undefined. However, type guards are extended to support non-null and non-undefined checks.


// Compiled with --strictNullChecks declare function f(x: number): string; let x: number | null | undefined; if (x) { f(x); // Ok, type of x is number here } else { f(x); // Error, type of x is number? here } let a = x != null ? f(x) : ""; // Type of a is string let b = x && f(x); // Type of b is string | 0 | null | undefined

Non-null and non-undefined type guards may use the ==, !=, ===, or !== operator to compare to null or undefined, as in x != null or x === undefined. The effects on subject variable types accurately reflect JavaScript semantics (e.g. double-equals operators check for both values no matter which one is specified whereas triple-equals only checks for the specified value).

Dotted names in type guards

Type guards previously only supported checking local variables and parameters. Type guards now support checking “dotted names” consisting of a variable or parameter name followed one or more property accesses.


interface Options { location?: { x?: number; y?: number; }; } function foo(options?: Options) { if (options && options.location && options.location.x) { const x = options.location.x; // Type of x is number } }

Type guards for dotted names also work with user defined type guard functions and the typeof and instanceof operators and do not depend on the --strictNullChecks compiler option.

A type guard for a dotted name has no effect following an assignment to any part of the dotted name. For example, a type guard for x.y.z will have no effect following an assignment to x, x.y, or x.y.z.

Expression operators

Expression operators permit operand types to include null and/or undefined but always produce values of non-null and non-undefined types.

// Compiled with --strictNullChecks function sum(a: number | null, b: number | null) { return a + b; // Produces value of type number }

The && operator adds null and/or undefined to the type of the right operand depending on which are present in the type of the left operand, and the || operator removes both null and undefined from the type of the left operand in the resulting union type.

// Compiled with --strictNullChecks interface Entity { name: string; } let x: Entity | null; let s = x && x.name; // s is of type string | null let y = x || { name: "test" }; // y is of type Entity

Type widening

The null and undefined types are not widened to any in strict null checking mode.

let z = null; // Type of z is null

In regular type checking mode the inferred type of z is any because of widening, but in strict null checking mode the inferred type of z is null (and therefore, absent a type annotation, null is the only possible value for z).

Non-null assertion operator

A new ! post-fix expression operator may be used to assert that its operand is non-null and non-undefined in contexts where the type checker is unable to conclude that fact. Specifically, the operation x! produces a value of the type of x with null and undefined excluded. Similar to type assertions of the forms <T>x and x as T, the ! non-null assertion operator is simply removed in the emitted JavaScript code.

// 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 }


The new features are designed such that they can be used in both strict null checking mode and regular type checking mode. In particular, the null and undefined types are automatically erased from union types in regular type checking mode (because they are subtypes of all other types), and the ! non-null assertion expression operator is permitted but has no effect in regular type checking mode. Thus, declaration files that are updated to use null- and undefined-aware types can still be used in regular type checking mode for backwards compatibility.

In practical terms, strict null checking mode requires that all files in a compilation are null- and undefined-aware.

Control flow based type analysis

TypeScript 2.0 implements a control flow-based type analysis for local variables and parameters. Previously, the type analysis performed for type guards was limited to if statements and ?: conditional expressions and didn’t include effects of assignments and control flow constructs such as return and break statements. With TypeScript 2.0, the type checker analyses all possible flows of control in statements and expressions to produce the most specific type possible (the narrowed type) at any given location for a local variable or parameter that is declared to have a union type.


function foo(x: string | number | boolean) { if (typeof x === "string") { x; // type of x is string here x = 1; x; // type of x is number here } x; // type of x is number | boolean here } function bar(x: string | number) { if (typeof x === "number") { return; } x; // type of x is string here }

Control flow based type analysis is particuarly relevant in --strictNullChecks mode because nullable types are represented using union types:

function test(x: string | null) { if (x === null) { return; } x; // type of x is string in remainder of function }

Furthermore, in --strictNullChecks mode, control flow based type analysis includes definite assignment analysis for local variables of types that don’t permit the value undefined.

function mumble(check: boolean) { let x: number; // Type doesn't permit undefined x; // Error, x is undefined if (check) { x = 1; x; // Ok } x; // Error, x is possibly undefined x = 2; x; // Ok }

Tagged union types

TypeScript 2.0 implements support for tagged (or discriminated) union types. Specifically, the TS compiler now support type guards that narrow union types based on tests of a discriminant property and furthermore extend that capability to switch statements.


interface Square { kind: "square"; size: number; } interface Rectangle { kind: "rectangle"; width: number; height: number; } interface Circle { kind: "circle"; radius: number; } type Shape = Square | Rectangle | Circle; function area(s: Shape) { // In the following switch statement, the type of s is narrowed in each case clause // according to the value of the discriminant property, thus allowing the other properties // of that variant to be accessed without a type assertion. switch (s.kind) { case "square": return s.size * s.size; case "rectangle": return s.width * s.height; case "circle": return Math.PI * s.radius * s.radius; } } function test1(s: Shape) { if (s.kind === "square") { s; // Square } else { s; // Rectangle | Circle } } function test2(s: Shape) { if (s.kind === "square" || s.kind === "rectangle") { return; } s; // Circle }

A discriminant property type guard is an expression of the form x.p == v, x.p === v, x.p != v, or x.p !== v, where p and v are a property and an expression of a string literal type or a union of string literal types. The discriminant property type guard narrows the type of x to those constituent types of x that have a discriminant property p with one of the possible values of v.

Note that we currently only support discriminant properties of string literal types. We intend to later add support for boolean and numeric literal types.

The never type

TypeScript 2.0 introduces a new primitive type never. The never type represents the type of values that never occur. Specifically, never is the return type for functions that never return and never is the type of variables under type guards that are never true.

The never type has the following characteristics:

  • never is a subtype of and assignable to every type.
  • No type is a subtype of or assignable to never (except never itself).
  • In a function expression or arrow function with no return type annotation, if the function has no return statements, or only return statements with expressions of type never, and if the end point of the function is not reachable (as determined by control flow analysis), the inferred return type for the function is never.
  • In a function with an explicit never return type annotation, all return statements (if any) must have expressions of type never and the end point of the function must not be reachable.

Because never is a subtype of every type, it is always omitted from union types and it is ignored in function return type inference as long as there are other types being returned.

Some examples of functions returning never:

// Function returning never must have unreachable end point function error(message: string): never { throw new Error(message); } // Inferred return type is never function fail() { return error("Something failed"); } // Function returning never must have unreachable end point function infiniteLoop(): never { while (true) {} }

Some examples of use of functions returning never:

// Inferred return type is number function move1(direction: "up" | "down") { switch (direction) { case "up": return 1; case "down": return -1; } return error("Should never get here"); } // Inferred return type is number function move2(direction: "up" | "down") { return direction === "up" ? 1 : direction === "down" ? -1 : error("Should never get here"); } // Inferred return type is T function check<T>(x: T | undefined) { return x || error("Undefined value"); }

Because never is assignable to every type, a function returning never can be used when a callback returning a more specific type is required:

function test(cb: () => string) { let s = cb(); return s; } test(() => "hello"); test(() => fail()); test(() => { throw new Error(); });

Read-only properties and index signatures

A property or index signature can now be declared with the readonly modifier is considered read-only.

Read-only properties may have initializers and may be assigned to in constructors within the same class declaration, but otherwise assignments to read-only properties are disallowed.

In addition, entities are implicitly read-only in several situations:

  • A property declared with a get accessor and no set accessor is considered read-only.
  • In the type of an enum object, enum members are considered read-only properties.
  • In the type of a module object, exported const variables are considered read-only properties.
  • An entity declared in an import statement is considered read-only.
  • An entity accessed through an ES2015 namespace import is considered read-only (e.g. foo.x is read-only when foo is declared as import * as foo from "foo").


interface Point { readonly x: number; readonly y: number; } var p1: Point = { x: 10, y: 20 }; p1.x = 5; // Error, p1.x is read-only var p2 = { x: 1, y: 1 }; var p3: Point = p2; // Ok, read-only alias for p2 p3.x = 5; // Error, p3.x is read-only p2.x = 5; // Ok, but also changes p3.x because of aliasing
class Foo { readonly a = 1; readonly b: string; constructor() { this.b = "hello"; // Assignment permitted in constructor } }
let a: Array<number> = [0, 1, 2, 3, 4]; let b: ReadonlyArray<number> = a; b[5] = 5; // Error, elements are read-only b.push(5); // Error, no push method (because it mutates array) b.length = 3; // Error, length is read-only a = b; // Error, mutating methods are missing

Specifying the type of this for functions

Following up on specifying the type of this in a class or an interface, functions and methods can now declare the type of this they expect.

By default the type of this inside a function is any. Starting with TypeScript 2.0, you can provide an explicit this parameter. this parameters are fake parameters that come first in the parameter list of a function:

function f(this: void) { // make sure `this` is unusable in this standalone function }

this parameters in callbacks

Libraries can also use this parameters to declare how callbacks will be invoked.


interface UIElement { addClickListener(onclick: (this: void, e: Event) => void): void; }

this: void means that addClickListener expects onclick to be a function that does not require a this type.

Now if you annotate calling code with this:

class Handler { info: string; onClickBad(this: Handler, e: Event) { // oops, used this here. using this callback would crash at runtime this.info = e.message; } } let h = new Handler(); uiElement.addClickListener(h.onClickBad); // error!


A new flag is also added in TypeScript 2.0 to flag all uses of this in functions without an explicit type annotation.

Glob support in tsconfig.json

Glob support is here!! Glob support has been one of the most requested features.

Glob-like file patterns are supported two properties "include" and "exclude".


{ "compilerOptions": { "module": "commonjs", "noImplicitAny": true, "removeComments": true, "preserveConstEnums": true, "outFile": "../../built/local/tsc.js", "sourceMap": true }, "include": ["src/**/*"], "exclude": ["node_modules", "**/*.spec.ts"] }

The supported glob wildcards are:

  • * matches zero or more characters (excluding directory separators)
  • ? matches any one character (excluding directory separators)
  • **/ recursively matches any subdirectory

If a segment of a glob pattern includes only * or .*, then only files with supported extensions are included (e.g. .ts, .tsx, and .d.ts by default with .js and .jsx if allowJs is set to true).

If the "files" and "include" are both left unspecified, the compiler defaults to including all TypeScript (.ts, .d.ts and .tsx) files in the containing directory and subdirectories except those excluded using the "exclude" property. JS files (.js and .jsx) are also included if allowJs is set to true.

If the "files" or "include" properties are specified, the compiler will instead include the union of the files included by those two properties. Files in the directory specified using the "outDir" compiler option are always excluded unless explicitly included via the "files" property (even when the ”exclude” property is specified).

Files included using "include" can be filtered using the "exclude" property. However, files included explicitly using the "files" property are always included regardless of "exclude". The "exclude" property defaults to excluding the node_modules, bower_components, and jspm_packages directories when not specified.

Module resolution enhancements: BaseUrl, Path mapping, rootDirs and tracing

TypeScript 2.0 provides a set of additional module resolution knops to inform the compiler where to find declarations for a given module.

See Module Resolution documentation for more details.

Base URL

Using a baseUrl is a common practice in applications using AMD module loaders where modules are “deployed” to a single folder at run-time. All module imports with non-relative names are assumed to be relative to the baseUrl.


{ "compilerOptions": { "baseUrl": "./modules" } }

Now imports to "moduleA" would be looked up in ./modules/moduleA

import A from "moduleA";

Path mapping

Sometimes modules are not directly located under baseUrl. Loaders use a mapping configuration to map module names to files at run-time, see RequireJs documentation and SystemJS documentation.

The TypeScript compiler supports the declaration of such mappings using "paths" property in tsconfig.json files.


For instance, an import to a module "jquery" would be translated at runtime to "node_modules/jquery/dist/jquery.slim.min.js".

{ "compilerOptions": { "baseUrl": "./node_modules", "paths": { "jquery": ["jquery/dist/jquery.slim.min"] } }

Using "paths" also allow for more sophisticated mappings including multiple fall back locations. Consider a project configuration where only some modules are available in one location, and the rest are in another.

Virtual Directories with rootDirs

Using ‘rootDirs’, you can inform the compiler of the roots making up this “virtual” directory; and thus the compiler can resolve relative modules imports within these “virtual” directories as if were merged together in one directory.


Given this project structure:

 └── views
     └── view1.ts (imports './template1')
     └── view2.ts

 └── templates
         └── views
             └── template1.ts (imports './view2')

A build step will copy the files in /src/views and /generated/templates/views to the same directory in the output. At run-time, a view can expect its template to exist next to it, and thus should import it using a relative name as "./template".

"rootDirs" specify a list of roots whose contents are expected to merge at run-time. So following our example, the tsconfig.json file should look like:

{ "compilerOptions": { "rootDirs": ["src/views", "generated/templates/views"] } }

Tracing module resolution

--traceResolution offers a handy way to understand how modules have been resolved by the compiler.

tsc --traceResolution

Shorthand ambient module declarations

If you don’t want to take the time to write out declarations before using a new module, you can now just use a shorthand declaration to get started quickly.


declare module "hot-new-module";

All imports from a shorthand module will have the any type.

import x, { y } from "hot-new-module"; x(y);

Wildcard character in module names

Importing none-code resources using module loaders extension (e.g. AMD or SystemJS) has not been easy before; previously an ambient module declaration had to be defined for each resource.

TypeScript 2.0 supports the use of the wildcard character (*) to declare a “family” of module names; this way, a declaration is only required once for an extension, and not for every resource.


declare module "*!text" { const content: string; export default content; } // Some do it the other way around. declare module "json!*" { const value: any; export default value; }

Now you can import things that match "*!text" or "json!*".

import fileContent from "./xyz.txt!text"; import data from "json!http://example.com/data.json"; console.log(data, fileContent);

Wildcard module names can be even more useful when migrating from an un-typed code base. Combined with Shorthand ambient module declarations, a set of modules can be easily declared as any.


declare module "myLibrary/*";

All imports to any module under myLibrary would be considered to have the type any by the compiler; thus, shutting down any checking on the shapes or types of these modules.

import { readFile } from "myLibrary/fileSystem/readFile"; readFile(); // readFile is 'any'

Support for UMD module definitions

Some libraries are designed to be used in many module loaders, or with no module loading (global variables). These are known as UMD or Isomorphic modules. These libraries can be accessed through either an import or a global variable.

For example:

export const isPrime(x: number): boolean; export as namespace mathLib;

The library can then be used as an import within modules:

import { isPrime } from "math-lib"; isPrime(2); mathLib.isPrime(2); // ERROR: can't use the global definition from inside a module

It can also be used as a global variable, but only inside of a script. (A script is a file with no imports or exports.)


Optional class properties

Optional properties and methods can now be declared in classes, similar to what is already permitted in interfaces.


class Bar { a: number; b?: number; f() { return 1; } g?(): number; // Body of optional method can be omitted h?() { return 2; } }

When compiled in --strictNullChecks mode, optional properties and methods automatically have undefined included in their type. Thus, the b property above is of type number | undefined and the g method above is of type (() => number) | undefined. Type guards can be used to strip away the undefined part of the type:

function test(x: Bar) { x.a; // number x.b; // number | undefined x.f; // () => number x.g; // (() => number) | undefined let f1 = x.f(); // number let g1 = x.g && x.g(); // number | undefined let g2 = x.g ? x.g() : 0; // number }

Private and Protected Constructors

A class constructor may be marked private or protected. A class with private constructor cannot be instantiated outside the class body, and cannot be extended. A class with protected constructor cannot be instantiated outside the class body, but can be extended.


class Singleton { private static instance: Singleton; private constructor() {} static getInstance() { if (!Singleton.instance) { Singleton.instance = new Singleton(); } return Singleton.instance; } } let e = new Singleton(); // Error: constructor of 'Singleton' is private. let v = Singleton.getInstance();

Abstract properties and accessors

An abstract class can declare abstract properties and/or accessors. Any sub class will need to declare the abstract properties or be marked as abstract. Abstract properties cannot have an initializer. Abstract accessors cannot have bodies.


abstract class Base { abstract name: string; abstract get value(); abstract set value(v: number); } class Derived extends Base { name = "derived"; value = 1; }

Implicit index signatures

An object literal type is now assignable to a type with an index signature if all known properties in the object literal are assignable to that index signature. This makes it possible to pass a variable that was initialized with an object literal as parameter to a function that expects a map or dictionary:

function httpService(path: string, headers: { [x: string]: string }) {} const headers = { "Content-Type": "application/x-www-form-urlencoded", }; httpService("", { "Content-Type": "application/x-www-form-urlencoded" }); // Ok httpService("", headers); // Now ok, previously wasn't

Including built-in type declarations with --lib

Getting to ES6/ES2015 built-in API declarations were only limited to target: ES6. Enter --lib; with --lib you can specify a list of built-in API declaration groups that you can chose to include in your project. For instance, if you expect your runtime to have support for Map, Set and Promise (e.g. most evergreen browsers today), just include --lib es2015.collection,es2015.promise. Similarly you can exclude declarations you do not want to include in your project, e.g. DOM if you are working on a node project using --lib es5,es6.

Here is a list of available API groups:

  • dom
  • webworker
  • es5
  • es6 / es2015
  • es2015.core
  • es2015.collection
  • es2015.iterable
  • es2015.promise
  • es2015.proxy
  • es2015.reflect
  • es2015.generator
  • es2015.symbol
  • es2015.symbol.wellknown
  • es2016
  • es2016.array.include
  • es2017
  • es2017.object
  • es2017.sharedmemory
  • scripthost


tsc --target es5 --lib es5,es2015.promise
"compilerOptions": { "lib": ["es5", "es2015.promise"] }

Flag unused declarations with --noUnusedParameters and --noUnusedLocals

TypeScript 2.0 has two new flags to help you maintain a clean code base. --noUnusedParameters flags any unused function or method parameters errors. --noUnusedLocals flags any unused local (un-exported) declaration like variables, functions, classes, imports, etc… Also, unused private members of a class would be flagged as errors under --noUnusedLocals.


import B, { readFile } from "./b"; // ^ Error: `B` declared but never used readFile(); export function write(message: string, args: string[]) { // ^^^^ Error: 'arg' declared but never used. console.log(message); }

Parameters declaration with names starting with _ are exempt from the unused parameter checking. e.g.:

function returnNull(_a) { // OK return null; }

Module identifiers allow for .js extension

Before TypeScript 2.0, a module identifier was always assumed to be extension-less; for instance, given an import as import d from "./moduleA.js", the compiler looked up the definition of "moduleA.js" in ./moduleA.js.ts or ./moduleA.js.d.ts. This made it hard to use bundling/loading tools like SystemJS that expect URI’s in their module identifier.

With TypeScript 2.0, the compiler will look up definition of "moduleA.js" in ./moduleA.ts or ./moduleA.d.ts.

Support ‘target : es5’ with ‘module: es6’

Previously flagged as an invalid flag combination, target: es5 and ‘module: es6’ is now supported. This should facilitate using ES2015-based tree shakers like rollup.

Trailing commas in function parameter and argument lists

Trailing comma in function parameter and argument lists are now allowed. This is an implementation for a Stage-3 ECMAScript proposal that emits down to valid ES3/ES5/ES6.