async function

An async function is a function declared with the async keyword. Async functions are instances of the AsyncFunction constructor, and the await keyword is permitted within them. The async and await keywords enable asynchronous, promise-based behavior to be written in a cleaner style, avoiding the need to explicitly configure promise chains.

Async functions may also be defined as expressions.

Syntax

async function name([param[, param[, ...param]]]) {
   statements
}

Parameters

name
The function’s name.
param
The name of an argument to be passed to the function.
statements
The statements comprising the body of the function.  The await mechanism may be used.

Return value

A Promise which will be resolved with the value returned by the async function, or rejected with an exception thrown from, or uncaught within, the async function.

Description

Async functions can contain zero or more await expressions. Await expressions suspend progress through an async function, yielding control and subsequently resuming progress only when an awaited promise-based asynchronous operation is either fulfilled or rejected. The resolved value of the promise is treated as the return value of the await expression. Use of async / await enables the use of ordinary try / catch blocks around asynchronous code.

The await keyword is only valid inside async functions. If you use it outside of an async function's body, you will get a SyntaxError.

The purpose of async/await is to simplify the syntax necessary to consume promise-based APIs. The behavior of async/await is similar to combining generators and promises.

Async functions always return a promise. If the return value of an async function is not explicitly a promise, it will be implicitly wrapped in a promise.

For example, the following:

async function foo() {
   return 1
}

...is equivalent to:

function foo() {
   return Promise.resolve(1)
}

The body of an async function can be thought of as being split by zero or more await expressions. Top-level code, up to and including the first await expression (if there is one), is run synchronously. In this way, an async function without an await expression will run synchronously. If there is an await expression inside the function body, however, the async function will always complete asynchronously.

For example:

async function foo() {
   await 1
}

...is equivalent to:

function foo() {
   return Promise.resolve(1).then(() => undefined)
}

Code after each await expression can be thought of as existing in a .then callback. In this way a promise chain is progressively constructed with each reentrant step through the function. The return value forms the final link in the chain.

In the following example, we successively await two promises. Progress moves through function foo in three stages.

  1. The first line of the body of function foo is executed synchronously, with the await expression configured with the pending promise. Progress through foo is then suspended and control is yielded back to the function that called foo.
  2. Some time later, when the first promise has either been fulfilled or rejected, control moves back into foo. The result of the first promise fulfilment (if it was not rejected) is returned from the await expression. Here 1 is assigned to result1. Progress continues, and the second await expression is evaluated. Again, progress through foo is suspended and control is yielded.
  3. Some time later, when the second promise has either been fulfilled or rejected, control re-enters foo. The result of the second promise resolution is returned from the second await expression. Here 2 is assigned to result2. Control moves to the return expression (if any). The default return value of undefined is returned as the resolution value of the current promise.
async function foo() {
   const result1 = await new Promise((resolve) => setTimeout(() => resolve('1')))
   const result2 = await new Promise((resolve) => setTimeout(() => resolve('2')))
}
foo()

Note how the promise chain is not built-up in one go. Instead, the promise chain is constructed in stages as control is successively yielded from and returned to the async function. As a result, we must be mindful of error handling behavior when dealing with concurrent asynchronous operations.

For example, in the following code an unhandled promise rejection error will be thrown, even if a .catch handler has been configured further along the promise chain. This is because p2 will not be "wired into" the promise chain until control returns from p1.

async function foo() {
   const p1 = new Promise((resolve) => setTimeout(() => resolve('1'), 1000))
   const p2 = new Promise((_,reject) => setTimeout(() => reject('2'), 500))   
   const results = [await p1, await p2] // Do not do this! Use Promise.all or Promise.allSettled instead.
}
foo().catch(() => {}) // Attempt to swallow all errors...

Examples

Async functions and execution order

function resolveAfter2Seconds() {
  console.log("starting slow promise")
  return new Promise(resolve => {
    setTimeout(function() {
      resolve("slow")
      console.log("slow promise is done")
    }, 2000)
  })
}

function resolveAfter1Second() {
  console.log("starting fast promise")
  return new Promise(resolve => {
    setTimeout(function() {
      resolve("fast")
      console.log("fast promise is done")
    }, 1000)
  })
}

async function sequentialStart() {
  console.log('==SEQUENTIAL START==')

  // 1. Execution gets here almost instantly
  const slow = await resolveAfter2Seconds()
  console.log(slow) // 2. this runs 2 seconds after 1.

  const fast = await resolveAfter1Second()
  console.log(fast) // 3. this runs 3 seconds after 1.
}

async function concurrentStart() {
  console.log('==CONCURRENT START with await==');
  const slow = resolveAfter2Seconds() // starts timer immediately
  const fast = resolveAfter1Second() // starts timer immediately

  // 1. Execution gets here almost instantly
  console.log(await slow) // 2. this runs 2 seconds after 1.
  console.log(await fast) // 3. this runs 2 seconds after 1., immediately after 2., since fast is already resolved
}

function concurrentPromise() {
  console.log('==CONCURRENT START with Promise.all==')
  return Promise.all([resolveAfter2Seconds(), resolveAfter1Second()]).then((messages) => {
    console.log(messages[0]) // slow
    console.log(messages[1]) // fast
  })
}

async function parallel() {
  console.log('==PARALLEL with await Promise.all==')
  
  // Start 2 "jobs" in parallel and wait for both of them to complete
  await Promise.all([
      (async()=>console.log(await resolveAfter2Seconds()))(),
      (async()=>console.log(await resolveAfter1Second()))()
  ])
}

sequentialStart() // after 2 seconds, logs "slow", then after 1 more second, "fast"

// wait above to finish
setTimeout(concurrentStart, 4000) // after 2 seconds, logs "slow" and then "fast"

// wait again
setTimeout(concurrentPromise, 7000) // same as concurrentStart

// wait again
setTimeout(parallel, 10000) // truly parallel: after 1 second, logs "fast", then after 1 more second, "slow"

await and parallelism

In sequentialStart, execution suspends 2 seconds for the first await, and then another second for the second await. The second timer is not created until the first has already fired, so the code finishes after 3 seconds.

In concurrentStart, both timers are created and then awaited. The timers run concurrently, which means the code finishes in 2 rather than 3 seconds, i.e. the slowest timer.
However, the await calls still run in series, which means the second await will wait for the first one to finish. In this case, the result of the fastest timer is processed after the slowest.

If you wish to safely perform two or more jobs in parallel, you must await a call to Promise.all, or Promise.allSettled.

Error Handling

Note that functions concurrentStart and concurrentPromise are not functionally equivalent.

In concurrentStart, if promise fast rejects before promise slow is fulfilled, then an unhandled promise rejection error will be raised, regardless of whether the caller has configured a catch clause.

In concurrentPromise, Promise.all wires up the promise chain in one go, meaning that the operation will fail-fast regardless of the order of rejection of the promises, and the error will always occur within the configured promise chain, enabling it to be caught in the normal way.

Rewriting a Promise chain with an async function

An API that returns a Promise will result in a promise chain, and it splits the function into many parts. Consider the following code:

function getProcessedData(url) {
  return downloadData(url) // returns a promise
    .catch(e => {
      return downloadFallbackData(url)  // returns a promise
    })
    .then(v => {
      return processDataInWorker(v)  // returns a promise
    }) 
}

it can be rewritten with a single async function as follows:

async function getProcessedData(url) {
  let v
  try {
    v = await downloadData(url)
  } catch(e) {
    v = await downloadFallbackData(url)
  }
  return processDataInWorker(v)
}

In the above example, notice there is no await statement after the return keyword, although that would be valid too: The return value of an async function is implicitly wrapped in Promise.resolve - if it's not already a promise itself (as in this example).

return await promiseValue vs. return promiseValue

The implicit wrapping of return values in Promise.resolve does not imply that return await promiseValue is functionally equivalent to return promiseValue.

Consider the following rewrite of the above code. It returns null if processDataInWorker rejects with an error:

async function getProcessedData(url) {
  let v
  try {
    v = await downloadData(url)
  } catch(e) {
    v = await downloadFallbackData(url)
  }
  try {
    return await processDataInWorker(v)  // Note the `return await` vs. just `return`
  } catch (e) {
    return null
  }
}

Writing return processDataInWorker(v) would have caused the Promise returned by the function to reject, instead of resolving to null if processDataInWorker(v) rejects.

This highlights the subtle difference between return foo; and return await foo;return foo immediately returns foo and never throws, even if foo is a Promise that rejects. return await foo will wait for foo to resolve or reject if it's a Promise, and throws before returning if it rejects.

Specifications

Specification
ECMAScript (ECMA-262)
The definition of 'async function' in that specification.

Browser compatibility

Update compatibility data on GitHub
DesktopMobileServer
ChromeEdgeFirefoxInternet ExplorerOperaSafariAndroid webviewChrome for AndroidFirefox for AndroidOpera for AndroidSafari on iOSSamsung InternetNode.js
async functionChrome Full support 55Edge Full support 15Firefox Full support 52IE No support NoOpera Full support 42Safari Full support 10.1WebView Android Full support 55Chrome Android Full support 55Firefox Android Full support 52Opera Android Full support 42Safari iOS Full support 10.3Samsung Internet Android Full support 6.0nodejs Full support 7.6.0
Full support 7.6.0
Full support 7.0.0
Disabled
Disabled From version 7.0.0: this feature is behind the --harmony runtime flag.

Legend

Full support  
Full support
No support  
No support
User must explicitly enable this feature.
User must explicitly enable this feature.

See also