Promise() constructor

executor

A function to be executed by the constructor. It receives two functions as parameters: resolveFunc and rejectFunc. Any errors thrown in the executor will cause the promise to be rejected, and the return value will be neglected. The semantics of executor are detailed below.

When called via new, the Promise constructor returns a promise object. The promise object will become resolved when either of the functions resolveFunc or rejectFunc are invoked. Note that if you call resolveFunc or rejectFunc and pass another Promise object as an argument, it can be said to be "resolved", but still not "settled". See the Promise description for more explanation.

Traditionally (before promises), asynchronous tasks were designed as callbacks.

readFile("./data.txt", (error, result) => {
  // This callback will be called when the task is done, with the
  // final `error` or `result`. Any operation dependent on the
  // result must be defined within this callback.
});
// Code here is immediately executed after the `readFile` request
// is fired. It does not wait for the callback to be called, hence
// making `readFile` "asynchronous".

To take advantage of the readability improvement and language features offered by promises, the Promise() constructor allows one to transform the callback-based API to a promise-based one.

The executor is custom code that ties an outcome in a callback to a promise. You, the programmer, write the executor. Its signature is expected to be:

function executor(resolveFunc, rejectFunc) {
  // Typically, some asynchronous operation that accepts a callback,
  // like the `readFile` function above
}

resolveFunc and rejectFunc are also functions, and you can give them whatever actual names you want. Their signatures are simple: they accept a single parameter of any type.

resolveFunc(value); // call on resolved
rejectFunc(reason); // call on rejected

The value parameter passed to resolveFunc can be another promise object, in which case the newly constructed promise's state will be "locked in" to the promise passed (as part of the resolution promise). The rejectFunc has semantics close to the throw statement, so reason is typically an Error instance. If either value or reason is omitted, the promise is fulfilled/rejected with undefined.

The executor's completion state has limited effect on the promise's state:

• The executor return value is ignored. return statements within the executor merely impact control flow and alter whether a part of the function is executed, but do not have any impact on the promise's fulfillment value. If executor exits and it's impossible for resolveFunc or rejectFunc to be called in the future (for example, there are no async tasks scheduled), then the promise remains pending forever.
• If an error is thrown in the executor, the promise is rejected, unless resolveFunc or rejectFunc has already been called.

Here's a summary of the typical flow:

1. At the time when the constructor generates the new Promise object, it also generates a corresponding pair of functions for resolveFunc and rejectFunc; these are "tethered" to the Promise object.
2. executor typically wraps some asynchronous operation which provides a callback-based API. The callback (the one passed to the original callback-based API) is defined within the executor code, so it has access to the resolveFunc and rejectFunc.
3. The executor is called synchronously (as soon as the Promise is constructed) with the resolveFunc and rejectFunc functions as arguments.
4. The code within the executor has the opportunity to perform some operation. The eventual completion of the asynchronous task is communicated with the promise instance via the side effect caused by resolveFunc or rejectFunc. The side effect is that the Promise object becomes "resolved".
   • If resolveFunc is called first, the value passed will be resolved. The promise may stay pending (in case another thenable is passed), become fulfilled (in most cases where a non-thenable value is passed), or become rejected (in case of an invalid resolution value).
   • If rejectFunc is called first, the promise instantly becomes rejected.
   • Once one of the resolving functions (resolveFunc or rejectFunc) is called, the promise stays resolved. Only the first call to resolveFunc or rejectFunc affects the promise's eventual state, and subsequent calls to either function can neither change the fulfillment value/rejection reason nor toggle its eventual state from "fulfilled" to "rejected" or opposite.
   • If executor exits by throwing an error, then the promise is rejected. However, the error is ignored if one of the resolving functions has already been called (so that the promise is already resolved).
   • Resolving the promise does not necessarily cause the promise to become fulfilled or rejected (i.e. settled). The promise may still be pending because it's resolved with another thenable, but its eventual state will match that of the resolved thenable.
5. Once the promise settles, it (asynchronously) invokes any further handlers associated through then(), catch(), or finally(). The eventual fulfillment value or rejection reason is passed to the invocation of fulfillment and rejection handlers as an input parameter (see Chained Promises).

For example, the callback-based readFile API above can be transformed into a promise-based one.

const readFilePromise = (path) =>
  new Promise((resolve, reject) => {
    readFile(path, (error, result) => {
      if (error) {
        reject(error);
      } else {
        resolve(result);
      }
    });
  });

readFilePromise("./data.txt")
  .then((result) => console.log(result))
  .catch((error) => console.error("Failed to read data"));

The resolver function resolveFunc has the following behaviors:

• If it's called with the same value as the newly created promise (the promise it's "tethered to"), the promise is rejected with a TypeError.
• If it's called with a non-thenable value (a primitive, or an object whose then property is not callable, including when the property is not present), the promise is immediately fulfilled with that value.
• If it's called with a thenable value (including another Promise instance), then the thenable's then method is saved and called in the future (it's always called asynchronously). The then method will be called with two callbacks, which are two new functions with the exact same behaviors as the resolveFunc and rejectFunc passed to the executor function. If calling the then method throws, then the current promise is rejected with the thrown error.

In the last case, it means code like:

new Promise((resolve, reject) => {
  resolve(thenable);
});

Is roughly equivalent to:

new Promise((resolve, reject) => {
  try {
    thenable.then(
      (value) => resolve(value),
      (reason) => reject(reason),
    );
  } catch (e) {
    reject(e);
  }
});

Except that in the resolve(thenable) case:

1. resolve is called synchronously, so that calling resolve or reject again has no effect, even when the handlers attached through anotherPromise.then() are not called yet.
2. The then method is called asynchronously, so that the promise will never be instantly resolved if a thenable is passed.

Because resolve is called again with whatever thenable.then() passes to it as value, the resolver function is able to flatten nested thenables, where a thenable calls its onFulfilled handler with another thenable. The effect is that the fulfillment handler of a real promise will never receive a thenable as its fulfillment value.

To provide a function with promise functionality, have it return a promise by calling the resolve and reject functions at the correct times.

function myAsyncFunction(url) {
  return new Promise((resolve, reject) => {
    const xhr = new XMLHttpRequest();
    xhr.open("GET", url);
    xhr.onload = () => resolve(xhr.responseText);
    xhr.onerror = () => reject(xhr.statusText);
    xhr.send();
  });
}

Calling resolveFunc causes the promise to become resolved, so that calling resolveFunc or rejectFunc again has no effect. However, the promise may be in any of the states: pending, fulfilled, or rejected.

This pendingResolved promise is resolved the time it's created, because it has already been "locked in" to match the eventual state of the inner promise, and calling resolveOuter or rejectOuter or throwing an error later in the executor has no effect on its eventual state. However, the inner promise is still pending until 100ms later, so the outer promise is also pending:

const pendingResolved = new Promise((resolveOuter, rejectOuter) => {
  resolveOuter(
    new Promise((resolveInner) => {
      setTimeout(() => {
        resolveInner("inner");
      }, 100);
    }),
  );
});

This fulfilledResolved promise becomes fulfilled the moment it's resolved, because it's resolved with a non-thenable value. However, when it's created, it's unresolved, because neither resolve nor reject has been called yet. An unresolved promise is necessarily pending:

const fulfilledResolved = new Promise((resolve, reject) => {
  setTimeout(() => {
    resolve("outer");
  }, 100);
});

Calling rejectFunc obviously causes the promise to reject. However, there are also two ways to cause the promise to instantly become rejected even when the resolveFunc callback is called.

// 1. Resolving with the promise itself
const rejectedResolved1 = new Promise((resolve) => {
  // Note: resolve has to be called asynchronously,
  // so that the rejectedResolved1 variable is initialized
  setTimeout(() => resolve(rejectedResolved1)); // TypeError: Chaining cycle detected for promise #<Promise>
});

// 2. Resolving with an object which throws when accessing the `then` property
const rejectedResolved2 = new Promise((resolve) => {
  resolve({
    get then() {
      throw new Error("Can't get then property");
    },
  });
});

• Polyfill of Promise in core-js
• Using Promises