Short History
In concurrent programming languages, at least three terms have been used for the same promising synchronization feature: future, promise, and delay. These terms refer to an object that acts as a proxy for a result that will be available only some time later because its computation takes place asynchronously. In this tutorial, we show how to use ES6 Promises and how they can help with asynchronous computations.
The term promise was coined by Daniel P. Friedman and David Wise in 1976, while Peter Hibbard called it eventual. The first future and/or promise constructs were implemented in programming languages such as MultiLisp and Act 1. Later, promises became popular and some high level programming languages, such as for example C++ and Java, started to provide built-in support for them. JavaScript promises started out in the DOM as Futures, before they were renamed to Promises, and finally moved into JavaScript, as part of the ECMAScript 6 (ES6) specification. Having them in JavaScript rather than in the DOM is great because this makes them available in other JavaScript contexts, like in NodeJS.
See the JavaScript Summary on web-engineering.info for an overview about the current state of JavaScript.
What is a Promise?
A Promise represents an operation that has not completed yet, but is expected in the future. It is a proxy for a value not necessarily known when the promise is created. It allows you to associate handlers to an asynchronous action's eventual success value or failure reason. This lets asynchronous methods return values like synchronous methods: instead of the final value, the asynchronous method returns a promise of having a value at some point in the future.
A Promise is created by using the Promise( function (resolve, reject) {...}) constructor. It takes one argument, a function with two parameters, resolve and reject. Notice that resolve and reject are also functions. The generic structure of a promise is shown in the example below:
var promise = new Promise( function (resolve, reject) {
if () {
resolve( params);
return resolve( params);
} else {
reject( err);
return reject( err);
}
});
Then, we can use the promise by invoking its then method (from here the thenable terminology):
promise.then( function (params) {
console.log(params);
}).catch( function (err) {
console.log( err);
});
The then method takes as parameter a method, invoked when the promise fulfilled, i.e., the resolve method was invoked. The catch method takes as parameter a method, invoked when the promise is rejected, i.e., the reject method was invoked. For both cases, the callback parameters are the ones used to call the resolve, respectively the reject methods. Alternatively, is possible to use only the then method for both, fulfilled and rejected cases:
promise.then( function (params) {
console.log( params);
},function (err) {
console.log( err);
});
While catch is almost the equivalent of then( undefined, catchCallback), using both, then and catch, makes the code easier to read and understand.
Promises can also be chained, so we can write code similar to:
promise.then( function (params) {
console.log(params);
}).catch( function (err) {
console.log(err);
}).then(function (params) {
});
Notice that we can chain then with then (i.e., .then(function (p){...}).then(function (p){...})) or catch with then (i.e., .then(function (p){...}).catch(function (p){...}).then(function (p){...})) as many times as needed.
States and Fates
Promises may be in one of the following three states:
fulfilled: if promise.then(f) will call f at some point in timerejected: if promise.catch(e) will call e at some point in timepending: if it is neither fulfilled nor rejected
Another term, i.e., settled, is used for something that's not really a promise state, but it expresses the fact that a promise has either fulfilled or rejected.
Note: A promise can be, at any point in time, only in one of the three above described states.
Promises have two possible mutually exclusive fates:
resolved: The promise either follows another promise, or has been fulfilled or rejectedunresolved: If trying to resolve or reject, it will have an impact on the promise
A promise can be resolved to either a promise or thenable, in which case it will store the promise or thenable for later unwrapping, or it can be resolved to a non-promise value, in which case it is fulfilled with that value.
Example 1: Asynchronous JavaScript File Loading
Before ECMAScript6, the JavaScript programmers used to implement promises in ad-hock manners, with the help of callback methods and/or using setTimeout/setInterval calls combined with flags and conditional checks.
One common example of asynchronous operation was dynamic JavaScript scripts loading. This is very useful to obtain a modular behavior for our JavaScript applications, and also to avoid the load of unnecessarily JavaScript code if not needed (i.e., save some bandwidth in case of mobile devices).
No-Promise JavaScript Implementation
To dynamically load a JavaScript file, we had a code similar to the following one:
function loadJsFile( filePath, onLoadFinishedCallback, onLoadErrorCallback) {
var scriptElem = document.createElement('script');
scriptElem.onload = onLoadFinishedCallback;
scriptElem.onerror = onLoadErrorCallback;
scriptElem.src = filePath;
document.head.appendChild( scriptElem );
};
Then, when needed, we may call the loadJsFile method:
loadJsFile("js/script.js",
function () {
},
function (err) {
alert("Failed to load JS file!);
// write to JS console the details about the error
console.log(err);
}
);
Note: The example code works in most of the modern web browsers. Some old web browsers may not be able to execute it. Check the JavaScript console for details in case of errors. With most of the browsers, it does NOT work locally, by using the file protocol (e.g., file://path/to/app/index.html).
ES6 Promise Implementation
Let's see now how we obtain the same behavior, this time with the help of ECMAScript6 Promises. First, we re-implement the loadJsFile method as follows:
function loadJsFile( filePath) {
return new Promise( function (resolve, reject) {
var scriptElem = document.createElement('script');
scriptElem.onload = resolve;
scriptElem.onerror = reject;
scriptElem.src = filePath;
document.head.appendChild(scriptElem );
});
};
When the file loaded successfully, the promise resolve is called. The same, if the file loading fails, e.g., the file does not exists, the promise reject is called. The difference is that our loadJsFile has now only one parameter, i.e., the filePath instead of three, which is what we normally expect from a "file loading" method.
Later, somewhere in our JavaScript code, we like to invoke the loadJsFile method:
loadJsFile("js/myScript.js").then(
function () {
})
.catch(function (err) {
alert("Failed to load JS file!");
console.log(err);
}
);
Instead of using success and error callback methods as parameters of the loadJsFile method, we use then and catch. If the promise is fulfilled, the method given as parameter to then method is executed. If the promise is rejected, the method given as parameter to catch method is invoked, expressing the failure/reject state.
Note: The example code works in most of the modern web browsers. Some old web browsers may not be able to execute it. Check the JavaScript console for details in case of errors. With most of the browsers, it does NOT work locally, by using the file protocol (e.g., file://path/to/app/index.html).
Example 2: XHR (XML HTTP Request)
XHR/AJAX requests are another important use case of JavaScript asynchronous code. Normally, we use such code to retrieve/send data from/to a server, as for example in the case of dynamic auto-completion form fields, or when we need to retrieve data which is stored into a server database. More complex applications, such as web file managers, use XHR/AJAX to inform a back-end web application that a file (or folder) needs to be created/deleted, or that we need the index of folder.
No-Promise JavaScript Implementation
We start by creating a minimalistic NodeJS based HTTP server which is able to reply to GET requests with a simple message in the format {timestamp: timestamp_value, message: "some message..."}. Fortunately, we don't have to install any additional module, but you'll have download and install NodeJS, in case you don't have it already.
The NodeJS server code is very simple:
var port = 3000,
http = require('http'),
httpServer = null;
httpServer = http.createServer(function (req, resp){
resp.end(JSON.stringify(
{timestamp: (new Date()).getTime(),
message: "Hi, I am the NodeJS HTTP server!"}));
});
httpServer.listen(port, function (){
console.log("HTTP Server started on port " + port);
});
Create a new file named server.js somewhere on your disk (or to a folder of an online server if you have one), and copy/paste the above example code. To start the HTTP server, use a command line tool, navigate to the folder where you have created the server.js file, and execute node server.js. To try if your HTTP server runs, open your favorite web browser and navigate to http://domain:3000. Remember to replace the domain with localhost, if you run the server on your local machine, or with the real web domain name, if you use an online server. Notice that 3000 is the port number we have used in our example code, so you may need to change it with something else, depending on your server configuration (e.g., the port is not open in the server firewall, or some other application uses port 3000).
It is now the time to write a method that uses XHR (XML HTTP Request), to make a GET request to our HTTP server:
function makeGetRequest(url, successCallback, errCallback) {
var req = new XMLHttpRequest();
req.open('GET', url);
req.onload = function () {
if (req.status == 200) successCallback(JSON.parse(req.response));
else errCallback({code: req.status, message: req.statusText});
};
req.send();
};
For simplicity reasons, we create a lightweight version of a GET request implementation, which does not allow request parameters and does not treat more specific error cases, such as the ones specific to network problems, normally being captured with the help of onerror (i.e., req.onerror = function (err) {...})).
Calling the makeGetRequest method somewhere in our JavaScript code can be done as in the following example:
makeGetRequest("/",
function (resp) {
},
function (err) {
alert("GET request failed!")
console.log(err);
}
);
Remember to change the request URL with whatever you need to, specially if the NodeJS HTTP server runs code behind (or together with, e.g., using a proxy) Apache or Nginx. If you are behind a corporate firewall, it is very likely that any other port excepting 80 and maybe 443 are closed. In this case, you may have to use Apache to create a proxy for your HTTP server application. This can be done by editing your httpd.conf file, and adding the following code as part of the virtual host of the used domain:
ProxyPass /example/promise http:
ProxyPassReverse /example/promise http:
<proxy>
Order Allow,Deny
Allow from all
</proxy>
Remember to restart the httpd service (e.g., by using service httpd restart) after saving the above modifications. Also, remember to call the makeGetRequest with the new URL as parameter, i.e., makeGetRequest("example/promise").then(...).
Note: The example client code works in most of the modern web browsers. Some old web browsers may not be able to execute it. Check the JavaScript console for details in case of errors. It does NOT execute locally, i.e., by using the file protocol!
ES6 Promise Implementation
Our NodeJS code remains unchanged, but the client makeGetRequest method needs to be re-implemented by using a Promise this time:
function makeGetRequest(url) {
return new Promise(function (resolve, reject) {
var req = new XMLHttpRequest();
req.open('GET', url);
req.onload = function () {
if (req.status == 200) resolve(JSON.parse(req.response));
else reject({code: req.status, message: req.statusText});
};
req.send();
});
};
Mainly, we check if the response code is 200 (HTTP code 200 means all went fine, "OK"). In case we get that code, we parse the response by using JSON.parse, since we know that we expect a string serialization of a JSON object from our server. Any HTTP code other than 200 in our example represents an error, so the promise reject is called.
In our HTML code, at some point, we need to invoke the makeGetRequest method:
makeGetRequest("/example/promise")
.then(function (resp) {
})
.catch(function (err) {
alert("AJAX GET request failed!")
console.log(err);
});
Note: The example code works in most of the modern web browsers. Some old web browsers may not be able to execute it. Check the JavaScript console for details in case of errors. It does NOT execute locally, i.e., by using the file protocol!
Example 3: Login Action with NodeJS and MongoDB Database
Latest NodeJS releases provide native support for promises. We've tested the example code fragments shown below with NodeJS v5.4.1. This is awesome, because we can use them with a JavaScript engine which does not run only in a web browser. A relevant use case for asynchronous code in a NodeJS server application is provided by database operations, such as connect and query.
In this section, we show how to use promises in combination with a MongoDB database. We'll try to connect to a MongoDB database, then check the credentials (username and password) for a login request. Such scenarios are often seen in applications that provide a user authentication mechanism. For this scenario, we'll use the mongodb NodeJS module, which can be installed by executing the npm install mongodb command, in the root folder of the application. Remember that you need to install MongoDB Database Engine before being able to use MongoDB with any application.
No-Promise JavaScript Implementation
Our database contains a collection named users which stores the registered users for our app. Due to JavaScript’s non-blocking nature, however, we can't just use find({username: val1, password: val2}). First, we should be sure that we have connected successfully to the database, so we need to use mongodb.connect and pass it a callback function where, depending on the success or error case, we look for the user or we have to deal with a connection error.
var MongoClient = require('mongodb').MongoClient;
function login(username, password, successCallback, errCallback) {
MongoClient.connect("mongodb://localhost:27017/myapp", function (err, db) {
var usersCollection = null, stream = null;
if (err) errCallback(err);
else {
usersCollection = db.collection('users');
stream = usersCollection.findOne(
{username: username, password: password},
function (err, user) {
if (err) errCallback(err);
else if (!user) errCallback();
else successCallback(user);
}
);
}
});
};
Our method takes 4 parameters: the username, associated password, a success callback and an error callback. While this code does not look bad, things can be worse in complex scenarios. Imagine for example that we have to connect to database, then check authentication, then make a modification (update) and then update a log collection, and all these operations have callbacks. In such a case, a "doom pyramid" code is obtained, which is really ugly and hard to read.
Later on, the login method can be invoked to check the credentials of a specific user:
login("johnny", "12CqqDt4e*",
function (user) {
console.log("Login successful");
},
function (err) {
console.log("Login failed.");
}
);
Create a file named mongotest.js and copy/paste the above two pieces of code to it. Using a command line tool (e.g., cmd for Windows or terminal for Linux), navigate to the folder where this file is located and install the mongodb NodeJS module by executing npm install mongodb.
To run the example, execute node mongotest.js. Before being able to have a succesful login, you need to create the Mongo collection and insert the user(s). For this, using your command tool, type mongo and press enter. In the interactive Mongo client mode, type the following code:
use myapp;
db.createCollection("users");
db.users.insert({username:"johnny", password:"12CqqDt4e"});
ES6 Promise Implementation
Our login method is now updated to use ECMAScript6 Promises:
var MongoClient = require('mongodb').MongoClient;
function login(username, password) {
return new Promise(function (resolve, reject) {
MongoClient.connect("mongodb://localhost:27017/myapp", function (err, db) {
var usersCollection = null, stream = null;
if (err) reject(err);
else {
usersCollection = db.collection('users');
stream = usersCollection.findOne(
{username: username, password: password},
function (err, user) {
if (err) reject(err);
else if (!user) reject();
else resolve(user);
}
);
}
});
});
};
Practically, we replaced all the successCallback and errCallback calls with resolve and respectively reject. The two callback methods used as parameters before are no longer needed. Using our login method is straight forward:
login("johnny", "12345")
.then(function (user) {
console.log("Login successful");
})
.catch(function (err) {
console.log("Login failed.");
});
Interesting Facts About Promises
Promises not only make your code look better and more readable, but they also provide additional advantages, such as a semantic handling of errors by using catch instead of our old errorCallback buddy. There are a few more important facts to consider when using promises, which are discussed in this section.
Do Promises Swallow Your Errors?
Note: This fact is inspired by a blog post of James K. Nelson. You may also want to check this youtube video related to this matter.
Considering the following promises:
Promise.resolve('promised value').then(function () {
throw new Error('error');
});
Promise.reject('error value').catch(function () {
throw new Error('error');
});
new Promise(function (resolve, reject) {
throw new Error('error');
});
You may think: "what? clearly, all three of them throw the error!". Well, that's not really the case, and actually you won't get the error in any of the cases. Solving this issue is possible by using catch after each of these promises, such as:
Promise.resolve('promised value').then(function () {
throw new Error('error');
}).catch(function (error) {
console.error(error.stack);
});
However, if this does not look like your style to work with promises, then try the bluebirdjs library, which allows you to solve the problem as shown below:
Promise.onPossiblyUnhandledRejection(function (error){
throw error;
});
Promise.reject('error value').catch(function () {});
Promises Can Be Looped
Supposing you have more than one promise which should be handled in the same way, then you can use Promise.all to loop them and execute your code:
var p1 = Promise.resolve("I am a solved promise!"),
p2 = 2.7889,
p3 = new Promise( function (resolve, reject) {
setTimeout(resolve, 1000, "My name is 'The Promise'!");
});
Promise.all( [p1, p2, p3]).then( function (values) {
console.log(values);
});
The result of executing the code is the array: ["I am a solved promise!", 2.7889, "My name is 'The Promise'!"].
Equivalent Promises
For specific cases, we can use Promise.resolve, or we can create a new Promise, or we can use the then method, or ... is all getting confusing. In fact, there are many ways to obtain the same promise behavior. For example, all the following promises are functionally equivalent:
Promise.resolve(v);
new Promise(function (resolve, reject) {resolve(v);});
Promise.resolve().then( function () {return v; });
Promise.all([v]).then( function (val){ return val[0]; });
The choices catch() and then(undefined,function(v){}) are not "exactly" the same
In the first part of the article, we discussed that using then and catch is "almost" equivalent with using then with two parameters. Let's see why "almost", and for this, we consider the following "equivalent" promises:
somePromise()
.then(function () {
.catch(function (err) {
somePromise().then(function () {
}, function (err) {
});
At first, they look to give the same results, and with an exception, they actually do so. The difference is made in the case when an exception/error is thrown in the then handled function:
somePromise()
.then(function () {
throw new Error('Bad, things are not good here!');
}).catch(function (err) {
});
somePromise().then(function () {
throw new Error('Bad, things are not good here!');
}, function (err) {
});
When using then, to handle both, the resolve and the reject cases, an error (exception) thrown in the resolve case is not passed to the reject handler.
Support for ES6 Promises
Web Browsers: The current versions of all major web browsers (Mozilla Firefox, Google Chrome, Microsoft Edge, Opera and Safari) support the new promise method. For a complete list of supportive browsers, check caniuse.com.
NodeJS: provides builtin support for promises, and in addition, a set of modules, such as bluebirdjs, make the use of promises even easier and appealing.
