This Pusher Channels client library supports web browsers, web workers, Node.js and React Native.
If you're looking for the Pusher Channels server library for Node.js, use pusher-http-node instead.
For tutorials and more in-depth information about Pusher Channels, visit our official docs.
The following topics are covered:
- Installation
- Initialization
- Configuration
- Global Configuration
- Connection
- Subscribing to Channels (Public and Private)
- Accessing Channels
- Binding to Events
- Default events
- Developing
-
Web
-
We test against Chrome, Firefox and Safari.
-
Works with all major web frameworks, including
- Angular (See Angular tutorial)
- React (See React tutorial)
- Vue.js (see Vue.js tutorial)
-
If you're using Pusher Channels on a web page, you can install the library via:
You can use any NPM-compatible package manager, including NPM itself and Yarn.
yarn add pusher-js
Then:
import Pusher from 'pusher-js';
Or, if you're not using ES6 modules:
const Pusher = require('pusher-js');
<script src="https://js.pusher.com/5.0/pusher.min.js"></script>
You can also use cdnjs.com if you prefer or as a fallback.
Or via Bower:
bower install pusher
and then:
<script src="bower_components/pusher/dist/web/pusher.min.js"></script>
Warning it's now necessary to install @react-native-community/netinfo in order to use pusher-js with react-native. pusher-js depends on NetInfo. NetInfo. NetInfo was included within react-native core until v0.60, when it was moved to the @react-native-community/netinfo library. Please follow the install instructions for the @react-native-community/netinfo library before trying to use pusher-js in your react-native project.
Use a package manager like Yarn or NPM to install pusher-js
and then import
it as follows:
import Pusher from 'pusher-js/react-native';
Notes:
- The fallbacks available for this runtime are HTTP streaming and polling.
- This build uses React Native's NetInfo API to detect changes on connectivity state. It will use this to automatically reconnect.
(pusher-js
's Web Workers implementation is currently not compatible with Internet Explorer)
You can import the worker script (pusher.worker.js
, not pusher.js
) from the CDN:
importScripts('https://js.pusher.com/5.0/pusher.worker.min.js');
Having installed pusher-js
via an NPM-compatible package manager, run:
import Pusher from 'pusher-js';
Notes:
- For standard
WebWorkers
, this build will use HTTP as a fallback. - For
ServiceWorkers
, as theXMLHttpRequest
API is unavailable, there is currently no support for HTTP fallbacks. However, we are open to requests for fallbacks usingfetch
if there is demand.
const socket = new Pusher(APP_KEY, {
cluster: APP_CLUSTER,
});
You can get your APP_KEY
and APP_CLUSTER
from the Pusher Channels dashboard.
There are a number of configuration parameters which can be set for the client, which can be passed as an object to the Pusher constructor, i.e.:
const socket = new Pusher(APP_KEY, {
cluster: APP_CLUSTER,
authEndpoint: 'http://example.com/pusher/auth',
forceTLS: true
});
For most users, there is little need to change these. See client API guide for more details.
Forces the connection to use TLS.
Endpoint on your server that will return the authentication signature needed for private and presence channels. Defaults to '/pusher/auth'
.
For more information see authenticating users.
Defines how the authentication endpoint, defined using authEndpoint, will be called. There are two options available: ajax
and jsonp
.
ajax
- The default option where anXMLHttpRequest
object will be used to make a request. The parameters will be passed asPOST
parameters.jsonp
- The authentication endpoint will be called by a<script>
tag being dynamically created pointing to the endpoint defined byauthEndpoint
. This can be used when the authentication endpoint is on a different domain to the web application. The endpoint will therefore be requested as aGET
and parameters passed in the query string.
For more information see the Channel authentication transport section of our authenticating users docs.
Allows passing additional data to authorizers. Supports query string params and headers (AJAX only). For example, following will pass foo=bar
via the query string and baz: boo
via headers:
const socket = new Pusher(APP_KEY, {
cluster: APP_CLUSTER,
auth: {
params: { foo: 'bar' },
headers: { baz: 'boo' }
}
});
Additional parameters to be sent when the channel authentication endpoint is called. When using ajax authentication the parameters are passed as additional POST
parameters. When using jsonp authentication the parameters are passed as GET
parameters. This can be useful with web application frameworks that guard against CSRF (Cross-site request forgery).
If you require a CSRF header for incoming requests to the private channel authentication endpoint on your server, you should add a CSRF token to the auth
hash under headers
. This is applicable to frameworks which apply CSRF protection by default.
const socket = new Pusher(APP_KEY, {
cluster: APP_CLUSTER,
auth: {
params: { foo: 'bar' },
headers: { 'X-CSRF-Token': 'SOME_CSRF_TOKEN' }
}
});
If you need custom authorization behavior you can provide your own authorizer
function as follows:
const socket = new Pusher(APP_KEY, {
cluster: APP_CLUSTER,
authorizer: function (channel, options) {
return {
authorize: function (socketId, callback) {
// Do some ajax to get the auth information
callback(false, authInformation);
}
};
}
})
Specifies the cluster that pusher-js should connect to. If you'd like to see a full list of our clusters, click here. If you do not specify a cluster, mt1
will be used by default.
const socket = new Pusher(APP_KEY, {
cluster: APP_CLUSTER,
});
Disables stats collection, so that connection metrics are not submitted to Pusher’s servers. These stats are used for internal monitoring only and they do not affect the account stats.
Specifies which transports should be used by pusher-js to establish a connection. Useful for applications running in controlled, well-behaving environments. Available transports for web: ws
, wss
, xhr_streaming
, xhr_polling
, sockjs
. If you specify your transports in this way, you may miss out on new transports we add in the future.
// Only use WebSockets
const socket = new Pusher(APP_KEY, {
cluster: APP_CLUSTER,
enabledTransports: ['ws']
});
Specifies which transports must not be used by pusher-js to establish a connection. This settings overwrites transports whitelisted via the enabledTransports
options. Available transports for web: ws
, wss
, xhr_streaming
, xhr_polling
, sockjs
. This is a whitelist, so any new transports we introduce in the future will be used until you explicitly add them to this list.
// Use all transports except for sockjs
const socket = new Pusher(APP_KEY, {
cluster: APP_CLUSTER,
disabledTransports: ['sockjs']
});
// Only use WebSockets
const socket = new Pusher(APP_KEY, {
cluster: APP_CLUSTER,
enabledTransports: ['ws', 'xhr_streaming'],
disabledTransports: ['xhr_streaming']
});
These can be changed to point to alternative Pusher Channels URLs (used internally for our staging server).
Useful in special scenarios if you're using the library against an endpoint you control yourself. This is used internally for testing.
Ignores null origin checks for HTTP fallbacks. Use with care, it should be disabled only if necessary (i.e. PhoneGap).
If there is no activity for this length of time (in milliseconds), the client will ping the server to check if the connection is still working. The default value is set by the server. Setting this value to be too low will result in unnecessary traffic.
Time before the connection is terminated after a ping is sent to the server. Default is 30000 (30s). Low values will cause false disconnections, if latency is high.
Enables logging to the browser console via calls to console.log
.
Assign a custom log handler for the pusher-js library logging. For example:
Pusher.log = (msg) => {
console.log(msg);
};
By setting the log
property you also override the use of Pusher.enableLogging
.
A connection to Pusher Channels is established by providing your APP_KEY
and APP_CLUSTER
to the constructor function:
const socket = new Pusher(APP_KEY, {
cluster: APP_CLUSTER,
});
This returns a socket object which can then be used to subscribe to channels.
One reason this connection might fail is your account being over its' limits. You can detect this in the client by binding to the error
event on the pusher.connection
object. For example:
var pusher = new Pusher('app_key');
pusher.connection.bind( 'error', function( err ) {
if( err.error.data.code === 4004 ) {
log('Over limit!');
}
});
You may disconnect again by invoking the disconnect
method:
socket.disconnect();
The connection can be in any one of these states.
State | Note |
---|---|
initialized | Initial state. No event is emitted in this state. |
connecting | All dependencies have been loaded and Channels is trying to connect. The connection will also enter this state when it is trying to reconnect after a connection failure. |
connected | The connection to Channels is open and authenticated with your app. |
unavailable | The connection is temporarily unavailable. In most cases this means that there is no internet connection. It could also mean that Channels is down |
failed | Channels is not supported by the browser. This implies that WebSockets are not natively available and an HTTP-based transport could not be found. |
disconnected | The Channels connection was previously connected and has now intentionally been closed. |
Making a connection provides the client with a new socket_id
that is assigned by the server. This can be used to distinguish the client's own events. A change of state might otherwise be duplicated in the client. More information on this pattern is available here.
It is also stored within the socket, and used as a token for generating signatures for private channels.
The default method for subscribing to a channel involves invoking the subscribe
method of your socket object:
const channel = socket.subscribe('my-channel');
This returns a Channel object which events can be bound to.
Private channels are created in exactly the same way as normal channels, except that they reside in the 'private-' namespace. This means prefixing the channel name:
const channel = socket.subscribe('private-my-channel');
Like private channels, encrypted channels have their own namespace, 'private-encrypted-'. For more information about encrypted channels, please see the docs.
Please note that encrypted channels are only officially supported for our 'web' and 'node' clients for now. We know for sure this won't work in React Native builds since the React Native runtime does not include the required crypto functionality we depend on. Please let us know if you need this functionality in our web-worker or React Native builds!
const channel = socket.subscribe('private-encrypted-my-channel');
It is possible to access channels by name, through the channel
function:
const channel = socket.channel('private-my-channel');
It is possible to access all subscribed channels through the allChannels
function:
socket.allChannels().forEach(channel => console.log(channel.name));
Private, presence and encrypted channels will make a request to your authEndpoint
(/pusher/auth
) by default, where you will have to authenticate the subscription. You will have to send back the correct auth response and a 200 status code.
To unsubscribe from a channel, invoke the unsubscribe
method of your socket object:
socket.unsubscribe('my-channel');
Unsubscribing from private channels is done in exactly the same way, just with the additional private-
prefix:
socket.unsubscribe('private-my-channel');
Event binding takes a very similar form to the way events are handled in jQuery. You can use the following methods either on a channel object, to bind to events on a particular channel; or on the pusher object, to bind to events on all subscribed channels simultaneously.
Binding to "new-message" on channel: The following logs message data to the console when "new-message" is received
channel.bind('new-message', function (data) {
console.log(data.message);
});
We can also provide the this
value when calling a handler as a third optional parameter. The following logs "hi Pusher" when "my-event" is fired.
channel.bind('my-event', function () {
console.log(`hi ${this.name}`);
}, { name: 'Pusher' });
For client-events on presence channels, bound callbacks will be called with an additional argument. This argument is an object containing the user_id
of the user who triggered the event
presenceChannel.bind('client-message', function (data, metadata) {
console.log('received data from', metadata.user_id, ':', data);
});
Unsubscribe behaviour varies depending on which parameters you provide it with. For example:
// Remove just `handler` for the `new-comment` event
channel.unbind('new-comment', handler);
// Remove all handlers for the `new-comment` event
channel.unbind('new-comment');
// Remove `handler` for all events
channel.unbind(null, handler);
// Remove all handlers for `context`
channel.unbind(null, null, context);
// Remove all handlers on `channel`
channel.unbind();
bind_global
and unbind_global
work much like bind
and unbind
, but instead of only firing callbacks on a specific event, they fire callbacks on any event, and provide that event along to the handler along with the event data. For example:
channel.bind_global(function (event, data) {
console.log(`The event ${event} was triggered with data ${data}`);
})
unbind_global
works similarly to unbind
.
// remove just `handler` from global bindings
channel.unbind_global(handler);
// remove all global bindings
channel.unbind_global();
The unbind_all
method is equivalent to calling unbind()
and unbind_global()
together; it removes all bindings, global and event specific.
Currently, pusher-js itself does not support authenticating multiple channels in one HTTP request. However, thanks to @dirkbonhomme you can use the pusher-js-auth plugin that buffers subscription requests and sends auth requests to your endpoint in batches.
There are a number of events which are used internally, but can also be of use elsewhere, for instance subscribe
. There is also a state_change
event - which fires whenever there is a state change. You can use it like this:
pusher.connection.bind('state_change', function(states) {
// states = {previous: 'oldState', current: 'newState'}
$('div#status').text("Channels current state is " + states.current);
});
To listen for when you connect to Pusher Channels:
socket.connection.bind('connected', callback);
And to bind to disconnections:
socket.connection.bind('disconnected', callback);
You can host JavaScript files yourself, but it's a bit more complicated than putting them somewhere and just linking pusher.js
in the source of your website. Because pusher-js loads fallback files dynamically, the dependency loader must be configured correctly or it will be using js.pusher.com
.
First, clone this repository and run npm install && git submodule init && git submodule update
. Then run:
$ CDN_HTTP='http://your.http.url' CDN_HTTPS='https://your.https.url' make web
In the dist/web
folder, you should see the files you need: pusher.js
, pusher.min.js
, json2.js
, json.min.js
, sockjs.js
and sockjs.min.js
. pusher.js
should be built referencing your URLs as the dependency hosts.
First, make sure you expose all files from the dist
directory. They need to be in a directory with named after the version number. For example, if you're hosting version 5.0.0 under http://example.com/pusher-js
(and https for SSL), files should be accessible under following URL's:
http://example.com/pusher-js/5.0.0/pusher.js
http://example.com/pusher-js/5.0.0/json2.js
http://example.com/pusher-js/5.0.0/sockjs.js
Minified files should have .min
in their names, as in the dist/web
directory:
http://example.com/pusher-js/5.0.0/pusher.min.js
http://example.com/pusher-js/5.0.0/json2.min.js
http://example.com/pusher-js/5.0.0/sockjs.min.js
Most browsers have a limit of 6 simultaneous connections to a single domain, but Internet Explorer 6 and 7 have a limit of just 2. This means that you can only use a single Pusher Channels connection in these browsers, because SockJS requires an HTTP connection for incoming data and another one for sending. Opening the second connection will break the first one as the client won't be able to respond to ping messages and get disconnected eventually.
All other browsers work fine with two or three connections.
Install all dependencies via Yarn:
yarn install
Run a development server which serves bundled javascript from http://localhost:5555/pusher.js so that you can edit files in /src freely.
make serve
You can optionally pass a PORT
environment variable to run the server on a different port. You can also pass CDN_HTTP
and CDN_HTTPS
variables if you wish the library to load dependencies from a new host.
This command will serve pusher.js
, sockjs.js
, json2.js
, and their respective minified versions.
New to pusher-js 3.1 is the ability for the library to produce builds for different runtimes: classic web, React Native, NodeJS and Web Workers.
In order for this to happen, we have split the library into two directories: core/
and runtimes/
. In core
we keep anything that is platform-independent. In runtimes
we keep code that depends on certain runtimes.
Throughout the core/
directory you'll find this line:
import Runtime from "runtime";
We use webpack module resolution to make the library look for different versions of this module depending on the build.
For web it will look for src/runtimes/web/runtime.ts
. For ReactNative, src/runtimes/react-native/runtime.ts
. For Node: src/runtimes/node/runtime.ts
. For worker: src/runtimes/worker/runtime.ts
.
Each of these runtime files exports an object (conforming to the interface you can see in src/runtimes/interface.ts
) that abstracts away everything platform-specific. The core library pulls this object in without any knowledge of how it implements it. This means web build can use the DOM underneath, the ReactNative build can use its native NetInfo API, Workers can use fetch
and so on.
In order to build SockJS, you must first initialize and update the Git submodule:
git submodule init
git submodule update
Then run:
make web
This will build the source files relevant for the web build into dist/web
.
In order to specify the library version, you can either update package.json
or pass a VERSION
environment variable upon building.
Other build commands include:
make react-native # for the React Native build
make node # for the NodeJS build
make worker # for the worker build
Each test environment contains two types of tests:
- unit tests,
- integration tests.
Unit tests are simple, fast and don't need any external dependencies. Integration tests usually connect to production and js-integration-api servers and can use a local server for loading JS files, so they need an Internet connection to work.
There are 3 different testing environments: one for web, one for NodeJS and one for workers. We may consider adding another one for React Native in the future.
The web and worker tests use Karma to execute specs in real browsers. The NodeJS tests use jasmine-node.
To run the tests:
# For web
make web_unit
make web_integration
# For NodeJS
make node_unit
make node_integration
# For workers
make worker_unit
make worker_integration
If you want your Karma tests to automatically reload, then in spec/karma/config.common.js
set singleRun
to false
.