@fastify/websocket
WebSocket support for Fastify. Built upon ws@8.
Install
npm i @fastify/websocket
# or
yarn add @fastify/websocketIf you're a TypeScript user, this package has its own TypeScript types built in, but you will also need to install the types for the ws package:
npm i @types/ws -D
# or
yarn add -D @types/wsUsage
After registering this plugin, you can choose on which routes the WS server will respond. This can be achieved by adding websocket: true property to routeOptions on a fastify's .get route. In this case two arguments will be passed to the handler, the socket connection, and the fastify request object:
'use strict'
const fastify = require('fastify')()
fastify.register(require('@fastify/websocket'))
fastify.register(async function (fastify) {
fastify.get('/', { websocket: true }, (connection /* SocketStream */, req /* FastifyRequest */) => {
connection.socket.on('message', message => {
// message.toString() === 'hi from client'
connection.socket.send('hi from server')
})
})
})
fastify.listen({ port: 3000 }, err => {
if (err) {
fastify.log.error(err)
process.exit(1)
}
})In this case, it will respond with a 404 error on every unregistered route, closing the incoming upgrade connection requests.
However, you can still define a wildcard route, that will be used as default handler:
'use strict'
const fastify = require('fastify')()
fastify.register(require('@fastify/websocket'), {
options: { maxPayload: 1048576 }
})
fastify.register(async function (fastify) {
fastify.get('/*', { websocket: true }, (connection /* SocketStream */, req /* FastifyRequest */) => {
connection.socket.on('message', message => {
// message.toString() === 'hi from client'
connection.socket.send('hi from wildcard route')
})
})
fastify.get('/', { websocket: true }, (connection /* SocketStream */, req /* FastifyRequest */) => {
connection.socket.on('message', message => {
// message.toString() === 'hi from client'
connection.socket.send('hi from server')
})
})
})
fastify.listen({ port: 3000 }, err => {
if (err) {
fastify.log.error(err)
process.exit(1)
}
})Attaching event handlers
It is important that websocket route handlers attach event handlers synchronously during handler execution to avoid accidentally dropping messages. If you want to do any async work in your websocket handler, say to authenticate a user or load data from a datastore, ensure you attach any on('message') handlers before you trigger this async work. Otherwise, messages might arrive whilst this async work is underway, and if there is no handler listening for this data it will be silently dropped.
Here is an example of how to attach message handlers synchronously while still accessing asynchronous resources. We store a promise for the async thing in a local variable, attach the message handler synchronously, and then make the message handler itself asynchronous to grab the async data and do some processing:
fastify.get('/*', { websocket: true }, (connection, request) => {
const sessionPromise = request.getSession() // example async session getter, called synchronously to return a promise
connection.socket.on('message', async (message) => {
const session = await sessionPromise()
// do something with the message and session
})
})Using hooks
Routes registered with @fastify/websocket respect the Fastify plugin encapsulation contexts, and so will run any hooks that have been registered. This means the same route hooks you might use for authentication or error handling of plain old HTTP handlers will apply to websocket handlers as well.
fastify.addHook('preValidation', async (request, reply) => {
// check if the request is authenticated
if (!request.isAuthenticated()) {
await reply.code(401).send("not authenticated");
}
})
fastify.get('/', { websocket: true }, (connection, req) => {
// the connection will only be opened for authenticated incoming requests
connection.socket.on('message', message => {
// ...
})
})NB
This plugin uses the same router as the fastify instance, this has a few implications to take into account:
- Websocket route handlers follow the usual
fastifyrequest lifecycle, which means hooks, error handlers, and decorators all work the same way as other route handlers. - You can access the fastify server via
thisin your handlers - When using
@fastify/websocket, it needs to be registered before all routes in order to be able to intercept websocket connections to existing routes and close the connection on non-websocket routes.
import Fastify from 'fastify'
import websocket from '@fastify/websocket'
const fastify = Fastify()
await fastify.register(websocket)
fastify.get('/', { websocket: true }, function wsHandler (connection, req) {
// bound to fastify server
this.myDecoration.someFunc()
connection.socket.on('message', message => {
// message.toString() === 'hi from client'
connection.socket.send('hi from server')
})
})
await fastify.listen({ port: 3000 })If you need to handle both HTTP requests and incoming socket connections on the same route, you can still do it using the full declaration syntax, adding a wsHandler property.
'use strict'
const fastify = require('fastify')()
function handle (conn, req) {
conn.pipe(conn) // creates an echo server
}
fastify.register(require('@fastify/websocket'), {
handle,
options: { maxPayload: 1048576 }
})
fastify.register(async function () {
fastify.route({
method: 'GET',
url: '/hello',
handler: (req, reply) => {
// this will handle http requests
reply.send({ hello: 'world' })
},
wsHandler: (conn, req) => {
// this will handle websockets connections
conn.setEncoding('utf8')
conn.write('hello client')
conn.once('data', chunk => {
conn.end()
})
}
})
})
fastify.listen({ port: 3000 }, err => {
if (err) {
fastify.log.error(err)
process.exit(1)
}
})Custom error handler:
You can optionally provide a custom errorHandler that will be used to handle any cleaning up of established websocket connections. The errorHandler will be called if any errors are thrown by your websocket route handler after the connection has been established. Note that neither Fastify's onError hook or functions registered with fastify.setErrorHandler will be called for errors thrown during a websocket request handler.
Neither the errorHandler passed to this plugin or fastify's onError hook will be called for errors encountered during message processing for your connection. If you want to handle unexpected errors within your message event handlers, you'll need to use your own try { } catch {} statements and decide what to send back over the websocket.
const fastify = require('fastify')()
fastify.register(require('@fastify/websocket'), {
errorHandler: function (error, conn /* SocketStream */, req /* FastifyRequest */, reply /* FastifyReply */) {
// Do stuff
// destroy/close connection
conn.destroy(error)
},
options: {
maxPayload: 1048576, // we set the maximum allowed messages size to 1 MiB (1024 bytes * 1024 bytes)
verifyClient: function (info, next) {
if (info.req.headers['x-fastify-header'] !== 'fastify is awesome !') {
return next(false) // the connection is not allowed
}
next(true) // the connection is allowed
}
}
})
fastify.get('/', { websocket: true }, (connection /* SocketStream */, req /* FastifyRequest */) => {
connection.socket.on('message', message => {
// message.toString() === 'hi from client'
connection.socket.send('hi from server')
})
})
fastify.listen({ port: 3000 }, err => {
if (err) {
fastify.log.error(err)
process.exit(1)
}
})Note: Fastify's onError and error handlers registered by setErrorHandler will still be called for errors encountered before the websocket connection is established. This means errors thrown by onRequest hooks, preValidation handlers, and hooks registered by plugins will use the normal error handling mechanisms in Fastify. Once the websocket is established and your websocket route handler is called, fastify-websocket's errorHandler takes over.
Options
@fastify/websocket accept these options for ws :
host- The hostname where to bind the server.port- The port where to bind the server.backlog- The maximum length of the queue of pending connections.server- A pre-created Node.js HTTP/S server.verifyClient- A function which can be used to validate incoming connections.handleProtocols- A function which can be used to handle the WebSocket subprotocols.clientTracking- Specifies whether or not to track clients.perMessageDeflate- Enable/disable permessage-deflate.maxPayload- The maximum allowed message size in bytes.
For more information, you can check ws options documentation.
NB By default if you do not provide a server option @fastify/websocket will bind your websocket server instance to the scoped fastify instance.
NB The path option from ws should not be provided since the routing is handled by fastify itself
NB The noServer option from ws should not be provided since the point of @fastify/websocket is to listen on the fastify server. If you want a custom server, you can use the server option, and if you want more control, you can use the ws library directly
You can also pass the following as connectionOptions for createWebSocketStream.
allowHalfOpenIf set to false, then the stream will automatically end the writable side when the readable side ends. Default: true.readableSets whether the Duplex should be readable. Default: true.writableSets whether the Duplex should be writable. Default: true.readableObjectModeSets objectMode for readable side of the stream. Has no effect if objectMode is true. Default: false.readableHighWaterMarkSets highWaterMark for the readable side of the stream.writableHighWaterMarkSets highWaterMark for the writable side of the stream.
ws does not allow you to set objectMode or writableObjectMode to true
Acknowledgements
This project is kindly sponsored by nearForm.
License
Licensed under MIT.