ws-messaging
Just a really thin abstraction layer on top of WebSocket for Node.js
and Web Browsers with Promise and EventEmitter based APIs.
Features
-
Send and receive notifications (events) via an EventEmitter API.
-
Request-reply API using promises (works in both directions, without
any connection blocking during processing).
-
Built-in auth via WebSocket messages exchange (no more query
strings).
-
Auto reconnection is provided.
-
Binary messages support via custom encoders/decoders.
-
Reasonable client size (also includes a Promise polyfill): around
16KB minified, 5KB gziped.
Table of Contents
Background
Read this
article
for more background information.
Installation
npm i ws-messaging
Usage
On a server:
const Server = require('ws-messaging')
const port = 8000
function connectionHook (client, authData) {
}
let server = new Server({port}, {connectionHook})
On a client:
const Client = require('ws-messaging/client')
const url = `ws://${HOST}:${PORT}`
const auth = { }
let client = new Client(url, auth)
client.on('someEvent', (...data) => { })
client.register('someMethod', (...args) => { })
client.on('connect', () => {
client.send('myEvent', ...someData)
client.invoke('myMethod', ...someArgs)
.then(result => { })
.catch(error => { })
})
Essentially there are two usage patterns that are working in both
directions. Fire and forget via send
/on
, and RPC-style via
invoke
/register
. Unlike on
, only a single handler function can
be registered per a method name.
API
Server API and
Client API
documentation is available online.
Network format description
This section describes what data is actually passed to an encoder.
There are only two types of messages. The first one is for normal
messages:
{
name: string,
args: Array,
id?: number
}
An id
field is present for invoke
calls. The second one is for
ack (replies for invoke
calls) messages:
{
id: number
result?: Object
error?: Object
}
Either a result
or an error
field is included. Note that an
error
is the value returned by an errorFormatter
, by default
String
is used as an errorFormatter
.
Data validation
All incoming data must be validated on a server side, including errors
that are passed to a catch callback. By default only the
network format itself is
validated. Validation can be made by a custom decoder (useful when a
decoder is already using some scheme) or via a receiveHook
, or
inside a handler itself (useful for registered procedures). When
validation is done inside decoder
/receiveHook
, just throw an error
or reject a promise to fail a validation and prevent handlers
execution. Also note that a promise returned by invoke
can be
rejected locally either with Client.ConnectionError
or with
Client.TimeoutError
.
Contribute
If you encounter a bug in this package, please submit a bug report to
github repo issues.
PRs are also accepted.
License
MIT