Koa.js service and a Middleware hook loader for @atlas.js.
npm i @atlas.js/koa
The service configuration allows you to define three things:
listen
: A bind configuration, ie. to which host and port to bindkoa
: Koa configuration, ie. all the properties you can set on the Koa instanceserver
: http configuration, ie. all the properties you can set on the http server instance
import { Atlas } from '@atlas.js/atlas'
import * as Koa from '@atlas.js/koa'
const atlas = new Atlas({
config: {
services: {
http: {
// Listen on this port and network interface
listen: {
port: 3000,
hostname: '127.0.0.1',
},
// These are set on the http.Server instance
server: {
timeout: 30000,
},
// These are set on the Koa instance
koa: {
proxy: true,
},
// If set to an object, will be used to load all middleware found in this module into the
// Koa instance
middleware: {
module: 'path/to/middleware',
config: {
bodyparser: {},
// ...
}
}
}
}
}
})
atlas.service('http', Koa.Server)
await atlas.start()
// The Koa instance is now available here:
atlas.services.http
// And the http.Server instance is also exposed:
atlas.services.http.server
Here is an example middleware module that the service supports.
// middleware.js
import forcehttps from './forcehttps'
import routes from './routes'
import notfound from './notfound'
export default {
forcehttps,
routes,
notfound,
}
The Atlas instance can be accessed through
ctx.atlas
in middleware or routes.
Here is an example middleware that makes use of the Atlas instance inside the route handler. It returns 400 status code with a custom message when the request is made on an insecure protocol.
// middleware/forcehttps.js
export default function mkforcehttps(config) {
return function forcehttps(ctx, next) {
// Here you can access the Atlas instance via `ctx.atlas`
if (ctx.atlas.env === 'production' && !ctx.secure) {
ctx.response.status = 426
ctx.response.set({
Upgrade: 'TLS/1.2, HTTP/1.1',
Connection: 'Upgrade',
})
ctx.response.body = {
message: 'I refuse to talk to you while anyone may be listening.',
}
return
}
await next()
}
}
This hook allows you to extend the Koa context object prototype with custom functions or properties. It might be useful to define response type aliases, such as ctx.ok()
, or ctx.forbidden()
.
service:koa
: A Koa service on which to extend the context
const atlas = new Atlas({
config: {
hooks: {
context: {
// The path to the module, relative to root, which should be loaded and
// properties/functions from that module added to koa.context
module: 'server/context',
}
}
}
})
atlas.service('http', Koa.Server)
atlas.hook('context', Koa.ContextHook, {
aliases: {
'service:koa': 'http'
}
})
await atlas.start()
// server/context.js
export default {
ok(body = {}) {
this.status = 200
this.body = body
},
forbidden() {
this.status = 403
this.body = {
error: 'Forbidden'
}
}
}
This hook extends the Koa instance with websocket protocol support, using koa-websocket.
service:koa
: A Koa service on which to add the websocket protocol support
const atlas = new Atlas({
config: {
hooks: {
websocket: {
// This has the same structure and purpose as the Koa service's middleware config: it allows
// you to add websocket-specific middleware to the server.
middleware: {
module: 'path/to/websocket/middleware',
config: {}
},
// This goes directly to the websocket protocol's constructor. Note that you should not
// use `host`, `port` or `server` options since the server instance is re-used from the
// underlying Koa server and creating a new http server could cause unwanted side-effects.
// See: https://github.com/websockets/ws/blob/master/doc/ws.md#new-websocketserveroptions-callback
listen: {},
}
}
}
})
Once attached to a Koa servise, the websocket interface is accessible as per the library's definition via koa.ws
, which in Atlas it would be:
atlas.services.http.ws
See the LICENSE file for information.