orq

orq - Observable Request Queue

HTTP Request Queue Optimized for Web Workers.

Features

• TTL cache
• RESTful cache policy
• Cancelable requests
• Duplicate request elimination
• Works with Web Workers
• Platform independent

Usage

worker.js

import { mkMemCache, mkCachePolicy, mkReceiver } from 'orq'
import request from '@orq/superagent'

const fiveMinutes = 300000
const applyCachePolicy = mkCachePolicy({ ttl: fiveMinutes })
const cache = applyCachePolicy(mkMemCache())
mkReceiver(self, request, cache)

main.js

import mkInterface from 'orq/interface'

const orqWorker = new Worker('worker.js')

const orq = mkInterface(orqWorker)

Featrues in detail

TTL cache

mkCachePolicy({
  ttl: 3600000, // 1 hour in milliseconds
  ressources: {
    'https://example.com/api': {
      '/fish': {
        // override default ttl, since fish get stale quickly
        ttl: 180000, // 3 minutes in milliseconds
      },
    },
  },
})

RESTful cache policy

⚠️ Pseudo code ahead

orq.addRequest('https://example.com/api/fish')
  .subscribe() // network call made, will now be served from cache
orq.addRequest('https://example.com/api/fish/42', { method: 'PUT' })
  .subscribe() // invalidates /fish/* and /fish caches
orq.addRequest('https://example.com/api/fish/42')
  .subscribe() // another network call is made, since cache was invalidated before

Cancelable requests

const requestSub = orq.addRequest('https://example.com/api/fish', { cancelable: true })
  .subscribe()
setTimeout(() => {
  requestSub.unsubscribe() // Signals the worker to cancel the request
}, 1000)

By default all requests are cancelable except GET requests. The reasoning is, that the user might request the same resource again, at which point the response can be served from cache. So that's why we have to pass the option explicitly in the above GET request.

Duplicate request elimination

orq.addRequest('https://example.com/api/fish')
  .subscribe() // Starts network call from worker
orq.addRequest('https://example.com/api/fish')
  .subscribe() // Doesn't start another network call, since the first hasn't been completed. Instead the result of the first will be served to this request too.

Platform independent

The request implementation isn't implemented by orq itself. So you may write your own. For example by wrapping the node internal http/https modules. orq only uses a subset of the worker API. You could easily wrap the node cluster module to provide a worker like API. Those wrapped node master/worker can then be passed to orq mkInterface and mkReceiver. See test helpers to see how this could be done.

TODO

• update deps
• write @orq/superagent
• cannot add type constraints on request and response bodies
• cache policy should limit cache item size

Developed at Vimcar.

Changelog

0.3.0 (2018-05-31)

Bug Fixes

• use flow-copy-source instead of flow gen-flow-files (516a28b)

Features

• allow passing redirects option to request client (f67c8a4)
• allow retrying requests (94b9881)

<a name="0.2.1"></a>