unliar/cacheable-response

An HTTP compliant route path middleware for serving cache response with invalidation support.

cacheable-response

An HTTP compliant route path middleware for serving cache response with invalidation support.

Why

Server Side Rendering (SSR) is a luxurious but necessary thing if you want to have a first class user experience.

The main issue of doing server-side things is the extra cost associated with dynamic things: The server will take CPU cycles to compute the value to be served, probably discarded in the next page reload, losing precious resources in the process.

Instead of serving a real time™ – and costly – response, we can say it is OK serving a pre-calculated response but much much cheaper.

That will save CPU cycles, saving them for things that really matters.

Install

$ npm install cacheable-response --save

Get Started

cacheable-response is a HTTP middleware for a serving pre-calculated response.

It's like a LRU cache but with all the logic necessary for auto-invalidate response copies and refresh them.

Imagine you are currently running an HTTP microservice to compute something heavy in terms of CPU

const server = ({ req, res }) => {
  const data = doSomething(req)
  res.send(data)
}

To leverage caching capabilities, just you need to adapt your HTTP based project a bit for following cacheable-response interface

const cacheableResponse = require('cacheable-response')

const ssrCache = cacheableResponse({
  get: ({ req, res }) => ({
    data: doSomething(req),
    ttl: 86400000 // 24 hours
  }),
  send: ({ data, res, req }) => res.send(data)
})

At least, cacheable-response needs two things:

• get: It creates a fresh cacheable response associated with the current route path.
• send: It determines how the response should be rendered.

cacheable-response is framework agnostic: It could be used with any library that accepts (request, response) as input.

const micro = require('micro')

/* Explicitly pass `cacheable-response` as server */
micro((req, res) => ssrCache({ req, res }))

That's include any express-like framework as well.

const express = require('express')
const app = express()

/* Passing `cacheable-response` instance as middleware */
app.use((req, res) => ssrCache({ req, res }))

See more examples.

At all times the cache status is reflected as x-cache headers in the response.

The first resource access will be a MISS.

HTTP/2 200
cache-control: public, max-age=7200, stale-while-revalidate=300
ETag: "d-pedE0BZFQNM7HX6mFsKPL6l+dUo"
x-cache-status: MISS
x-cache-expired-at: 1h 59m 60s

Successive resource access under the ttl period returns a HIT

HTTP/2 200
cache-control: public, max-age=7170, stale-while-revalidate=298
ETag: "d-pedE0BZFQNM7HX6mFsKPL6l+dUo"
x-cache-status: HIT
x-cache-expired-at: 1h 59m 30s

After ttl period expired, the cache will be invalidated and refreshed in the next request.

In case you need you can force invalidate a cache response passing force=true as part of your query parameters.

curl https://myserver.dev/user # MISS (first access)
curl https://myserver.dev/user # HIT (served from cache)
curl https://myserver.dev/user # HIT (served from cache)
curl https://myserver.dev/user?force=true # BYPASS (skip cache copy)

In that case, the x-cache-status will reflect a 'BYPASS' value.

Additionally, you can configure a stale ttl:

const cacheableResponse = require('cacheable-response')

const ssrCache = cacheableResponse({
  get: ({ req, res }) => ({
    data: doSomething(req),
    ttl: 86400000, // 24 hours
    staleTtl: 3600000 // 1h
  }),
  send: ({ data, res, req }) => res.send(data)
})

The stale ttl maximizes your cache HITs, allowing you to serve a no fresh cache copy while doing revalidation on the background.

curl https://myserver.dev/user # MISS (first access)
curl https://myserver.dev/user # HIT (served from cache)
curl https://myserver.dev/user # STALE (23 hours later, background revalidation)
curl https://myserver.dev/user?force=true # HIT (fresh cache copy for the next 24 hours)

The library provides enough good sensible defaults for most common scenarios and you can tune these values based on your use case.

API

cacheableResponse([options])

options

bypassQueryParameter

Type: boolean
Default: 'force'

The name of the query parameter to be used for skipping the cache copy in an intentional way.

cache

Type: boolean
Default: new Keyv({ namespace: 'ssr' })

The cache instance used for backed your pre-calculated server side response copies.

The library delegates in keyv, a tiny key value store with multi adapter support.

If you don't specify it, a memory cache will be used.

compress

Type: boolean
Default: false

Enable compress/decompress data using brotli compression format.

If you enable it, you need to an additional iltorb package:

npm install iltorb

get

Required
Type: function

The method to be called for creating a fresh cacheable response associated with the current route path.

async function get ({ req, res }) {
  const data = doSomething(req, res)
  const ttl = 86400000 // 24 hours
  const headers = { userAgent: 'cacheable-response' }
  return { data, ttl, headers }
}

The method will received ({ req, res }) and it should be returns:

• data object|string: The content to be saved on the cache.
• ttl number: The quantity of time in milliseconds the content is considered valid on the cache. Don't specify it means use default ttl.
• createdAt date: The timestamp associated with the content (Date.now() by default).

Any other property can be specified and will passed to .send.

In case you want to bypass the cache, preventing caching a value (e.g., when an error occurred), you should return undefined or null.

key

Type: function
Default: ({ req }) => req.url)

It determinates how the cache key should be computed, receiving req, res as input.

send

Required
Type: function

The method used to determinate how the content should be rendered.

async function send ({ req, res, data, headers }) {
  res.setHeader('user-agent', headers.userAgent)
  res.send(data)
}

It will receive ({ req, res, data, ...props }) being props any other data supplied to .get.

staleTtl

Type: number|boolean
Default: 3600000

Number of milliseconds that indicates grace period after response cache expiration for refreshing it in the background. the latency of the refresh is hidden from the user.

The value will be associated with stale-while-revalidate directive.

You can pass a false to disable it.

ttl

Type: number
Default: 86400000

Number of milliseconds a cache response is considered valid.

After this period of time, the cache response should be refreshed.

This value can be specified as well providing it as part of .get output.

If you don't provide one, this be used as fallback for avoid keep things into cache forever.

serialize

Type: function
Default: JSON.stringify

Set the serializer method to be used before compress.

deserialize

Type: function
Default: JSON.parse

Set the deserialize method to be used after decompress.

Pro-tip: Distributed cache with CloudFlare™️

This content is not sponsored; Just I consider CloudFlare is doing a good job offering a cache layer as part of their free tier.

Imagine what could be better than having one cache layer?

Exactly, two cache layers.

If your server domain is connected with CloudFlare you can take advantage of having a distributed CDN that also caches your responses.

For doing that, you need to setup a Page Rule over your domain specifing you want to enable cache. Read more how to do that.

Next time you query about a resource, a new cf-cache-status appeared as part of your headers response.

HTTP/2 200
cache-control: public, max-age=7200, stale-while-revalidate=300
ETag: "d-pedE0BZFQNM7HX6mFsKPL6l+dUo"
x-cache-status: MISS
x-cache-expired-at: 1h 59m 60s
cf-cache-status: MISS

CloudFlare will respect your cache-control policy, creating another caching layer reflected by cf-cache-status

HTTP/2 200
cache-control: public, max-age=7200, stale-while-revalidate=300
ETag: "d-pedE0BZFQNM7HX6mFsKPL6l+dUo"
x-cache-status: MISS
x-cache-expired-at: 1h 59m 60s
cf-cache-status: HIT

Note how in this second request x-cache-status is still MISS.

That's because CloudFlare way for caching the content includes caching the response headers.

The headers associated with the cache copy will the headers from the first request. You need to look at cf-cache-status instead.

You can have a better overview of the percentage of success by looking your CloudFlare domain analytics

Examples

Make a PR for adding your project!

Express

• unavatar.now.sh
• html-microservice

Next.js

• ssr-caching

Bibliography

• Server rendered pages are not optional, by Guillermo Rauch.
• Increasing the Performance of Dynamic Next.JS Websites, by scale AI.
• The Benefits of Microcaching, by NGINX.
• Cache-Control for Civilians, by Harry Robert

License

cacheable-response © Kiko Beats, released under the MIT License.
Authored and maintained by Kiko Beats with help from contributors.

kikobeats.com · GitHub Kiko Beats · Twitter @Kikobeats