Skip to content

vidarc/async-cache-dedupe

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

46 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

async-cache-dedupe

async-cache-dedupe is a cache for asynchronous fetching of resources with full deduplication, i.e. the same resource is only asked once at any given time.

Install

npm i async-cache-dedupe

Example

import { createCache } from 'async-cache-dedupe'

const cache = createCache({
  ttl: 5, // seconds
  storage: { type: 'memory' },
})

cache.define('fetchSomething', async (k) => {
  console.log('query', k)
  // query 42
  // query 24

  return { k }
})

const p1 = cache.fetchSomething(42)
const p2 = cache.fetchSomething(24)
const p3 = cache.fetchSomething(42)

const res = await Promise.all([p1, p2, p3])

console.log(res)
// [
//   { k: 42 },
//   { k: 24 }
//   { k: 42 }
// ]

Commonjs/require is also supported.

API

createCache(opts)

Creates a new cache.

Options:

  • ttl: the maximum time a cache entry can live, default 0; if 0, an element is removed from the cache as soon as the promise resolves.
  • onDedupe: a function that is called every time it is defined is deduped.
  • onError: a function that is called every time there is a cache error.
  • onHit: a function that is called every time there is a hit in the cache.
  • onMiss: a function that is called every time the result is not in the cache.
  • storage: the storage options; default is { type: "memory" } Storage options are:
    • type: memory (default) or redis
    • options: by storage type
      • for memory type

        • size: maximum number of items to store in the cache per resolver. Default is 1024.
        • invalidation: enable invalidation, see invalidation. Default is disabled.
        • log: logger instance pino compatible, default is disabled.

        Example

        createCache({ storage: { type: 'memory', options: { size: 2048 } } })
      • for redis type

        • client: a redis client instance, mandatory. Should be an ioredis client or compatible.
        • invalidation: enable invalidation, see invalidation. Default is disabled.
        • invalidation.referencesTTL: references TTL in seconds, it means how long the references are alive; it should be set at the maximum of all the caches ttl.
        • log: logger instance pino compatible, default is disabled.

        Example

        createCache({ storage: { type: 'redis', options: { client: new Redis(), invalidation: { referencesTTL: 60 } } } })

cache.define(name[, opts], original(arg, cacheKey))

Define a new function to cache of the given name.

The define method adds a cache[name] function that will call the original function if the result is not present in the cache. The cache key for arg is computed using safe-stable-stringify and it is passed as the cacheKey argument to the original function.

Options:

  • ttl: the maximum time a cache entry can live, default as defined in the cache; default is zero, so cache is disabled, the function will be only the deduped.

  • serialize: a function to convert the given argument into a serializable object (or string).

  • onDedupe: a function that is called every time there is defined is deduped.

  • onError: a function that is called every time there is a cache error.

  • onHit: a function that is called every time there is a hit in the cache.

  • onMiss: a function that is called every time the result is not in the cache.

  • storage: the storage to use, same as above. It's possible to specify different storages for each defined function for fine-tuning.

  • references: sync or async function to generate references, it receives (args, key, result) from the defined function call and must return an array of strings or falsy; see invalidation to know how to use them. Example

      const cache = createCache({ ttl: 60 })
    
      cache.define('fetchUser', {
        references: (args, key, result) => result ? [`user~${result.id}`] : null
      }, 
      (id) => database.find({ table: 'users', where: { id }}))
    
      await cache.fetchUser(1)

cache.clear([name], [arg])

Clear the cache. If name is specified, all the cache entries from the function defined with that name are cleared. If arg is specified, only the elements cached with the given name and arg are cleared.

cache.invalidateAll(references, [storage])

cache.invalidateAll perform invalidation over the whole storage; if storage is not specified - using the same name as the defined function, invalidation is made over the default storage.

references can be:

  • a single reference
  • an array of references (without wildcard)
  • a matching reference with wildcard, same logic for memory and redis

Exmple

const cache = createCache({ ttl: 60 })

cache.define('fetchUser', {
  references: (args, key, result) => result ? [`user:${result.id}`] : null
}, (id) => database.find({ table: 'users', where: { id }}))

cache.define('fetchCountries', {
  storage: { type: 'memory', size: 256 },
  references: (args, key, result) => [`countries`]
}, (id) => database.find({ table: 'countries' }))

// ...

// invalidate all users from default storage
cache.invalidateAll('user:*')

// invalidate user 1 from default storage
cache.invalidateAll('user:1')

// invalidate user 1 and user 2 from default storage
cache.invalidateAll(['user:1', 'user:2'])

// note "fetchCountries" uses a different storage
cache.invalidateAll('countries', 'fetchCountries')

See below how invalidation and references work.

Invalidation

Along with time to live invalidation of the cache entries, we can use invalidation by keys.
The concept behind invalidation by keys is that entries have an auxiliary key set that explicitly links requests along with their own result. These auxiliary keys are called here references.
A scenario. Let's say we have an entry user {id: 1, name: "Alice"}, it may change often or rarely, the ttl system is not accurate:

  • it can be updated before ttl expiration, in this case the old value is shown until expiration by ttl.
  • it's not been updated during ttl expiration, so in this case, we don't need to reload the value, because it's not changed

To solve this common problem, we can use references.
We can say that the result of defined function getUser(id: 1) has reference user~1, and the result of defined function findUsers, containing {id: 1, name: "Alice"},{id: 2, name: "Bob"} has references [user~1,user~2]. So we can find the results in the cache by their references, independently of the request that generated them, and we can invalidate by references.

So, when a writing event involving user {id: 1} happens (usually an update), we can remove all the entries in the cache that have references to user~1, so the result of getUser(id: 1) and findUsers, and they will be reloaded at the next request with the new data - but not the result of getUser(id: 2).

Explicit invalidation is disabled by default, you have to enable it in storage settings.

See mercurius-cache-example for a complete example.

Redis

Using a redis storage is the best choice for a shared and/or large cache.
All the references entries in redis have referencesTTL, so they are all cleaned at some time. referencesTTL value should be set at the maximum of all the ttls, to let them be available for every cache entry, but at the same time, they expire, avoiding data leaking.
Anyway, we should keep references up-to-date to be more efficient on writes and invalidation, using the garbage collector function, that prunes the expired references: while expired references do not compromise the cache integrity, they slow down the I/O operations.
Storage memory doesn't have gc.

Redis garbage collector

As said, While the garbage collector is optional, is highly recommended to keep references up to date and improve performances on setting cache entries and invalidation of them.

storage.gc([mode], [options])

  • mode: lazy (default) or strict. In lazy mode, only a chunk of the references are randomly checked, and probably freed; running lazy jobs tend to eventually clear all the expired references. In strict mode, all the references are checked and freed, and after that, references and entries are perfectly clean. lazy mode is the light heuristic way to ensure cached entries and references are cleared without stressing too much redis, strict mode at the opposite stress more redis to get a perfect result. The best strategy is to combine them both, running often lazy jobs along with some strict ones, depending on the size of the cache.

Options:

  • chunk: the chunk size of references analyzed per loops, default 64
  • lazy~chunk: the chunk size of references analyzed per loops in lazy mode, default 64; if both chunk and lazy.chunk is set, the maximum one is taken
  • lazy~cursor: the cursor offset, default zero; cursor should be set at report.cursor to continue scanning from the previous operation

Return report of the gc job, as follows

"report":{
  "references":{
      "scanned":["r:user:8", "r:group:11", "r:group:16"],
      "removed":["r:user:8", "r:group:16"]
  },
  "keys":{
      "scanned":["users~1"],
      "removed":["users~1"]
  },
  "loops":4,
  "cursor":0,
  "error":null
}

Example

import { createCache, createStorage } from 'async-cache-dedupe'

const cache = createCache({
  ttl: 5,
  storage: { type: 'redis', options: { client: redisClient, invalidation: true } },
})
// ... cache.define('fetchSomething'

const storage = createStorage('redis', { client: redisClient, invalidation: true })

let cursor
setInterval(() => {
  const report = await storage.gc('lazy', { lazy: { cursor } })
  if(report.error) {
    console.error('error on redis gc', error)
    return
  }
  console.log('gc report (lazy)', report)
  cursor = report.cursor
}, 60e3).unref()

setInterval(() => {
  const report = await storage.gc('strict', { chunk: 128 })
  if(report.error) {
    console.error('error on redis gc', error)
    return
  }
  console.log('gc report (strict)', report)
}, 10 * 60e3).unref()

Maintainers


Breaking Changes

  • version 0.5.0 -> 0.6.0
    • options.cacheSize is dropped in favor of storage

License

MIT

About

Async cache with dedupe support

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • JavaScript 99.6%
  • Shell 0.4%