Batching and Caching layer using Redis as the Caching layer. Redis Dataloader wraps Facebook Dataloader, adding Redis as a caching layer.
npm install redis-dataloader
or
yarn add redis-dataloader
const redisClient = require('redis').createClient();
// the "ioredis" module is also supported
// const Redis = require('ioredis');
// const redisClient = new Redis();
const DataLoader = require('dataloader');
const RedisDataLoader = require('redis-dataloader')({ redis: redisClient });
const loader = new RedisDataLoader(
// set a prefix for the keys stored in redis. This way you can avoid key
// collisions for different data-sets in your redis instance.
'redis-key-prefix',
// create a regular dataloader. This should always be set with caching disabled.
new DataLoader(myBatchLoadFunction, { cache: false }),
// The options here are the same as the regular dataloader options, with
// the additional option "expire"
{
// caching here is a local in memory cache. Caching is always done
// to redis.
cache: true,
// if set redis keys will be set to expire after this many seconds
// this may be useful as a fallback for a redis cache.
expire: 60,
// can include a custom serialization and deserialization for
// storage in redis.
serialize: date => date.getTime(),
deserialize: timestamp => new Date(timestamp),
// Set this to true to return Buffer objects to the deserialize function
// when using the ioredis driver.
buffer: false
}
);
// load an individual item by its key
loader.load(5).then(resp => console.log(resp));
//clear an individiaul item from the local and redis cache.
loader.clear(5).then(() => {})
In general, RedisDataLoader has the same API as the Facebook Dataloader Api, with a few differences. Read through the Facebook Dataloader documentation and then note the differences mentioned here.
clear
returns a promise (waits until redis succeeds at deleting the key). Facebook Dataloader'sclear
method is synchronous.clearAll
is not available (redis does not have an efficient way to do this?)prime
will always overwrite the cache. Facebook Dataloader will only write to its cache if a value is not already present. Prime is asyncronous and returns a Promise.- dataloader results must be either
null
or a JSON object. - two functions:
clearLocal(key)
andclearAllLocal()
allow you to clear the local cache only.
const redis = require('redis').createClient();
const RedisDataLoader = require('redis-dataloader')({ redis: redis });
Each Dataloader holds its own local in memory cache (Same as Facebook Dataloader), and additionally caches to your Redis instance.
const loader = new RedisDataLoader('<redis key prefix>', '<Facebook Dataloader>', '<Options>');
Specify a Prefix that will be appended to each key when storing in Redis.
So for example if your prefix is "bar" and you call loader.load('foo')
, this key
will be stored in Redis as bar:foo
A regular Facebook Dataloader is passed in as the second parameter. It will be used to fetch data from your underlying datastore (mongo, sql, whatever). It is very important to disable the cache on this dataloader. Redis dataloader will already do local in memory caching (unless you disable it).
All the options available to Facebook Dataloader can be passed in here. An additional option called expire is also available, and will set a ttl in seconds on all keys set in redis if this option is passed.
The cacheKeyFn
will default to serialize objects and arrays using json-stable-stringify and allow all other values to pass through unchanged.
buffer
will pass a Buffer object to the deserialize function rather than a string.
If not using the ioredis driver this will throw an error at instation.
The purpose of Redis Dataloader is to provide a caching layer in redis on top of the Facebook Dataloader. Facebook's Dataloader provides a local in memory cache. This may be ok for short lived per-request caches, but may not be sufficient if you need a long lived cache and/or you have multiple webservers that need to share data.
Redis Dataloader will additionally use the same local cache that Facebook Dataloader provides. It will first check the local cache, then check the redis cache, before finally checking your underlying datastore. This pattern may be desirable if for example you create a new DataLoader for each request. If your dataloader is long-lived you may want to disable to the local cache, and just rely on the redis cache instead
const loader = new RedisDataLoader('prefix', new DataLoader(), { cache: false });
- Install Dependencies
npm install
- Start Redis
docker-compose stop && docker-compose rm && docker-compose build && docker-compose up -d
- Run Tests
npm test