Add support for asynchronous cacheMap

[stanogurnik]

When dealing with large datasets it would be nice to be able to use also other cache drivers as Redis, memcache etc. as it could be too much for just memory storage, so DataLoader should be able to handle Promise specifically when using get() to retrieve cached Promise on cacheMap ...

[vieks]

This is exactly the issue I have created #23 but I auto-closed it because it's silly.

The purpose of Dataloader is to prevent at most round trips generated by a data source (Redis, PostgreSQL, whatever...) induced by their network nature. So permitting to create a async remote cache goes against the benefit of using Dataloader hence becoming useless.

Strictly speaking Dataloader doesn't realy implement a cache but a memoize mechanism.

Furthermore, if you look at the source: _promiseCache: CacheMap<K, Promise<V>>; Dataloader memoize/cache Promise objects, no pure raw data types, so you can't serialize them into a remote cache.

If you want a real cache in front of a slow data source/store/base, make it in an upstream batch function.

as it could be too much for just memory storage

Or at least use the custom cache api with an in-memory LRU local cache:

LRU node LRU memoize

[stanogurnik]

@vieks it makes sense in how you put it, thank you for explanation & closing issue.

[leebyron]

@vieks is correct here. DataLoader's cache is purely a local (in memory per instance) cache. It is not a distributed cache like memcached.

Here is a rough example of how you might implement a "read through" distributed cache with DataLoader:

function loadValue(id) {
  return fastLoader.load(id).catch(err => slowLoader.load(id));
}

let fastLoader = new DataLoader(ids => batchLoadFromMemcache(ids));

let slowLoader = new DataLoader(ids => batchLoadFromSQL(ids).then(values => {
  writeToMemcache(ids, values); // fire and forget
  return values;
}));

[eugenehp]

Hey Lee, @leebyron

Do you think this concept could be used on the level of the internal Cache Storage?

Thanks.