This module targets at relational database such as MySQL, SQL Server. Heavily inspired by Facebook DataLoader.

$ npm i rdb-dataloader

First i'd like to say DataLoader is a great module that, like it's said, aims to provide per-request cache to save network IO and extremly useful for data service implemented by GraphQL which intergreted with my project recently. I use MySQL as data storage layer which is a common case in real world i believe. However, I found something inconvenient when playing with MySQL:

It is common to have one or more unique key(s) in a table. For example, an User table with these fields: id, name, mobile can have id as the Primary Key and mobile as a Unique Key. So the data service has to response to fetchByIds and fetchByMobiles calls. Then it have to initiate two instances when using DataLoader:

  const DataLoader = require('dataloader');
  const idLoader = new DataLoader(fetchByIds);
  const mobileLoader = new DataLoader(fetchByMobiles);

  idLoader.load(1);
  mobileLoader.load('+86123456');

  function fetchByIds(ids) {
    return db.Users.getByIds(ids);
  }

  function fetchByMobiles(mobiles) {
    return db.Users.getByMobiles(mobiles);
  }

This can work but it's not good enough and the logic is far more complex in real world. Two instances meaning two separate cache, you can imagine that there would be many duplecate records cached in both instance cache just because they are fetched by a different key(id or mobile) and Network Still happens for both sides. Also there would be more instances to initiate as the number of unique keys increases.

There is an old issue and DataLoader fixes now.

This module aims to solve the inconvenience metioned above when playing with relational database. It is recommended that one loader per table. It'd use a single cache and whenever a record is fetched by PK or UK, it'd contribute to the cache. Accordingly, it'd support load by PK or UK to take full advantage of cache.

  const uniqueKeyMap = new Map();
  uniqueKeyMap.set('name', db.fetchByNames); // UK
  uniqueKeyMap.set('email', db.fetchByEmails);  // UK
  const loader = new DataLoader(db.fetchByIds, { uniqueKeyMap });
  loader.load('luckydrq', 'name') // the second argument `name` is to tell loader to fetchByNames
    .then(record => {
      assert(record.id === 1);
      assert(loader._promiseCache.size === 1);
      done();
    }).catch(done);

More examples.

Since this module inherits from DataLoader, it is compatible with the apis of DataLoader. But as you can see from former example, the api is extended. Extended apis are the following:

The constructor accepts all the options that DataLoader accepts. options.cache = true and options.batch = true is set by default and there is no way to override it because it'd be useless without these two features disabled. Alse, there are additional options as below:

• options.primaryKey(optional): String Specify PK name, default is "id";

• options.uniqueKeyMap(optional): ES6 Map

  • key: String UK name
  • value: Function Batch function for UK Use it when you want to fetch by UK.

If you want to load by UK, the second argument should be specified, omitted meaning load by PK.

If you want to load by UK, the second argument should be specified, omitted meaning load by PK.

MIT