Apollo data source for MongoDB

npm i apollo-datasource-mongodb

This package uses DataLoader for batching and per-request memoization caching. It also optionally (if you provide a ttl) does shared application-level caching (using either the default Apollo InMemoryLRUCache or the cache you provide to ApolloServer()). It does this only for these two methods:

• findOneById(id, options)
• findManyByIds(ids, options)

Contents:

• Usage
  • Basic
  • Batching
  • Caching
• API
  • findOneById
  • findManyByIds
  • deleteFromCacheById

Usage

Basic

The basic setup is subclassing MongoDataSource, passing your collection or Mongoose model to the constructor, and using the API methods:

data-sources/Users.js

import { MongoDataSource } from 'apollo-datasource-mongodb'

export default class Users extends MongoDataSource {
  getUser(userId) {
    return this.findOneById(userId)
  }
}

and:

import { MongoClient } from 'mongodb'

import Users from './data-sources/Users.js'

const client = new MongoClient('mongodb://localhost:27017/test')
client.connect()

const server = new ApolloServer({
  typeDefs,
  resolvers,
  dataSources: () => ({
    users: new Users(client.db().collection('users'))
    // OR
    // users: new Users(UserModel)
  })
})

Inside the data source, the collection is available at this.collection (e.g. this.collection.update({_id: 'foo, { $set: { name: 'me' }}})). The model (if applicable) is available at this.model (new this.model({ name: 'Alice' })). The request's context is available at this.context. For example, if you put the logged-in user's ID on context as context.currentUserId:

class Users extends MongoDataSource {
  ...

  async getPrivateUserData(userId) {
    const isAuthorized = this.context.currentUserId === userId
    if (isAuthorized) {
      const user = await this.findOneById(userId)
      return user && user.privateData
    }
  }
}

If you want to implement an initialize method, it must call the parent method:

class Users extends MongoDataSource {
  initialize(config) {
    super.initialize(config)
    ...
  }
}

If you're passing a Mongoose model rather than a collection, Mongoose will be used for data fetching. All transformations definded on that model (virtuals, plugins, etc.) will be applied to your data before caching, just like you would expect it. If you're using reference fields, you might be interested in checking out mongoose-autopopulate.

Batching

This is the main feature, and is always enabled. Here's a full example:

class Users extends MongoDataSource {
  getUser(userId) {
    return this.findOneById(userId)
  }
}

class Posts extends MongoDataSource {
  getPosts(postIds) {
    return this.findManyByIds(postIds)
  }
}

const resolvers = {
  Post: {
    author: (post, _, { dataSources: { users } }) => users.getUser(post.authorId)
  },
  User: {
    posts: (user, _, { dataSources: { posts } }) => posts.getPosts(user.postIds)
  }
}

const server = new ApolloServer({
  typeDefs,
  resolvers,
  dataSources: () => ({
    users: new Users(db.collection('users')),
    posts: new Posts(db.collection('posts'))
  })
})

Caching

To enable shared application-level caching, you do everything from the above section, and you add the ttl option to findOneById():

const MINUTE = 60

class Users extends MongoDataSource {
  getUser(userId) {
    return this.findOneById(userId, { ttl: MINUTE })
  }

  updateUserName(userId, newName) {
    this.deleteFromCacheById(userId)
    return this.collection.updateOne({
      _id: userId
    }, {
      $set: { name: newName }
    })
  }
}

const resolvers = {
  Post: {
    author: (post, _, { users }) => users.getUser(post.authorId)
  },
  Mutation: {
    changeName: (_, { userId, newName }, { users, currentUserId }) =>
      currentUserId === userId && users.updateUserName(userId, newName)
  }
}

Here we also call deleteFromCacheById() to remove the user from the cache when the user's data changes. If we're okay with people receiving out-of-date data for the duration of our ttl—in this case, for as long as a minute—then we don't need to bother adding calls to deleteFromCacheById().

API

findOneById

this.findOneById(id, { ttl })

Resolves to the found document. Uses DataLoader to load id. DataLoader uses collection.find({ _id: { $in: ids } }). Optionally caches the document if ttl is set (in whole seconds).

findManyByIds

this.findManyByIds(ids, { ttl })

Calls findOneById() for each id. Resolves to an array of documents.

deleteFromCacheById

this.deleteFromCacheById(id)

Deletes a document from the cache.