MongoDB

Support for MongoDB is provided in Feathers via the @feathersjs/mongodb database adapter which uses the MongoDB Client for Node.js. The adapter uses the MongoDB Aggregation Framework, internally, and enables using Feathers' friendly syntax with the full power of aggregation operators. The adapter automatically uses the MongoDB Query API when you need features like Collation.

bash

$ npm install --save @feathersjs/mongodb

tip

The MongoDB adapter implements the common database adapter API and querying syntax.

API

`MongoDBService(options)`

new MongoDBService(options) returns a new service instance initialized with the given options. The following example extends the MongoDBService and then uses the mongodbClient from the app configuration and provides it to the Model option, which is passed to the new MessagesService.

import type { Params } from '@feathersjs/feathers'
import { MongoDBService } from '@feathersjs/mongodb'
import type { MongoDBAdapterParams, MongoDBAdapterOptions } from '@feathersjs/mongodb'

import type { Application } from '../../declarations'
import type { Messages, MessagesData, MessagesQuery } from './messages.schema'

export interface MessagesParams extends MongoDBAdapterParams<MessagesQuery> {}

export class MessagesService<ServiceParams extends Params = MessagesParams> extends MongoDBService<
  Messages,
  MessagesData,
  ServiceParams
> {}

export const messages = (app: Application) => {
  const options: MongoDBAdapterOptions = {
    paginate: app.get('paginate'),
    Model: app.get('mongodbClient').then((db) => db.collection('messages'))
  }
  app.use('messages', new MessagesService(options))
}

Here's an overview of the options object:

Options

MongoDB adapter specific options are:

Model {Promise<MongoDBCollection>} (required) - A Promise that resolves with the MongoDB collection instance. This can also be the return value of an async function without await
disableObjectify {boolean} (optional, default false) - This will disable conversion of the id field to a MongoDB ObjectID if you want to e.g. use normal strings
useEstimatedDocumentCount {boolean} (optional, default false) - If true document counting will rely on estimatedDocumentCount instead of countDocuments

The common API options are:

id {string} (optional, default: '_id') - The name of the id field property. By design, MongoDB will always add an _id property.
id {string} (optional) - The name of the id field property (usually set by default to id or _id).
paginate {Object} (optional) - A pagination object containing a default and max page size
multi {string[]|boolean} (optional, default: false) - Allow create with arrays and patch and remove with id null to change multiple items. Can be true for all methods or an array of allowed methods (e.g. [ 'remove', 'create' ])

There are additionally several legacy options in the common API options

getModel()

getModel([params]) returns a Promise that resolves with the MongoDB collection object. The optional params is the service parameters which may allow to override the collection via params.adapter.

aggregateRaw(params)

The find method has been split into separate utilities for converting params into different types of MongoDB requests. By default, requests are processed by this method and are run through the MongoDB Aggregation Pipeline. This method returns a raw MongoDB Cursor object, which can be used to perform custom pagination or in custom server scripts, if desired.

findRaw(params)

findRaw(params) is used when params.mongodb is set to retrieve data using params.mongodb as the FindOptions object. This method returns a raw MongoDB Cursor object, which can be used to perform custom pagination or in custom server scripts, if desired.

makeFeathersPipeline(params)

makeFeathersPipeline(params) takes a set of Feathers params and converts them to a pipeline array, ready to pass to collection.aggregate. This utility comprises the bulk of the aggregateRaw functionality, but does not use params.pipeline.

Custom Params

The @feathersjs/mongodb adapter utilizes two custom params which control adapter-specific features: params.pipeline and params.mongodb.

params.adapter

Allows to dynamically set the adapter options (like the Model collection) for a service method call.

params.pipeline

Used for aggregation pipelines.

params.mongodb

When making a service method call, params can contain anmongodb property (for example, {upsert: true}) which allows modifying the options used to run the MongoDB query. The adapter will use the collection.find method and not the aggregation pipeline when you use params.mongodb.

Transactions

MongoDB Transactions can be used by passing a session in params.mongodb. For example in a hook:

import { ObjectId } from 'mongodb'
import { HookContext } from '../declarations'

export const myHook = async (context: HookContext) => {
  const { app } = context
  const session = app.get('mongoClient').startSession()

  try {
    await session.withTransaction(async () => {
      const fooData = { message: 'Data for foo' }
      const barData = { text: 'Data for bar' }

      await app.service('fooService').create(fooData, {
        mongodb: { session }
      })
      await app.service('barService').create(barData, {
        mongodb: { session }
      })
    })
  } finally {
    await session.endSession()
  }
}

Indexes

Indexes and unique constraints can be added to the Model Promise, usually in the getOptions in <service>.class:

export const getOptions = (app: Application): MongoDBAdapterOptions => {
  return {
    paginate: app.get('paginate'),
    Model: app
      .get('mongodbClient')
      .then((db) => db.collection('myservice'))
      .then((collection) => {
        collection.createIndex({ email: 1 }, { unique: true })

        return collection
      })
  }
}

info

Note that creating indexes for an existing collection with many entries should be done as a separate operation instead. See the MongoDB createIndex documentation for more information.

Querying

Additionally to the common querying mechanism this adapter also supports MongoDB's query syntax and the update method also supports MongoDB update operators.

Search

Important

Note that in a normal application all MongoDB specific operators have to explicitly be added to the TypeBox query schema or JSON query schema.

There are two ways to perform search queries with MongoDB:

Perform basic Regular Expression matches using the $regex filter.
Perform full-text search using the $search filter.

Basic Regex Search

You can perform basic search using regular expressions with the $regex operator. Here's an example query.

{
  text: { $regex: 'feathersjs', $options: 'igm' },
}

Full-Text Search

See the MongoDB documentation for instructions on performing full-text search using the $search operator:

Perform full-text queries on self-hosted MongoDB.
Perform full-text queries on MongoDB Atlas (MongoDB's first-party hosted database).
Perform full-text queries with the MongoDB Pipeline

Aggregation Pipeline

In Feathers v5 Dove, we added support for the full power of MongoDB's Aggregation Framework and blends it seamlessly with the familiar Feathers Query syntax. All find queries now use the Aggregation Framework, by default.

The Aggregation Framework is accessed through the mongoClient's collection.aggregate method, which accepts an array of "stages". Each stage contains an operator which describes an operation to apply to the previous step's data. Each stage applies the operation to the results of the previous step. It’s now possible to perform any of the Aggregation Stages like $lookup and $unwind, integration with the normal Feathers queries.

Here's how it works with the operators that match the Feathers Query syntax. Let's convert the following Feathers query:

const query = {
  text: { $regex: 'feathersjs', $options: 'igm' },
  $sort: { createdAt: -1 },
  $skip: 20,
  $limit: 10
}

The above query looks like this when converted to aggregation pipeline stages:

;[
  // returns the set of records containing the word "feathersjs"
  { $match: { text: { $regex: 'feathersjs', $options: 'igm' } } },
  // Sorts the results of the previous step by newest messages, first.
  { $sort: { createdAt: -1 } },
  // Skips the first 20 records of the previous step
  { $skip: 20 },
  // returns the next 10 records
  { $limit: 10 }
]

Pipeline Queries

You can use the params.pipeline array to append additional stages to the query. This next example uses the $lookup operator together with the $unwind operator to populate a user attribute onto each message based on the message's userId property.

const result = await app.service('messages').find({
  query: { $sort: { name: 1 } },
  pipeline: [
    {
      $lookup: {
        from: 'users',
        localField: 'userId',
        foreignField: '_id',
        as: 'user'
      }
    },
    { $unwind: { path: '$user' } }
  ],
  paginate: false
})

Aggregation Stages

In the example, above, the query is added to the pipeline, first. Then additional stages are added in the pipeline option:

The $lookup stage creates an array called user which contains any matches in message.userId, so if userId were an array of ids, any matches would be in the users array. However, in this example, the userId is a single id, so...
The $unwind stage turns the array into a single user object.

The above is like doing a join, but without the data transforming overhead like you'd get with an SQL JOIN. If you have properly applied index to your MongoDB collections, the operation will typically execute extremely fast for a reasonable amount of data.

A couple of other notable query stages:

$graphLookup lets you recursively pull in a tree of data from a single collection.
$search lets you do full-text search on fields

All stages of the pipeline happen directly on the MongoDB server.

Read through the full list of supported stages in the MongoDB documentation.

The `$feathers` Stage

The previous section showed how to append stages to a query using params.pipeline. Well, params.pipeline also supports a custom $feathers operator/stage which allows you to specify exactly where in the pipeline the Feathers Query gets injected.

Example: Proxy Permissions

Imagine a scenario where you want to query the pages a user can edit by referencing a permissions collection to find out which pages the user can actually edit. Each record in the permissions record has a userId and a pageId. So we need to find and return only the pages to which the user has access by calling GET /pages from the client.

We could put the following query in a hook to pull the correct pages from the database in a single query THROUGH the permissions collection. Remember, the request is coming in on the pages service, but we're going to query for pages through the permissions collection. Assume we've already authenticated the user, so the user will be found at context.params.user.

// Assume this query on the client
const pages = await app.service('pages').find({ query: {} })

// And put this query in a hook to populate pages "through" the permissions collection
const result = await app.service('permissions').find({
  query: {},
  pipeline: [
    // query all permissions records which apply to the current user
    {
      $match: { userId: context.params.user._id }
    },
    // populate the pageId onto each `permission` record, as an array containing one page
    {
      $lookup: {
        from: 'pages',
        localField: 'pageId',
        foreignField: '_id',
        as: 'page'
      }
    },
    // convert the `page` array into an object, so now we have an array of permissions with permission.page on each.
    {
      $unwind: { path: '$page' }
    },
    // Add a permissionId to each page
    {
      $addFields: {
        'page.permissionId': '$_id'
      }
    },
    // discard the permission and only keep the populated `page`, and bring it top level in the array
    {
      $replaceRoot: { newRoot: '$page' }
    },
    // apply the feathers query stages to the aggregation pipeline.
    // now the query will apply to the pages, since we made the pages top level in the previous step.
    {
      $feathers: {}
    }
  ],
  paginate: false
})

Notice the $feathers stage in the above example. It will apply the query to that stage in the pipeline, which allows the query to apply to pages even though we had to make the query through the permissions service.

If we were to express the above query with JavaScript, the final result would the same as with the following example:

// perform a db query to get the permissions
const permissions = await context.app.service('permissions').find({
  query: {
    userId: context.params.user._id
  },
  paginate: false
})
// make a list of pageIds
const pageIds = permissions.map((permission) => permission.pageId)
// perform a db query to get the pages with matching `_id`
const pages = await context.app.service('pages').find({
  query: {
    _id: {
      $in: pageIds
    }
  },
  paginate: false
})
// key the permissions by pageId for easy lookup
const permissionsByPageId = permissions.reduce((byId, current) => {
  byId[current.pageId] = current
  return byId
}, {})
// Add the permissionId to each `page` record.
const pagesWithPermissionId = pages.map((page) => {
  page.permissionId = permissionByPageId[page._id]._id
  return page
})
// And now apply the original query, whatever the client may have sent, to the pages.
// It might require another database query

Both examples look a bit complex, but te one using aggregation stages will be much quicker because all stages run in the database server. It will also be quicker because it all happens in a single database query!

One more obstacle for using JavaScript this way is that if the user's query changed (from the front end), we would likely be required to edit multiple different parts of the JS logic in order to correctly display results. With the pipeline example, above, the query is very cleanly applied.

Collation

This adapter includes support for collation and case insensitive indexes available in MongoDB v3.4. Collation parameters may be passed using the special collation parameter to the find(), remove() and patch() methods.

Example: Patch records with case-insensitive alphabetical ordering

The example below would patch all student records with grades of 'c' or 'C' and above (a natural language ordering). Without collations this would not be as simple, since the comparison { $gt: 'c' } would not include uppercase grades of 'C' because the code point of 'C' is less than that of 'c'.

const patch = { shouldStudyMore: true }
const query = { grade: { $gte: 'c' } }
const collation = { locale: 'en', strength: 1 }
const patchedStudent = await students.patch(null, patch, { query, collation })

Example: Find records with a case-insensitive search

Similar to the above example, this would find students with a grade of 'c' or greater, in a case-insensitive manner.

const query = { grade: { $gte: 'c' } }
const collation = { locale: 'en', strength: 1 }

const collatedStudents = await students.find({ query, collation })

For more information on MongoDB's collation feature, visit the collation reference page.

ObjectIds

MongoDB uses ObjectId object as primary keys. To store them in the right format they have to be converted from and to strings.

AJV keyword

To validate and convert strings to an object id using AJV, the keywordObjectId AJV keyword helper can be used. It is set up automatically in a generated application using MongoDB.

import { keywordObjectId } from '@feathersjs/mongodb'

const validator = new Ajv()

validator.addKeyword(keywordObjectId)

ObjectIdSchema

Both, @feathersjs/typebox and @feathersjs/schema export an ObjectIdSchema helper that creates a schema which can be both, a MongoDB ObjectId or a string that will be converted with the objectid keyword:

import { ObjectIdSchema } from '@feathersjs/typebox' // or '@feathersjs/schema'

const typeboxSchema = Type.Object({
  userId: ObjectIdSchema()
})

const jsonSchema = {
  type: 'object',
  properties: {
    userId: ObjectIdSchema()
  }
}

Important

The ObjectIdSchema helper will only work when the objectid AJV keyword is registered.

ObjectId resolvers

While the AJV format checks if an object id is valid, it still needs to be converted to the right type. An alternative the the AJV converter is to use Feathers resolvers. The following property resolver helpers can be used.

Important

ObjectId resolvers do not need to be used when using the AJV keyword. They are useful however when using another JSON schema validation library.

resolveObjectId

resolveObjectId resolves a property as an object id. It can be used as a direct property resolver or called with the original value.

import { resolveObjectId } from '@feathersjs/mongodb'

export const messageDataResolver = resolve<Message, HookContext>({
  properties: {
    userId: resolveObjectId
  }
})

export const messageDataResolver = resolve<Message, HookContext>({
  properties: {
    userId: async (value, _message, context) => {
      // If the user is an admin, allow them to create messages for other users
      if (context.params.user.isAdmin && value !== undefined) {
        return resolveObjectId(value)
      }
      // Otherwise associate the record with the id of the authenticated user
      return context.params.user._id
    }
  }
})

resolveQueryObjectId

resolveQueryObjectId allows to query for object ids. It supports conversion from a string to an object id as well as conversion for values from the $in, $nin and $ne query syntax.

import { resolveQueryObjectId } from '@feathersjs/mongodb'

export const messageQueryResolver = resolve<MessageQuery, HookContext>({
  properties: {
    userId: resolveQueryObjectId
  }
})

Dates

While MongoDB has a native Date type, the most reliable way to deal with dates is to send and store them as UTC millisecond timestamps e.g. returned by Date.now() or new Date().getTime() which is also used in the Feathers getting started guide. This has a few advantages:

No conversion between different string types
No timezone and winter/summer time issues
Easier calculations and query-ability

MongoDB ​

API ​

MongoDBService(options) ​

Options ​

getModel() ​

aggregateRaw(params) ​

findRaw(params) ​

makeFeathersPipeline(params) ​

Custom Params ​

params.adapter ​

params.pipeline ​

params.mongodb ​

Transactions ​

Indexes ​

Querying ​

Search ​

Basic Regex Search ​

Full-Text Search ​

Aggregation Pipeline ​

Pipeline Queries ​

Aggregation Stages ​

The $feathers Stage ​

Example: Proxy Permissions ​

Collation ​

ObjectIds ​

AJV keyword ​

ObjectIdSchema ​

ObjectId resolvers ​

resolveObjectId ​

resolveQueryObjectId ​

Dates ​

MongoDB

API

`MongoDBService(options)`

Options

getModel()

aggregateRaw(params)

findRaw(params)

makeFeathersPipeline(params)

Custom Params

params.adapter

params.pipeline

params.mongodb

Transactions

Indexes

Querying

Search

Basic Regex Search

Full-Text Search

Aggregation Pipeline

Pipeline Queries

Aggregation Stages

The `$feathers` Stage

Example: Proxy Permissions

Collation

ObjectIds

AJV keyword

ObjectIdSchema

ObjectId resolvers

resolveObjectId

resolveQueryObjectId

Dates