Populate

MongoDB has the join-like $lookup aggregation operator in versions >= 3.2. Mongoose has a more powerful alternative called populate(), which lets you reference documents in other collections.

Population is the process of automatically replacing the specified paths in the document with document(s) from other collection(s). We may populate a single document, multiple documents, a plain object, multiple plain objects, or all objects returned from a query. Let's look at some examples.

const mongoose = require('mongoose');
const { Schema } = mongoose;

const personSchema = Schema({
  _id: Schema.Types.ObjectId,
  name: String,
  age: Number,
  stories: [{ type: Schema.Types.ObjectId, ref: 'Story' }]
});

const storySchema = Schema({
  author: { type: Schema.Types.ObjectId, ref: 'Person' },
  title: String,
  fans: [{ type: Schema.Types.ObjectId, ref: 'Person' }]
});

const Story = mongoose.model('Story', storySchema);
const Person = mongoose.model('Person', personSchema);

So far we've created two Models. Our Person model has its stories field set to an array of ObjectIds. The ref option is what tells Mongoose which model to use during population, in our case the Story model. All _ids we store here must be document _ids from the Story model.

Saving refs

Saving refs to other documents works the same way you normally save properties, just assign the _id value:

const author = new Person({
  _id: new mongoose.Types.ObjectId(),
  name: 'Ian Fleming',
  age: 50
});

await author.save();

const story1 = new Story({
  title: 'Casino Royale',
  author: author._id // assign the _id from the person
});

await story1.save();
// that's it!

You can set the ref option on ObjectId, Number, String, and Buffer paths. populate() works with ObjectIds, numbers, strings, and buffers. However, we recommend using ObjectIds as _id properties (and thus ObjectIds for ref properties) unless you have a good reason not to. That is because MongoDB will set _id to an ObjectId if you create a new document without an _id property, so if you make your _id property a Number, you need to be extra careful not to insert a document without a numeric _id.

Population

So far we haven't done anything much different. We've merely created a Person and a Story. Now let's take a look at populating our story's author using the query builder:

const story = await Story.
  findOne({ title: 'Casino Royale' }).
  populate('author').
  exec();
// prints "The author is Ian Fleming"
console.log('The author is %s', story.author.name);

Populated paths are no longer set to their original _id , their value is replaced with the mongoose document returned from the database by performing a separate query before returning the results.

Arrays of refs work the same way. Just call the populate method on the query and an array of documents will be returned in place of the original _ids.

Setting Populated Fields

You can manually populate a property by setting it to a document. The document must be an instance of the model your ref property refers to.

const story = await Story.findOne({ title: 'Casino Royale' });
story.author = author;
console.log(story.author.name); // prints "Ian Fleming"

You can also push documents or POJOs onto a populated array, and Mongoose will add those documents if their ref matches.

const fan1 = await Person.create({ name: 'Sean' });
await Story.updateOne({ title: 'Casino Royale' }, { $push: { fans: { $each: [fan1._id] } } });

const story = await Story.findOne({ title: 'Casino Royale' }).populate('fans');
story.fans[0].name; // 'Sean'

const fan2 = await Person.create({ name: 'George' });
story.fans.push(fan2);
story.fans[1].name; // 'George'

story.fans.push({ name: 'Roger' });
story.fans[2].name; // 'Roger'

If you push a non-POJO and non-document value, like an ObjectId, Mongoose >= 8.7.0 will depopulate the entire array.

const fan4 = await Person.create({ name: 'Timothy' });
story.fans.push(fan4._id); // Push the `_id`, not the full document

story.fans[0].name; // undefined, `fans[0]` is now an ObjectId
story.fans[0].toString() === fan1._id.toString(); // true

Checking Whether a Field is Populated

You can call the populated() function to check whether a field is populated. If populated() returns a truthy value, you can assume the field is populated.

story.populated('author'); // truthy

story.depopulate('author'); // Make `author` not populated anymore
story.populated('author'); // undefined

A common reason for checking whether a path is populated is getting the author id. However, for your convenience, Mongoose adds a _id getter to ObjectId instances so you can use story.author._id regardless of whether author is populated.

story.populated('author'); // truthy
story.author._id; // ObjectId

story.depopulate('author'); // Make `author` not populated anymore
story.populated('author'); // undefined

story.author instanceof ObjectId; // true
story.author._id; // ObjectId, because Mongoose adds a special getter

What If There's No Foreign Document?

Mongoose populate doesn't behave like conventional SQL joins. When there's no document, story.author will be null. This is analogous to a left join in SQL.

await Person.deleteMany({ name: 'Ian Fleming' });

const story = await Story.findOne({ title: 'Casino Royale' }).populate('author');
story.author; // `null`

If you have an array of authors in your storySchema, populate() will give you an empty array instead.

const storySchema = Schema({
  authors: [{ type: Schema.Types.ObjectId, ref: 'Person' }],
  title: String
});

// Later

const story = await Story.findOne({ title: 'Casino Royale' }).populate('authors');
story.authors; // `[]`

Field Selection

What if we only want a few specific fields returned for the populated documents? This can be accomplished by passing the usual field name syntax as the second argument to the populate method:

const story = await Story.
  findOne({ title: /casino royale/i }).
  populate('author', 'name').
  exec(); // only return the Persons name
// prints "The author is Ian Fleming"
console.log('The author is %s', story.author.name);
// prints "The authors age is null"
console.log('The authors age is %s', story.author.age);

Populating Multiple Paths

What if we wanted to populate multiple paths at the same time?

await Story.
  find({ /* ... */ }).
  populate('fans').
  populate('author').
  exec();

If you call populate() multiple times with the same path, only the last one will take effect.

// The 2nd `populate()` call below overwrites the first because they
// both populate 'fans'.
await Story.
  find().
  populate({ path: 'fans', select: 'name' }).
  populate({ path: 'fans', select: 'email' });
// The above is equivalent to:
await Story.find().populate({ path: 'fans', select: 'email' });

Query conditions and other options

What if we wanted to populate our fans array based on their age and select just their names?

await Story.
  find().
  populate({
    path: 'fans',
    match: { age: { $gte: 21 } },
    // Explicitly exclude `_id`, see http://bit.ly/2aEfTdB
    select: 'name -_id'
  }).
  exec();

The match option doesn't filter out Story documents. If there are no documents that satisfy match, you'll get a Story document with an empty fans array.

For example, suppose you populate() a story's author and the author doesn't satisfy match. Then the story's author will be null.

const story = await Story.
  findOne({ title: 'Casino Royale' }).
  populate({ path: 'author', match: { name: { $ne: 'Ian Fleming' } } }).
  exec();
story.author; // `null`

In general, there is no way to make populate() filter stories based on properties of the story's author. For example, the below query won't return any results, even though author is populated.

const story = await Story.
  findOne({ 'author.name': 'Ian Fleming' }).
  populate('author').
  exec();
story; // null

If you want to filter stories by their author's name, you should use denormalization.

limit vs. perDocumentLimit

Populate does support a limit option, however, it currently does not limit on a per-document basis for backwards compatibility. For example, suppose you have 2 stories:

await Story.create([
  { title: 'Casino Royale', fans: [1, 2, 3, 4, 5, 6, 7, 8] },
  { title: 'Live and Let Die', fans: [9, 10] }
]);

If you were to populate() using the limit option, you would find that the 2nd story has 0 fans:

const stories = await Story.find().populate({
  path: 'fans',
  options: { limit: 2 }
});

stories[0].name; // 'Casino Royale'
stories[0].fans.length; // 2

// 2nd story has 0 fans!
stories[1].name; // 'Live and Let Die'
stories[1].fans.length; // 0

That's because, in order to avoid executing a separate query for each document, Mongoose instead queries for fans using numDocuments * limit as the limit. If you need the correct limit, you should use the perDocumentLimit option (new in Mongoose 5.9.0). Just keep in mind that populate() will execute a separate query for each story, which may cause populate() to be slower.

const stories = await Story.find().populate({
  path: 'fans',
  // Special option that tells Mongoose to execute a separate query
  // for each `story` to make sure we get 2 fans for each story.
  perDocumentLimit: 2
});

stories[0].name; // 'Casino Royale'
stories[0].fans.length; // 2

stories[1].name; // 'Live and Let Die'
stories[1].fans.length; // 2

Refs to children

We may find however, if we use the author object, we are unable to get a list of the stories. This is because no story objects were ever 'pushed' onto author.stories.

There are two perspectives here. First, you may want the author to know which stories are theirs. Usually, your schema should resolve one-to-many relationships by having a parent pointer in the 'many' side. But, if you have a good reason to want an array of child pointers, you can push() documents onto the array as shown below.

await story1.save();

author.stories.push(story1);
await author.save();

This allows us to perform a find and populate combo:

const person = await Person.
  findOne({ name: 'Ian Fleming' }).
  populate('stories').
  exec(); // only works if we pushed refs to children
console.log(person);

It is debatable that we really want two sets of pointers as they may get out of sync. Instead we could skip populating and directly find() the stories we are interested in.

const stories = await Story.
  find({ author: author._id }).
  exec();
console.log('The stories are an array: ', stories);

The documents returned from query population become fully functional, removeable, saveable documents unless the lean option is specified. Do not confuse them with sub docs. Take caution when calling its remove method because you'll be removing it from the database, not just the array.

Populating an existing document

If you have an existing mongoose document and want to populate some of its paths, you can use the Document#populate() method.

const person = await Person.findOne({ name: 'Ian Fleming' });

person.populated('stories'); // null

// Call the `populate()` method on a document to populate a path.
await person.populate('stories');

person.populated('stories'); // Array of ObjectIds
person.stories[0].name; // 'Casino Royale'

The Document#populate() method does not support chaining. You need to call populate() multiple times, or with an array of paths, to populate multiple paths

await person.populate(['stories', 'fans']);
person.populated('fans'); // Array of ObjectIds

Populating multiple existing documents

If we have one or many mongoose documents or even plain objects (like mapReduce output), we may populate them using the Model.populate() method. This is what Document#populate() and Query#populate() use to populate documents.

Populating across multiple levels

Say you have a user schema which keeps track of the user's friends.

const userSchema = new Schema({
  name: String,
  friends: [{ type: ObjectId, ref: 'User' }]
});

Populate lets you get a list of a user's friends, but what if you also wanted a user's friends of friends? Specify the populate option to tell mongoose to populate the friends array of all the user's friends:

await User.
  findOne({ name: 'Val' }).
  populate({
    path: 'friends',
    // Get friends of friends - populate the 'friends' array for every friend
    populate: { path: 'friends' }
  });

Cross Database Populate

Let's say you have a schema representing events, and a schema representing conversations. Each event has a corresponding conversation thread.

const db1 = mongoose.createConnection('mongodb://127.0.0.1:27000/db1');
const db2 = mongoose.createConnection('mongodb://127.0.0.1:27001/db2');

const conversationSchema = new Schema({ numMessages: Number });
const Conversation = db2.model('Conversation', conversationSchema);

const eventSchema = new Schema({
  name: String,
  conversation: {
    type: ObjectId,
    ref: Conversation // `ref` is a **Model class**, not a string
  }
});
const Event = db1.model('Event', eventSchema);

In the above example, events and conversations are stored in separate MongoDB databases. String ref will not work in this situation, because Mongoose assumes a string ref refers to a model name on the same connection. In the above example, the conversation model is registered on db2, not db1.

// Works
const events = await Event.
  find().
  populate('conversation');

This is known as a "cross-database populate," because it enables you to populate across MongoDB databases and even across MongoDB instances.

If you don't have access to the model instance when defining your eventSchema, you can also pass the model instance as an option to populate().

const events = await Event.
  find().
  // The `model` option specifies the model to use for populating.
  populate({ path: 'conversation', model: Conversation });

Dynamic References via refPath

Mongoose can also populate from multiple collections based on the value of a property in the document. Let's say you're building a schema for storing comments. A user may comment on either a blog post or a product.

const commentSchema = new Schema({
  body: { type: String, required: true },
  doc: {
    type: Schema.Types.ObjectId,
    required: true,
    // Instead of a hardcoded model name in `ref`, `refPath` means Mongoose
    // will look at the `docModel` property to find the right model.
    refPath: 'docModel'
  },
  docModel: {
    type: String,
    required: true,
    enum: ['BlogPost', 'Product']
  }
});

const Product = mongoose.model('Product', new Schema({ name: String }));
const BlogPost = mongoose.model('BlogPost', new Schema({ title: String }));
const Comment = mongoose.model('Comment', commentSchema);

The refPath option is a more sophisticated alternative to ref. If ref is a string, Mongoose will always query the same model to find the populated subdocs. With refPath, you can configure what model Mongoose uses for each document.

const book = await Product.create({ name: 'The Count of Monte Cristo' });
const post = await BlogPost.create({ title: 'Top 10 French Novels' });

const commentOnBook = await Comment.create({
  body: 'Great read',
  doc: book._id,
  docModel: 'Product'
});

const commentOnPost = await Comment.create({
  body: 'Very informative',
  doc: post._id,
  docModel: 'BlogPost'
});

// The below `populate()` works even though one comment references the
// 'Product' collection and the other references the 'BlogPost' collection.
const comments = await Comment.find().populate('doc').sort({ body: 1 });
comments[0].doc.name; // "The Count of Monte Cristo"
comments[1].doc.title; // "Top 10 French Novels"

An alternative approach is to define separate blogPost and product properties on commentSchema, and then populate() on both properties.

const commentSchema = new Schema({
  body: { type: String, required: true },
  product: {
    type: Schema.Types.ObjectId,
    required: true,
    ref: 'Product'
  },
  blogPost: {
    type: Schema.Types.ObjectId,
    required: true,
    ref: 'BlogPost'
  }
});

// ...

// The below `populate()` is equivalent to the `refPath` approach, you
// just need to make sure you `populate()` both `product` and `blogPost`.
const comments = await Comment.find().
  populate('product').
  populate('blogPost').
  sort({ body: 1 });
comments[0].product.name; // "The Count of Monte Cristo"
comments[1].blogPost.title; // "Top 10 French Novels"

Defining separate blogPost and product properties works for this simple example. But, if you decide to allow users to also comment on articles or other comments, you'll need to add more properties to your schema. You'll also need an extra populate() call for every property, unless you use mongoose-autopopulate. Using refPath means you only need 2 schema paths and one populate() call regardless of how many models your commentSchema can point to.

You could also assign a function to refPath, which means Mongoose selects a refPath depending on a value on the document being populated.

const commentSchema = new Schema({
  body: { type: String, required: true },
  commentType: {
    type: String,
    enum: ['comment', 'review']
  },
  entityId: {
    type: Schema.Types.ObjectId,
    required: true,
    refPath: function () {
      return this.commentType === 'review' ? this.reviewEntityModel : this.commentEntityModel; // 'this' refers to the document being populated
    }
  },
  commentEntityModel: {
    type: String,
    required: true,
    enum: ['BlogPost', 'Review']
  },
  reviewEntityModel: {
    type: String,
    required: true,
    enum: ['Vendor', 'Product']
  }
});

Dynamic References via ref

Just like refPath, ref can also be assigned a function.

const commentSchema = new Schema({
  body: { type: String, required: true },
  verifiedBuyer: Boolean
  doc: {
    type: Schema.Types.ObjectId,
    required: true,
    ref: function() {
      return this.verifiedBuyer ? 'Product' : 'BlogPost'; // 'this' refers to the document being populated
    }
  },
});

Populate Virtuals

So far you've only populated based on the _id field. However, that's sometimes not the right choice. For example, suppose you have 2 models: Author and BlogPost.

const AuthorSchema = new Schema({
  name: String,
  posts: [{ type: mongoose.Schema.Types.ObjectId, ref: 'BlogPost' }]
});

const BlogPostSchema = new Schema({
  title: String,
  comments: [{
    author: { type: mongoose.Schema.Types.ObjectId, ref: 'Author' },
    content: String
  }]
});

const Author = mongoose.model('Author', AuthorSchema, 'Author');
const BlogPost = mongoose.model('BlogPost', BlogPostSchema, 'BlogPost');

The above is an example of bad schema design. Why? Suppose you have an extremely prolific author that writes over 10k blog posts. That author document will be huge, over 12kb, and large documents lead to performance issues on both server and client. The Principle of Least Cardinality states that one-to-many relationships, like author to blog post, should be stored on the "many" side. In other words, blog posts should store their author, authors should not store all their posts.

const AuthorSchema = new Schema({
  name: String
});

const BlogPostSchema = new Schema({
  title: String,
  author: { type: mongoose.Schema.Types.ObjectId, ref: 'Author' },
  comments: [{
    author: { type: mongoose.Schema.Types.ObjectId, ref: 'Author' },
    content: String
  }]
});

Unfortunately, these two schemas, as written, don't support populating an author's list of blog posts. That's where virtual populate comes in. Virtual populate means calling populate() on a virtual property that has a ref option as shown below.

// Specifying a virtual with a `ref` property is how you enable virtual
// population
AuthorSchema.virtual('posts', {
  ref: 'BlogPost',
  localField: '_id',
  foreignField: 'author'
});

const Author = mongoose.model('Author', AuthorSchema, 'Author');
const BlogPost = mongoose.model('BlogPost', BlogPostSchema, 'BlogPost');

You can then populate() the author's posts as shown below.

const author = await Author.findOne().populate('posts');

author.posts[0].title; // Title of the first blog post

Keep in mind that virtuals are not included in toJSON() and toObject() output by default. If you want populate virtuals to show up when using functions like Express' res.json() function or console.log(), set the virtuals: true option on your schema's toJSON and toObject() options.

const authorSchema = new Schema({ name: String }, {
  toJSON: { virtuals: true }, // So `res.json()` and other `JSON.stringify()` functions include virtuals
  toObject: { virtuals: true } // So `console.log()` and other functions that use `toObject()` include virtuals
});

If you're using populate projections, make sure foreignField is included in the projection.

let authors = await Author.
  find({}).
  // Won't work because the foreign field `author` is not selected
  populate({ path: 'posts', select: 'title' }).
  exec();

authors = await Author.
  find({}).
  // Works, foreign field `author` is selected
  populate({ path: 'posts', select: 'title author' }).
  exec();

Populate Virtuals: The Count Option

Populate virtuals also support counting the number of documents with matching foreignField as opposed to the documents themselves. Set the count option on your virtual:

const PersonSchema = new Schema({
  name: String,
  band: String
});

const BandSchema = new Schema({
  name: String
});
BandSchema.virtual('numMembers', {
  ref: 'Person', // The model to use
  localField: 'name', // Find people where `localField`
  foreignField: 'band', // is equal to `foreignField`
  count: true // And only get the number of docs
});

// Later
const doc = await Band.findOne({ name: 'Motley Crue' }).
  populate('numMembers');
doc.numMembers; // 2

Populate Virtuals: The Match Option

Another option for Populate virtuals is match. This option adds an extra filter condition to the query Mongoose uses to populate():

// Same example as 'Populate Virtuals' section
AuthorSchema.virtual('posts', {
  ref: 'BlogPost',
  localField: '_id',
  foreignField: 'author',
  match: { archived: false } // match option with basic query selector
});

const Author = mongoose.model('Author', AuthorSchema, 'Author');
const BlogPost = mongoose.model('BlogPost', BlogPostSchema, 'BlogPost');

// After population
const author = await Author.findOne().populate('posts');

author.posts; // Array of not `archived` posts

You can also set the match option to a function. That allows configuring the match based on the document being populated. For example, suppose you only want to populate blog posts whose tags contain one of the author's favoriteTags.

AuthorSchema.virtual('posts', {
  ref: 'BlogPost',
  localField: '_id',
  foreignField: 'author',
  // Add an additional filter `{ tags: author.favoriteTags }` to the populate query
  // Mongoose calls the `match` function with the document being populated as the
  // first argument.
  match: author => ({ tags: author.favoriteTags })
});

You can overwrite the match option when calling populate() as follows.

// Overwrite the `match` option specified in `AuthorSchema.virtual()` for this
// single `populate()` call.
await Author.findOne().populate({ path: posts, match: {} });

You can also set the match option to a function in your populate() call. If you want to merge your populate() match option, rather than overwriting, use the following.

await Author.findOne().populate({
  path: posts,
  // Add `isDeleted: false` to the virtual's default `match`, so the `match`
  // option would be `{ tags: author.favoriteTags, isDeleted: false }`
  match: (author, virtual) => ({
    ...virtual.options.match(author),
    isDeleted: false
  })
});

Populating Maps

Maps are a type that represents an object with arbitrary string keys. For example, in the below schema, members is a map from strings to ObjectIds.

const BandSchema = new Schema({
  name: String,
  members: {
    type: Map,
    of: {
      type: 'ObjectId',
      ref: 'Person'
    }
  }
});
const Band = mongoose.model('Band', bandSchema);

This map has a ref, which means you can use populate() to populate all the ObjectIds in the map. Suppose you have the below band document:

const person1 = new Person({ name: 'Vince Neil' });
const person2 = new Person({ name: 'Mick Mars' });

const band = new Band({
  name: 'Motley Crue',
  members: {
    singer: person1._id,
    guitarist: person2._id
  }
});

You can populate() every element in the map by populating the special path members.$*. $* is a special syntax that tells Mongoose to look at every key in the map.

const band = await Band.findOne({ name: 'Motley Crue' }).populate('members.$*');

band.members.get('singer'); // { _id: ..., name: 'Vince Neil' }

You can also populate paths in maps of subdocuments using $*. For example, suppose you have the below librarySchema:

const librarySchema = new Schema({
  name: String,
  books: {
    type: Map,
    of: new Schema({
      title: String,
      author: {
        type: 'ObjectId',
        ref: 'Person'
      }
    })
  }
});
const Library = mongoose.model('Library', librarySchema);

You can populate() every book's author by populating books.$*.author:

const libraries = await Library.find().populate('books.$*.author');

Populate in Middleware

You can populate in either pre or post hooks. If you want to always populate a certain field, check out the mongoose-autopopulate plugin.

// Always attach `populate()` to `find()` calls
MySchema.pre('find', function() {
  this.populate('user');
});
// Always `populate()` after `find()` calls. Useful if you want to selectively populate
// based on the docs found.
MySchema.post('find', async function(docs) {
  for (const doc of docs) {
    if (doc.isPublic) {
      await doc.populate('user');
    }
  }
});
// `populate()` after saving. Useful for sending populated data back to the client in an
// update API endpoint
MySchema.post('save', function(doc, next) {
  doc.populate('user').then(function() {
    next();
  });
});

Populating Multiple Paths in Middleware

Populating multiple paths in middleware can be helpful when you always want to populate some fields. But, the implementation is just a tiny bit trickier than what you may think. Here's how you may expect it to work:

const userSchema = new Schema({
  email: String,
  password: String,
  followers: [{ type: mongoose.Schema.Types.ObjectId, ref: 'User' }],
  following: [{ type: mongoose.Schema.Types.ObjectId, ref: 'User' }]
});

userSchema.pre('find', function(next) {
  this.populate('followers following');
  next();
});

const User = mongoose.model('User', userSchema);

However, this will not work. By default, passing multiple paths to populate() in the middleware will trigger an infinite recursion, which means that it will basically trigger the same middleware for all of the paths provided to the populate() method - For example, this.populate('followers following') will trigger the same middleware for both followers and following fields and the request will just be left hanging in an infinite loop.

To avoid this, we have to add the _recursed option, so that our middleware will avoid populating recursively. The example below will make it work as expected.

userSchema.pre('find', function(next) {
  if (this.options._recursed) {
    return next();
  }
  this.populate({ path: 'followers following', options: { _recursed: true } });
  next();
});

Alternatively, you can check out the mongoose-autopopulate plugin.

Transform populated documents

You can manipulate populated documents using the transform option. If you specify a transform function, Mongoose will call this function on every populated document in the result with two arguments: the populated document, and the original id used to populate the document. This gives you more control over the result of the populate() execution. It is especially useful when you're populating multiple documents.

The original motivation for the transform option was to give the ability to leave the unpopulated _id if no document was found, instead of setting the value to null:

// With `transform`
doc = await Parent.findById(doc).populate([
  {
    path: 'child',
    // If `doc` is null, use the original id instead
    transform: (doc, id) => doc == null ? id : doc
  }
]);

doc.child; // 634d1a5744efe65ae09142f9
doc.children; // [ 634d1a67ac15090a0ca6c0ea, { _id: 634d1a4ddb804d17d95d1c7f, name: 'Luke', __v: 0 } ]

You can return any value from transform(). For example, you can use transform() to "flatten" populated documents as follows.

let doc = await Parent.create({ children: [{ name: 'Luke' }, { name: 'Leia' }] });

doc = await Parent.findById(doc).populate([{
  path: 'children',
  transform: doc => doc == null ? null : doc.name
}]);

doc.children; // ['Luke', 'Leia']

Another use case for transform() is setting $locals values on populated documents to pass parameters to getters and virtuals. For example, suppose you want to set a language code on your document for internationalization purposes as follows.

const internationalizedStringSchema = new Schema({
  en: String,
  es: String
});

const ingredientSchema = new Schema({
  // Instead of setting `name` to just a string, set `name` to a map
  // of language codes to strings.
  name: {
    type: internationalizedStringSchema,
    // When you access `name`, pull the document's locale
    get: function(value) {
      return value[this.$locals.language || 'en'];
    }
  }
});

const recipeSchema = new Schema({
  ingredients: [{ type: mongoose.ObjectId, ref: 'Ingredient' }]
});

const Ingredient = mongoose.model('Ingredient', ingredientSchema);
const Recipe = mongoose.model('Recipe', recipeSchema);

You can set the language code on all populated exercises as follows:

// Create some sample data
const { _id } = await Ingredient.create({
  name: {
    en: 'Eggs',
    es: 'Huevos'
  }
});
await Recipe.create({ ingredients: [_id] });

// Populate with setting `$locals.language` for internationalization
const language = 'es';
const recipes = await Recipe.find().populate({
  path: 'ingredients',
  transform: function(doc) {
    doc.$locals.language = language;
    return doc;
  }
});

// Gets the ingredient's name in Spanish `name.es`
recipes[0].ingredients[0].name; // 'Huevos'