Data Modeling in MongoDB

There are two ways to modeling related data in MongoDB:

via embedding
via referencing (linking)

Both approaches can be used simultaneously in the same document.

Embedded Documents

In MongoDB, by design, it is common to embed data in a parent document.

Modeling data with the embedded approach is different than what we've seen in a relational DB where we spread our data across multiple tables. However, this is the way MongoDB is designed to work and is the reason MongoDB can read and return large amounts of data far more quickly than a SQL DB that requires join operations.

To demonstrate embedding, we will add another person to our people collection, but this time we want to include contact info. A person may have several ways to contact them, so we will be modeling a typical one-to-many relationship.

Modeling Data

> db.people.insert({
    name: "Manny",
    age: 33,
    contacts: [
      {
        type: "email",
        contact: "manny@domain.com"
      },
      {
        type: "mobile",
        contact: "(555) 555-5555"
      }
    ]})

db.people.find({name:'Manny', 'contacts.type': 'email' })

What do you imagine could be a downside of embedding data?

If the embedded data's growth is unbound, MongoDB's maximum document size of 16 megabytes could be exceeded.

The above approach of embedding contact documents provides a great deal of flexibility in what types and how many contacts a person may have. However, this flexibility slightly complicates querying.

However, what if our app only wanted to work with a person's multiple emails and phoneNumbers?

Referencing Documents

We can model data relationships using a references approach where data is stored in separate documents. These documents, due to the fact that they hold different types of data, are likely be stored in separate collections.

It may help to think of this approach as linking documents together by including a reference to the related document's _id field.

Let's create a bankAccounts collection to demonstrate the references approach.

db.bankAccounts.insert({
  amount: 4403
})

This bank account might be a joint account, owned by more than one person.

For the sake of data consistency, keeping the account data in its own document would be a better design decision. In more clear terms, it would not be a good idea to store a bank account's balance in more than one place.

In our app, we have decided that all bank accounts will be retrieved through a person. This decision allows us to include a reference on the person document only.

> db.people.insert({
    name: "Miguel",
    age: 46,
    bankAccount: db.bankAccounts.findOne()._id
})

Implementing the above scenario is as simple as assigning a bankAccount document's _id to a new field in our person document:

db.people.find({name: 'Miguel', bankAccount.amount : 4403});

db.people.aggregate({
   $lookup:
     {
       from: "bankAccounts",
       localField: "bankAccount",
       foreignField: "amount",
       as: "account_amount"
     }
});

Data Modeling Best Practices - Discussion

MongoDB was designed from the ground up with application development in mind. More specifically, what can and can't be done in regards to data is enforced in your application, not the database itself (like in a SQL database).

Here are a few things to keep in mind:

For performance and simplicity reasons, lean toward embedding over referencing.
Prefer the reference approach when the child data could be unbound in amount.
Prefer the reference approach when multiple parent documents access the same child document and that child's document changes frequently.
Obtaining referenced documents requires multiple queries by your application.
In the references approach, depending upon your application's needs, you may choose to maintain links to the related document's _id in either document, or both.

For more details regarding data modeling in MongoDB, start with this section of mongoDB's documentation or this hour long YouTube video

Conclusion

What are some of the differences between Mongo & Postgres databases?
How do you add a document to a collection in the Mongo shell?
Describe the difference between embedding & referencing documents. Give an example of when you might use each.

Data Modeling