Agile Data Management with MongoDB

Hey visitor, welcome to the future! Do you come from the dark ages of the last decade where giant relational databases ruled? Don’t worry, it’s year 2023 now and we have moved on: we have something better for you. Check out document databases!

The best document databases that we have here in the future allow you to insert different kinds of complex objects into your tables collections without the headache of casting fixed schemas on your tables, and still have all the benefits of an SQL database - indexing, sophisticated queries, joins and still supporting the timeless relational model. Isn’t it fantastic to live in the future?

Let’s dive deeper!

But let me ask first,

Is the relational model going away?

Now, this is deep. Are the relational databases going away? Are they some ancient artifacts of the past? I don’t think so. Let me explain why! A database that supports the relational database model is a database that supports relational concepts like one-to-one, one-to-many and many-to-many.

Let’s think about it. Isn’t it a really natural way to see the world? Actually, can you think about the world without relations at all?

For example, think about a school. A school can have many pupils. Typically, a pupil belongs to one school. That’s a typical one-to-many relation. A course can be attended by many pupils. Yet one pupil may be attending many courses. This is a many-to-many relation.

Relational databases organize object similarly into object types that are stored in their own groups, or “tables”. These databases offer a simple and powerful concept to organize the relations between these objects in a way that is very natural, mimicking the way we see the world as well. Therefore, the relational databases will hardly go away at all!

In a way, MongoDB is a relational database

It’s quite common to think that a relational database equals an SQL database. SQL is a query language used to make database queries to relational databases. I challenge this view because MongoDB, not being an SQL database since it doesn’t support SQL queries, makes it possible to combine data in a way very similar to typical relational databases. Therefore, I would still call MongoDB a relational database - since it makes possible to form relations between objects similarly to SQL-based relational databases. If you wish!

Getting hands dirty

That’s all about philosophy for now. I’ll show next how to make the most of the relational databases with MongoDB, while still enjoying the fruit of flexibility of a document based database. Let’s get started and implement…

One-to-many and many-to-many models with MongoDB!

Another practical example of a many-to-many scenario in the real world would be a school that can have many courses and pupils. Yet each pupil can attend many courses and each course will be attended by many pupils.

To start with, let’s describe the scenario first with typescript models. This makes sense, because we will be soon loading data from MongoDB into those typescript objects:

interface ISchool {
  _id: string;
  name: string;
  courses?: ICourse[];
  pupils?: IPupil[]
}

interface IPupil {
  _id: string;
  schoolId: string;
  name: string;
  school?: ISchool;
  courses?: ICourse[]
}

interface ICourse {
  _id: string;
  schoolId: string; 
  name: string;
  school?: ISchool;
  pupils?: ICourse[];
}

Note that I added a prefix “I” to each model here! Why did I do it? I am just using here a convention that separates data objects from their actual handlers. If I don’t mix methods into the data objects but keep the data objects separate and plain, it will be a lot easier to serialize them directly into JSON objects. And JSON objects can be quite handily transferred over REST (or other) APIs over the Internet. Not to forget how this approach can boost your unit tests when you have a clear separation between data objects and their handlers.

To put it short, I added the “I” prefix to these data objects so I can have separate handler objects such as School, Pupil and Course.

Next time, let’s check out how to build and keep relationships between these objects on database level when working with MongoDB.

Finding schools, courses and students

Now, let’s find some schools, pupils and courses. I implement a handler called “School” that contains the functions to do that. typescript is a fantastic programming language because it supports nicely both a more “functional” like approach, yet also providing structures that allow a more traditional “object-oriented” approach, familiar to those coming from planet Java. I’ll play a bit here and mix in something from both worlds.

You could use class instances, static classes or even namespaces to implement the handlers to persist our objects into MongoDB. I will use namespaces here this time.

Have a look at my School handler that finds a school by ID and merges school’s pupils and courses into the result with a lookup:

namespace School {

  export const findByIdWithCoursesAndPupils = async (schoolId: string): 
	Promise<ISchool[]> => await MongoClient.aggregate("schools", [
	  {
		$match: {
		  _id: schoolId
		}
	  },
	  {
		$lookup: {
		  from: "courses,
		  localField: "_id",
		  foreignField: "schoolId",
		  as: "courses"
		}
	  },
	  {
		$lookup: {
		  from: "pupils",
		  localField: "_id",
		  foreignField: "schoolId",
	 	  as: "pupils"
	    }
	  }
  ]);
}

Populating collections with data

For our example, let’s add two imaginary schools, and populate the first one with courses and pupils. Finally, we add them all into MongoDB with handler utilities that we just implemented:

  // Defining the first school:
  const schoolMalmo: ISchool = {
	_id: "malmo",
	name: "Malmö Grundskula Svärje"
  };
 
  // Defining another school:
  const schoolHermosa: ISchool = {
	_id: "hermosa",
	name: "Escuela Hermosa Colombiana"
  };

  // Placing the chools into an array:
  const schools: ISchool[] = [schoolMalmo, schoolHermosa];

  // Adding a pupil to the first school:
  const pupilOne: IPupil = {
	_id: "elev1",
	schoolId: schoolMalmo._id,
	name: "Elva Tolva"
  };

  // Adding another pupil to the first school:
  const pupilTwo: IPupil = {
	_id: "elev2",
	schoolId: schoolMalmo._id,
	name: "Åckso Någonannan"
  };

  // Assinging the pupils into an array: 
  const pupils: IPupil[] = [pupilOne, pupilTwo];

  // Adding two courses to the first school:
  const courses: ICourse[] = [
	{
	  _id: "engelska",
	  schoolId: schoolMalmo._id,
	  name: "läser vi engelska mykke vidare"
	},
	{
	  _id: "matematiken",
	  schoolId: schoolMalmo._id,
	  name: "matematiken för alla"
	}
  ];

Next, let’s implement DB injectors for all our entities that we created:

namespace School {
 
  const collectionName: string = "schools";

  export const insertMany = (schools: ISchool[]) => MongoClient.insertMany(collectionName, schools);

}

One-to-many with MongoDB

And write similar type-checking insert methods for other classes. If you are lazy you can also just use the generic method MongoClient.insertMany(collectionName, entries) directly but then you lose type-checking. With TypeScrpt, type-safe with boilerplate or relax-style with compact code, the choice is yours!

Alright, I cut the crap now and add the items. Just do it!

await School.insertMany(schools);
await Course.insertMany(courses);
await Pupil.insertMany(pupils);

And finally, get the results:

await School.findByIdWithCoursesAndPupils("malmo");

The database returns our school with all its courses and pupils just the way we wanted:

{
    "_id": "malmo",
    "name": "Malmö Grundskula Svärje",
    "courses": [{
        "_id": "engelska",
        "schoolId": "malmo",
        "name": "läser vi engelska mykke vidare"
    }, {
        "_id": "matematiken",
        "schoolId": "malmo",
        "name": "matematiken för alla"
    }],
    "pupils": [{
        "_id": "elev1",
        "schoolId": "malmo",
        "name": "Elva Tolva"
    }, {
        "_id": "elev2",
        "schoolId": "malmo",
        "name": "Åckso Någonannan"
    }]
}

Many-to-many with MongoDB

In our example, a course should be able to have many pupils and one pupil should be attending many courses. It can be done by adding an association object, a design pattern well known in relational databases:

interface IPupilCourse {
  _id?: string;
  courseId: string;
  pupilId: string;
  courses?: ICourse[];
  pupils?: IPupil[];
}

Now, adding a pupil to a course:

const pupilCourse: IPupilCourse = {
  _id: `${courseMatematik._id}:${pupilOne._id}`,
  courseId: courseMatematik._id,
  pupilId: pupilOne._id,
}

await PupilCourse.insertMany([pupilCourse]);

And finally, let’s implement a query to find all pupils by course ID:

export namespace PupilCourse {

  const collectionName: string = "pupils-courses";

  export const insertMany = (pupilCourses: IPupilCourse[]) => MongoClient.insertMany(collectionName, pupilCourses);

  export const findPupilsByCourseId =  async (courseId: string) => await MongoClient.aggregate(collectionName, [
    {
      $match: {
        courseId
      },
    },
    {
      $lookup: {
        from: "pupils",
        localField: "pupilId",
        foreignField: "_id",
        as: "pupils"
      }
    }
  ]);

}

The ease of a schemaless document database peppered with agile queries comparable to SQL databases - that’s MongoDB!

I’ll be back with more examples from the exciting world of managing data with MongoDB.

Cheers!

See you soon!