Relationships with PostgreSQL and MikroORM

May 30, 2022

A significant advantage of SQL databases is handling relationships between various tables. Since, in web applications, entities often relate to each other, designing relationships is a big part of working with SQL databases. In this article, we continue learning MikroORM and use it to form relationships.

You can find the code from this article in this repository.

One-To-One#

With a one-to-one relationship, a row from the first table has just one matching row from the second table and the other way around. An elementary example of that is creating an address entity.

address.entity.ts#

import { Entity, Property, PrimaryKey } from "@mikro-orm/core"
@Entity()
class Address {
  @PrimaryKey()
  id: number
  @Property()
  street: string
  @Property()
  city: string
  @Property()
  country: string
}
export default Address

Once we do that, we can use the @OneToOne() decorator to declare the one-to-one relationship.

user.entity.ts#

import { Entity, Property, PrimaryKey, OneToOne } from "@mikro-orm/core"
import Address from "./address.entity"
@Entity()
class User {
  @PrimaryKey()
  id: number
  @Property({ unique: true })
  email: string
  @Property()
  name: string
  @Property({ hidden: true })
  password: string
  @OneToOne({ nullable: true })
  address?: Address
}
export default User

In our application, we assign just one user to a particular address. Because of that, it is a fitting example of a one-to-one relationship.

A thing worth noting is that above, we use the serialization built into MikroORM to hide the password property using the hidden: true option. Instead, we would use the serialization built into NestJS, but unfortunately, that would not work correctly with MikroORM.

When creating the user, we don’t need to create the address explicitly. Instead, MikroORM does that for us by default.

users.service.ts#

import { Injectable } from "@nestjs/common"
import { CreateUserDto } from "./dto/createUser.dto"
import { InjectRepository } from "@mikro-orm/nestjs"
import { EntityRepository } from "@mikro-orm/core"
import User from "./user.entity"
@Injectable()
class UsersService {
  constructor(
    @InjectRepository(User)
    private readonly userRepository: EntityRepository<User>,
  ) {}
  async create(user: CreateUserDto) {
    const newUser = await this.userRepository.create(user)
    await this.userRepository.persistAndFlush(newUser)
    return newUser
  } // ...
}
export default UsersService

The migration#

Doing all of the above and running npx mikro-orm migration:create gives us the following migration:

Migration20220529210655.ts#

import { Migration } from "@mikro-orm/migrations"
 
export class Migration20220529210655 extends Migration {
  async up(): Promise<void> {
    this.addSql(
      'create table "address" ("id" serial primary key, "street" varchar(255) not null, "city" varchar(255) not null, "country" varchar(255) not null);',
    )
    this.addSql('alter table "user" add column "address_id" int null;')
    this.addSql(
      'alter table "user" add constraint "user_address_id_foreign" foreign key ("address_id") references "address" ("id") on update cascade on delete set null;',
    )
    this.addSql('alter table "user" add constraint "user_address_id_unique" unique ("address_id");')
  }
  async down(): Promise<void> {
    this.addSql('alter table "user" drop constraint "user_address_id_foreign";')
    this.addSql('drop table if exists "address" cascade;')
    this.addSql('alter table "user" drop constraint "user_address_id_unique";')
    this.addSql('alter table "user" drop column "address_id";')
  }
}

The crucial thing to notice is that MikroORM adds the address_id column to the user table and makes sure it references the address table.

MikroORM also adds a unique constraint to the address_id column. Doing that makes sure the relationship is one-to-one because only one user can refer to a particular address. Trying to relate a second user to the same address would result in an error.

We need to explicitly tell MikroORM to fetch the address of a given user if we want to do that.

users.service.ts#

import { Injectable } from "@nestjs/common"
import { UserNotFoundException } from "./exceptions/userNotFound.exception"
import { InjectRepository } from "@mikro-orm/nestjs"
import { EntityRepository } from "@mikro-orm/core"
import User from "./user.entity"
@Injectable()
class UsersService {
  constructor(
    @InjectRepository(User)
    private readonly userRepository: EntityRepository<User>,
  ) {}
  async getByEmail(email: string) {
    const user = await this.userRepository.findOne(
      {
        email,
      },
      {
        populate: ["address"],
      },
    )
    if (!user) {
      throw new UserNotFoundException()
    }
    return user
  } // ...
}
export default UsersService

Thanks to doing the above, whenever we call the getByEmail function, we get both the user and the address.

One-To-Many and Many-To-One#

We need the one-to-many relationship when we want a row from the first table to link to multiple rows in the second table. With this approach, the rows from the second table can be linked to just one row from the first table, though.

A great example is a post entity and an author. The user can be an author of multiple posts, but a particular post can only have one author. We need to use the ManyToOne() decorator to create a relationship like that.

post.entity.ts#

import { Entity, Property, PrimaryKey, ManyToOne } from "@mikro-orm/core"
import User from "../users/user.entity"
@Entity()
class PostEntity {
  @PrimaryKey()
  id: number
  @Property()
  title: string
  @Property()
  content: string
  @ManyToOne()
  author: User
}
export default PostEntity

To define an author of a post, we need to make sure that we pass it to our PostsService.

posts.controller.ts#

import { Body, Controller, Post, UseGuards, Req } from "@nestjs/common"
import { PostsService } from "./posts.service"
import CreatePostDto from "./dto/createPost.dto"
import JwtAuthenticationGuard from "../authentication/jwt-authentication.guard"
import RequestWithUser from "../authentication/requestWithUser.interface"
@Controller("posts")
export default class PostsController {
  constructor(private readonly postsService: PostsService) {}
  @Post()
  @UseGuards(JwtAuthenticationGuard)
  async createPost(@Body() post: CreatePostDto, @Req() request: RequestWithUser) {
    return this.postsService.createPost(post, request.user)
  } // ...
}

We also need to create the data for the post entity properly and add the information about the author.

posts.service.ts#

import { Injectable } from "@nestjs/common"
import { InjectRepository } from "@mikro-orm/nestjs"
import { EntityRepository } from "@mikro-orm/core"
import PostEntity from "./post.entity"
import CreatePostDto from "./dto/createPost.dto"
import User from "../users/user.entity"
@Injectable()
export class PostsService {
  constructor(
    @InjectRepository(PostEntity)
    private readonly postRepository: EntityRepository<PostEntity>,
  ) {}
  async createPost(post: CreatePostDto, user: User) {
    const postData = {
      ...post,
      author: user,
    }
    const newPost = await this.postRepository.create(postData)
    await this.postRepository.persistAndFlush(newPost)
    return newPost
  } // ...
}

Thanks to the above, whenever an authenticated user creates a post, we store the information about the user.

The migration#

Thanks to creating the above entities, MikroORM generates the following migration:

Migration20220529225632.ts#

import { Migration } from "@mikro-orm/migrations"
 
export class Migration20220529225632 extends Migration {
  async up(): Promise<void> {
    this.addSql('alter table "post_entity" add column "author_id" int not null;')
    this.addSql(
      'alter table "post_entity" add constraint "post_entity_author_id_foreign" foreign key ("author_id") references "user" ("id") on update cascade;',
    )
  }
  async down(): Promise<void> {
    this.addSql('alter table "post_entity" drop constraint "post_entity_author_id_foreign";')
    this.addSql('alter table "post_entity" drop column "author_id";')
  }
}

Like before, MikroORM added the author_id column to the post_entity table that references the user.

It is crucial to notice that the above migration does not make the author_id unique in contrast to the one-to-one relationship. Thanks to that, a particular user can be the author of multiple posts.

MikroORM, by default, does not query the details of the related entities.

We can alter this behavior by telling MikroORM explicitly to populate the additional information.

posts.service.ts#

import { Injectable } from "@nestjs/common"
import { InjectRepository } from "@mikro-orm/nestjs"
import { EntityRepository } from "@mikro-orm/core"
import PostEntity from "./post.entity"
@Injectable()
export class PostsService {
  constructor(
    @InjectRepository(PostEntity)
    private readonly postRepository: EntityRepository<PostEntity>,
  ) {}
  getPosts() {
    return this.postRepository.findAll({
      populate: ["author"],
    })
  } // ...
}

We can go even further and populate the nested properties too.

posts.service.ts#

import { Injectable } from "@nestjs/common"
import { InjectRepository } from "@mikro-orm/nestjs"
import { EntityRepository } from "@mikro-orm/core"
import PostEntity from "./post.entity"
@Injectable()
export class PostsService {
  constructor(
    @InjectRepository(PostEntity)
    private readonly postRepository: EntityRepository<PostEntity>,
  ) {}
  getPosts() {
    return this.postRepository.findAll({
      populate: ["author", "author.address"],
    })
  } // ...
}

Fortunately, the populate property is built in a type-safe way, and TypeScript will ensure we don’t make a mistake here.

Many-To-Many#

When we define a many-to-many relationship, a row from the first table can relate to multiple rows of the second table and the other way around.

A straightforward example is a post that can belong to multiple categories. On the other hand, a category can be associated with various posts.

category.entity.ts#

import { Entity, Property, PrimaryKey } from "@mikro-orm/core"
@Entity()
class Category {
  @PrimaryKey()
  public id: number
  @Property()
  public name: string
}
export default Category

post.entity.ts#

import { Entity, Property, PrimaryKey, ManyToOne, Collection, ManyToMany } from "@mikro-orm/core"
import User from "../users/user.entity"
import Category from "../categories/category.entity"
@Entity()
class PostEntity {
  @PrimaryKey()
  id: number
  @Property()
  title: string
  @Property()
  content: string
  @ManyToOne()
  author: User
  @ManyToMany(() => Category)
  categories: Collection<Category>
}
export default PostEntity

MikroORM wraps the categories property in the Collection wrapper, which is not a regular array.

The easiest way to let the user assign categories to a post is to expect the POST request to contain an array of category ids.

createPost.dto.ts#

import { IsString, IsNotEmpty, IsNumber } from "class-validator"
class CreatePostDto {
  @IsString()
  @IsNotEmpty()
  title: string
  @IsString()
  @IsNotEmpty()
  content: string
  @IsNumber({}, { each: true })
  categories: number[]
}
export default CreatePostDto

Providing the categories array when calling the this.postRepository.create method assigns categories to the created post.

To ensure that the above works properly, we can fetch information about the related categories.

posts.service.ts#

getPosts() {
  return this.postRepository.findAll({
    populate: ['author', 'author.address', 'categories'],
  });
}

With the above approach, we send quite a lot of redundant data. Instead, we might want avoid populating the relationship’s data and fetch it separately.

The migration#

By creating entities like that and generating a migration, we end up with the following file:

Migration20220530004103.ts#

import { Migration } from "@mikro-orm/migrations"
 
export class Migration20220530004103 extends Migration {
  async up(): Promise<void> {
    this.addSql('create table "category" ("id" serial primary key, "name" varchar(255) not null);')
    this.addSql(
      'create table "post_entity_categories" ("post_entity_id" int not null, "category_id" int not null);',
    )
    this.addSql(
      'alter table "post_entity_categories" add constraint "post_entity_categories_pkey" primary key ("post_entity_id", "category_id");',
    )
    this.addSql(
      'alter table "post_entity_categories" add constraint "post_entity_categories_post_entity_id_foreign" foreign key ("post_entity_id") references "post_entity" ("id") on update cascade on delete cascade;',
    )
    this.addSql(
      'alter table "post_entity_categories" add constraint "post_entity_categories_category_id_foreign" foreign key ("category_id") references "category" ("id") on update cascade on delete cascade;',
    )
  }
  async down(): Promise<void> {
    this.addSql(
      'alter table "post_entity_categories" drop constraint "post_entity_categories_category_id_foreign";',
    )
    this.addSql('drop table if exists "category" cascade;')
    this.addSql('drop table if exists "post_entity_categories" cascade;')
  }
}

When we defined a many-to-many relationship between posts and categories, MikroORM created a post_entity_categories table. To store the information about a many-to-many relation, MikroORM uses the post_entity_id and category_id columns.

Bidirectional relationships#

So far, we’ve been creating unidirectional relationships. It means that we’ve defined them only on one side of the relationship.

For example, we’ve used the @ManyToOne() decorator to assign an author to a post.

post.entity.ts#

import { Entity, PrimaryKey, ManyToOne } from "@mikro-orm/core"
import User from "../users/user.entity"
@Entity()
class PostEntity {
  @PrimaryKey()
  id: number
  @ManyToOne()
  author: User // ...
}
export default PostEntity

Doing the above and creating a migration caused our post_entity table to have the author_id column.

Because the post entity contains the author_id, we call it the owning side of the relationship.

We could make the above relationship bidirectional by defining it on the user’s side, using the @OneToMany() decorator.

user.entity.ts#

import { Entity, PrimaryKey, OneToMany, Collection } from "@mikro-orm/core"
import PostEntity from "../posts/post.entity"
@Entity()
class User {
  @PrimaryKey()
  id: number // ...
  @OneToMany(() => PostEntity, (post: PostEntity) => post.author)
  posts: Collection<PostEntity>
}
export default User

The crucial thing about defining a bidirectional relationship is that it does not modify the structure of the database. For example, creating a bidirectional relationship does not cause the user table’s rows to contain the posts’ ids.

Thanks to having the posts property in the user’s entity, we can easily populate it.

users.service.ts#

import { Injectable } from "@nestjs/common"
import { UserNotFoundException } from "./exceptions/userNotFound.exception"
import { InjectRepository } from "@mikro-orm/nestjs"
import { EntityRepository } from "@mikro-orm/core"
import User from "./user.entity"
@Injectable()
class UsersService {
  constructor(
    @InjectRepository(User)
    private readonly userRepository: EntityRepository<User>,
  ) {}
  async getByEmail(email: string) {
    const user = await this.userRepository.findOne(
      {
        email,
      },
      {
        populate: ["address", "posts"],
      },
    )
    if (!user) {
      throw new UserNotFoundException()
    }
    return user
  } // ...
}
export default UsersService

We can also create bidirectional relationships using One-To-One and Many-To-Many.

The potential issues with bidirectional relationships#

While the above solution gives us an effortless way to fetch additional data, it comes with a cost. Even though it looks straightforward, it contains quite a bit of implicit complexity.

When we fetch the user with a list of posts, MikroORM has to go through all posts in the posts table to find matching entities. So it might hurt our performance at some point if we use this feature carelessly.

Also, in our case, to achieve the bidirectional relationship, the users.entity.ts file imports the post.entity.ts file and vice versa. The above creates a circular dependency, which is usually frowned upon.

Summary#

We’ve gone through all the relationship types in this article and implemented them with MikroORM and PostgreSQL. Besides regular relationships, we’ve also defined a bidirectional relationship. As a result, we’ve learned quite a bit about how MikroORM works and how it affects our database structure. Thanks to that, we can better grasp how our application communicates with the database and how it can affect its performance.

One-To-One#

address.entity.ts#

user.entity.ts#

users.service.ts#

The migration#

Migration20220529210655.ts#

Fetching the related entities#

users.service.ts#

One-To-Many and Many-To-One#

post.entity.ts#

posts.controller.ts#

posts.service.ts#

The migration#

Migration20220529225632.ts#

Fetching the related entities#

posts.service.ts#

posts.service.ts#

Many-To-Many#

category.entity.ts#

post.entity.ts#

createPost.dto.ts#

Fetching the related entities#

posts.service.ts#

The migration#

Migration20220530004103.ts#

Bidirectional relationships#

post.entity.ts#

user.entity.ts#

users.service.ts#

The potential issues with bidirectional relationships#

Summary#