3.5.3. Data Model Relationships
In this chapter, you’ll learn how to define relationships between data models in your module.
What is a Relationship Property?#
A relationship property defines an association in the database between two models. It's created using the Data Model Language (DML) methods, such as hasOne
or belongsTo
.
When you generate a migration for these data models, the migrations include foreign key columns or pivot tables, based on the relationship's type.
One-to-One Relationship#
A one-to-one relationship indicates that one record of a data model belongs to or is associated with another.
To define a one-to-one relationship, create relationship properties in the data models using the following methods:
hasOne
: indicates that the model has one record of the specified model.belongsTo
: indicates that the model belongs to one record of the specified model.
For example:
1import { model } from "@medusajs/framework/utils"2 3const User = model.define("user", {4 id: model.id().primaryKey(),5 email: model.hasOne(() => Email, {6 mappedBy: "user",7 }),8})9 10const Email = model.define("email", {11 id: model.id().primaryKey(),12 user: model.belongsTo(() => User, {13 mappedBy: "email",14 }),15})
In the example above, a user has one email, and an email belongs to one user.
The hasOne
and belongsTo
methods accept a function as the first parameter. The function returns the associated data model.
Both methods also accept a second parameter object with the property mappedBy
. Its value is the name of the relationship property in the other data model.
Optional Relationship#
To make the relationship optional on the hasOne
or belongsTo
side, use the nullable
method on either property as explained in this chapter.
One-to-One Relationship in the Database#
When you generate the migrations of data models that have a one-to-one relationship, the migration adds to the table of the data model that has the belongsTo
property:
- A column of the format
{relation_name}_id
to store the ID of the record of the related data model. For example, theemail
table will have auser_id
column. - A foreign key on the
{relation_name}_id
column to the table of the related data model.
One-sided One-to-One Relationship#
In some use cases, you may want to define a one-to-one relationship only on one side. This means that the other data model does not have a relationship property pointing to the first one.
You can do this either from the hasOne
or the belongsTo
side.
hasOne Side
By default, the foreign key column is added to the table of the data model that has the belongsTo
property. For example, if the Email
data model belongs to the User
data model, then the foreign key column is added to the email
table.
If you want to define a one-to-one relationship only on the User
data model's side (hasOne
side), you can do so by passing the following properties to the second parameter of the hasOne
method:
foreignKey
: A boolean indicating whether the foreign key column should be added to the table of the data model.mappedBy
: Set toundefined
, since the relationship is only defined on one side.
For example:
In the example above, you add a one-to-one relationship from the User
data model to the Email
data model.
The foreign key column is added to the user
table, and the Email
data model does not have a relationship property pointing to the User
data model.
belongsTo Side
To define the one-to-one relationship on the belongsTo
side, pass undefined
to the mappedBy
property in the belongsTo
method's second parameter.
For example:
In the example above, you add a one-to-one relationship from the Email
data model to the User
data model.
The User
data model does not have a relationship property pointing to the Email
data model.
One-to-Many Relationship#
A one-to-many relationship indicates that one record of a data model has many records of another data model.
To define a one-to-many relationship, create relationship properties in the data models using the following methods:
hasMany
: indicates that the model has more than one record of the specified model.belongsTo
: indicates that the model belongs to one record of the specified model.
For example:
1import { model } from "@medusajs/framework/utils"2 3const Store = model.define("store", {4 id: model.id().primaryKey(),5 products: model.hasMany(() => Product, {6 mappedBy: "store",7 }),8})9 10const Product = model.define("product", {11 id: model.id().primaryKey(),12 store: model.belongsTo(() => Store, {13 mappedBy: "products",14 }),15})
In this example, a store has many products, but a product belongs to one store.
Optional Relationship#
To make the relationship optional on the belongsTo
side, use the nullable
method on the property as explained in this chapter.
One-to-Many Relationship in the Database#
When you generate the migrations of data models that have a one-to-many relationship, the migration adds to the table of the data model that has the belongsTo
property:
- A column of the format
{relation_name}_id
to store the ID of the record of the related data model. For example, theproduct
table will have astore_id
column. - A foreign key on the
{relation_name}_id
column to the table of the related data model.
Many-to-Many Relationship#
A many-to-many relationship indicates that many records of a data model can be associated with many records of another data model.
To define a many-to-many relationship, create relationship properties in the data models using the manyToMany
method.
For example:
1import { model } from "@medusajs/framework/utils"2 3const Order = model.define("order", {4 id: model.id().primaryKey(),5 products: model.manyToMany(() => Product, {6 mappedBy: "orders",7 pivotTable: "order_product",8 joinColumn: "order_id",9 inverseJoinColumn: "product_id",10 }),11})12 13const Product = model.define("product", {14 id: model.id().primaryKey(),15 orders: model.manyToMany(() => Order, {16 mappedBy: "products",17 }),18})
The manyToMany
method accepts two parameters:
- A function that returns the associated data model.
- An object of optional configuration. Only one of the data models in the relation can define the
pivotTable
,joinColumn
, andinverseJoinColumn
configurations, and it's considered the owner data model. The object can accept the following properties:mappedBy
: The name of the relationship property in the other data model. If not set, the property's name is inferred from the associated data model's name.pivotTable
: The name of the pivot table created in the database for the many-to-many relation. If not set, the pivot table is inferred by combining the names of the data models' tables in alphabetical order, separating them by_
, and pluralizing the last name. For example,order_products
.joinColumn
: The name of the column in the pivot table that points to the owner model's primary key.inverseJoinColumn
: The name of the column in the pivot table that points to the owned model's primary key.
pivotTable
, joinColumn
, and inverseJoinColumn
properties are only available after Medusa v2.0.7.pivotTable
, joinColumn
, and inverseJoinColumn
aren't specified on either model, the owner is decided based on alphabetical order. So, in the example above, the Order
data model would be the owner.In this example, an order is associated with many products, and a product is associated with many orders. Since the pivotTable
, joinColumn
, and inverseJoinColumn
configurations are defined on the order, it's considered the owner data model.
Many-to-Many Relationship in the Database#
When you generate the migrations of data models that have a many-to-many relationship, the migration adds a new pivot table. Its name is either the name you specify in the pivotTable
configuration or the inferred name combining the names of the data models' tables in alphabetical order, separating them by _
, and pluralizing the last name. For example, order_products
.
The pivot table has a column with the name {data_model}_id
for each of the data model's tables. It also has foreign keys on each of these columns to their respective tables.
The pivot table has columns with foreign keys pointing to the primary key of the associated tables. The column's name is either:
- The value of the
joinColumn
configuration for the owner table, and theinverseJoinColumn
configuration for the owned table; - Or the inferred name
{table_name}_id
.
Many-To-Many with Custom Columns#
To add custom columns to the pivot table between two data models having a many-to-many relationship, you must define a new data model that represents the pivot table.
For example:
1import { model } from "@medusajs/framework/utils"2 3export const Order = model.define("order_test", {4 id: model.id().primaryKey(),5 products: model.manyToMany(() => Product, {6 pivotEntity: () => OrderProduct,7 }),8})9 10export const Product = model.define("product_test", {11 id: model.id().primaryKey(),12 orders: model.manyToMany(() => Order),13})14 15export const OrderProduct = model.define("orders_products", {16 id: model.id().primaryKey(),17 order: model.belongsTo(() => Order, {18 mappedBy: "products",19 }),20 product: model.belongsTo(() => Product, {21 mappedBy: "orders",22 }),23 metadata: model.json().nullable(),24})
The Order
and Product
data models have a many-to-many relationship. To add extra columns to the created pivot table, you pass a pivotEntity
option to the products
relation in Order
(since Order
is the owner). The value of pivotEntity
is a function that returns the data model representing the pivot table.
The OrderProduct
model defines, aside from the ID, the following properties:
order
: A relation that indicates this model belongs to theOrder
data model. You set themappedBy
option to the many-to-many relation's name in theOrder
data model.product
: A relation that indicates this model belongs to theProduct
data model. You set themappedBy
option to the many-to-many relation's name in theProduct
data model.metadata
: An extra column to add to the pivot table of typejson
. You can add other columns as well to the model.
Cascades#
When an operation is performed on a data model, such as record deletion, the relationship cascade specifies what related data model records should be affected by it.
For example, if a store is deleted, its products should also be deleted.
The cascades
method used on a data model configures which child records an operation is cascaded to.
For example:
1import { model } from "@medusajs/framework/utils"2 3const Store = model.define("store", {4 id: model.id().primaryKey(),5 products: model.hasMany(() => Product),6})7.cascades({8 delete: ["products"],9})10 11const Product = model.define("product", {12 id: model.id().primaryKey(),13 store: model.belongsTo(() => Store, {14 mappedBy: "products",15 }),16})
The cascades
method accepts an object. Its key is the operation’s name, such as delete
. The value is an array of relationship property names that the operation is cascaded to.
In the example above, when a store is deleted, its associated products are also deleted.