Model factories

Model factories

Have you ever written tests, in which the first 15-20 lines of each test are dedicated to just setting up the database state by using multiple models? With model factories, you can extract all this set up to a dedicated file and then write the bare minimum code to set up the database state.

By the end of this guide, you will know:

  • How to create and use factories
  • How to define factory states
  • Working with model relationships
  • Using the faker API to generate and use random data

Creating factories

Model factories are stored inside databases/factories directory. You can define all factories within a single file or create dedicated files for each model, the choice is yours.

You can use the make:factory command to create a new factory. The command accepts the model name for which you want to create the factory.

Unlike seeders or models, the factories are declarative in nature as shown in the following example:

database/factories/user.ts
import User from '#models/user'
import Factory from '@adonisjs/lucid/factories'
export const UserFactory = Factory.define(User, ({ faker }) => {
return {
username: faker.internet.userName(),
email: faker.internet.email(),
password: faker.internet.password(),
}
}).build()
  • The Factory.define method accepts a total of two arguments.
  • The first argument is reference to the Lucid model.
  • The second argument is a callback that returns an object of properties to be used when persisting the model instance. Make sure that you return an object with all the required properties, otherwise the database will raise not null exceptions. The faker option passed into the callback is an instance of Faker.js
  • Finally, make sure to call the build method.

Using factories

Using factories is quite simple. Just import the file and use the exported factories.

import { UserFactory } from '#database/factories/user'
const user = await UserFactory.create()

In order to create multiple instances, you can make use of the createMany method.

const users = await UserFactory.createMany(10)

Merging attributes

You can override the default set of attributes using the .merge method. For example:

await UserFactory.merge({ email: 'test@example.com' }).create()

When creating multiple instances, you can define an array of attributes and they will merge based upon their indices. For example:

await UserFactory.merge([{ email: 'foo@example.com' }, { email: 'bar@example.com' }]).createMany(3)

In the above example

  • The first user will have the email of foo@example.com.
  • The second user will have the email of bar@example.com.
  • And, the third user will use the default email address, since the merge array has a length of 2.

Factory states

Factory states allow you to define variations of your factories as states. For example: On a Post factory, you can have different states to represent published and draft posts.

import Post from '#models/post'
import { Factory } from '@adonisjs/lucid/factories'
export const PostFactory = Factory.define(Post, ({ faker }) => {
return {
title: faker.lorem.sentence(),
content: faker.lorem.paragraphs(4),
status: 'DRAFT',
}
})
.state('published', (post) => (post.status = 'PUBLISHED')) // 👈
.build()

By default, all posts will be created with DRAFT status. However, you can explicitly apply the published state to create posts with PUBLISHED status.

await PostFactory.apply('published').createMany(3)
await PostFactory.createMany(3)

Relationships

Model factories makes it super simple to work with relationships. Consider the following example:

export const PostFactory = Factory.define(Post, ({ faker }) => {
return {
title: faker.lorem.sentence(),
content: faker.lorem.paragraphs(4),
status: 'DRAFT',
}
}).build()
export const UserFactory = Factory.define(User, ({ faker }) => {
return {
username: faker.internet.userName(),
email: faker.internet.email(),
password: faker.internet.password(),
}
})
.relation('posts', () => PostFactory) // 👈
.build()

Now, you can create a user and its posts all together in one call.

const user = await UserFactory.with('posts', 3).create()
user.posts.length // 3
  • The factory will find the type of relationship by inspecting the Lucid model. For example: If your model defines a hasMany relationship on posts, then factory will infer the same.

  • A relationship first needs to be defined on the model and then only it can be defined on the Factory.

  • Lucid will internally wrap all the database operations inside a transaction. So if a relationship persistence fails, the parent model persistence will be rolled back too.

Applying relationship states

You can also apply states on a relationship by passing a callback to the with method.

const user = await UserFactory.with('posts', 3, (post) => post.apply('published')).create()

Similarly, if you want, you can create few posts with the published state and few without it.

const user = await UserFactory.with('posts', 3, (post) => post.apply('published'))
.with('posts', 2)
.create()
user.posts.length // 5

Finally, you can also create nested relationships. For example: Create a user with two posts and five comments for each post.

const user = await UserFactory.with('posts', 2, (post) => post.with('comments', 5)).create()

Pivot attributes

When creating a many to many relationship, you can define the attributes for the pivot table using the pivotAttributes method.

In the following example, the User model has a many to many relationship with the Team model and we define the user role within a given team.

await UserFactory.with('teams', 1, (team) => {
team.pivotAttributes({ role: 'admin' })
}).create()

You can pass an array of objects to the pivotAttributes method when creating multiple instances of the relationship.

The size of array should match the count of relationship rows you are about to create.

await UserFactory.with('teams', 2, (team) => {
team.pivotAttributes([{ role: 'admin' }, { role: 'moderator' }])
}).create()

Stubbing database calls

In some cases, you may prefer to stub out the database calls and just want to create in-memory model instances. This is can achieved using the makeStubbed and makeStubbedMany methods.

const user = await UserFactory.with('posts', 2).makeStubbed()
console.log(user.id) // <some-id>
console.log(user.$isPersisted) // false

The stubbed calls will never hit the database and will assign an in-memory numeric id to the model instances.

Customizing stub id

When we say id. We mean the primary key of a model and not a fixed named attribute id.

The stub id is just an in-memory counter, that keeps on increasing with every call. If required, you can define a custom method to generate stub ids in a different manner.

For example: Generating ids as a BigInt when using PostgreSQL bigInteger data type.

import { Factory } from '@adonisjs/lucid/factories'
Factory.stubId((counter, model) => {
return BigInt(counter)
})

You can make use of the makeStubbed hook to customize the id generation behavior for an individual factory.

import { randomUUID } from 'node:crypto'
Factory.define(Post, () => {
return {}
}).before('makeStubbed', (_, model) => {
model.id = randomUUID()
})

Runtime context

Every time you create a model instance from a factory, a runtime context is also created at the same time. The context is then passed to all the hooks, the define method callback and also the relationships.

Most of the time, you just want to access the faker object from the context. However, following are the available properties.

  • isStubbed: A boolean to know, if the factory was instantiated in stub mode.
  • $trx: A transaction object, under which all the database operations are wrapped. If you are running any database queries inside the factory hooks, then make sure to also wrap them inside the transaction.

Following is an example showcasing the callbacks that receives the runtime context (ctx).

Factory.define(User, (ctx) => {})
.before('create', (factory, model, ctx) => {})
.after('create', (factory, model, ctx) => {})
.state('admin', (model, ctx) => {})
.build()

Hooks

The factory exposes the following hooks to perform actions before or after certain events. You can also define multiple hooks for a single event.

Factory.define(Post, () => {})
.before('create', () => {})
.after('create', () => {})
LifecycleEventDescription
beforecreateInvoked before the insert query.
aftercreateInvoked after the insert query.
beforemakeStubbedInvoked before the stubbed call.
aftermakeStubbedInvoked after the stubbed call.
aftermakeInvoked only after the model instance has been created. This hook is also invoked before the before create and before makeStubbed hooks.

Custom connections

Factories allows you to define a custom connection or query client at the time using them. For example:

await Factory.connection('tenant-1').create()

Also, you can pass a custom query client instance.

const queryClient = Database.connection('tenant-1')
await Factory.client(queryClient).create()

For the sake of API uniformity among the factories and the Lucid models, you can also define the connection or the client using the query method.

await Factory.query({ connection: 'tenant-1' }).create()

Customizations

Finally, you can optionally customize the behavior of certain operations performed under the hood.

newUp

By defining the newUp handler, you can customize the process of instantiating a model instance for a given factory.

Factory.define(User, () => {})
.newUp((attributes, ctx) => {
const user = new User()
user.fill(attributes)
return user
})
.build()

merge

By defining the merge handler, you can customize the merge behavior.

Factory.define(User, () => {})
.merge((user, attributes, ctx) => {
user.merge(attributes)
})
.build()