Data querying library for testing JavaScript applications.

This library exists to help developers model and query data when testing and developing their applications. It acts as a convenient way of creating schema-based fixtures and querying them with a familiar ORM-inspired syntax. It can be used standalone or in conjuncture with Mock Service Worker for seamless mocking experience both on the network and the data layers.

• Relies on Standard Schema instead of inventing a proprietary modeling syntax you have to learn. You can use any Standard Schema-compliant object modeling library to describe your data, like Zod, ArkType, Valibot, yup, and many others.
• Full runtime and type-safety.
• Provides a powerful querying syntax inspired by Prizma;
• Supports relations for database-like behaviors (inspired by Drizzle);
• Supports extensions (including custom extensions) for things like cross-tab collection synchronization or record persistence.

npm i @msw/data

You start by defining a data collection.

import { Collection } from '@msw/data'
import z from 'zod'

const users = new Collection({
  schema: z.object({
    id: z.number(),
    name: z.string(),
  }),
})

Above, I'm using Zod to describe the schema for my users collection. You can use whichever Standard Schema compliant library of your choice instead.

Next, let's put some values into our collection. Those values are called records and you can create individual records via the .create() method or create a bunch of them with .createMany().

await users.create({ id: 1, name: 'John' })

await users.createMany(5, (index) => ({
  id: index + 1,
  name: faker.person.firstName(),
}))

Combine .createMany() with tools like Faker for random values in your records.

From this point on, you can use your users collection for anything data-related. You can create more records, query them, define relations to other collections, update and delete records, etc. Learn more you can do with the library in the documentation below. Good luck!

Whenever you have to target a record(s), you construct a query. A query acts as a predicate that a record must match in order to be targeted. The most basic query is that describing expected values of the record's properties:

users.findFirst((q) => q.where({ name: 'John' }))

Above, we are defining a query using the q builder that targets the first user whose name property equals to 'John'.

Additionally, any property in a query can be expanded into a function that accepts the value and returns a boolean, indicating whether the record matches:

users.findFirst((q) =>
  q.where({
    name: (name) => name.startsWith('John'),
  }),
)

This query matches the first user whose name starts with 'John'. Use functions as predicates to express more advanced logic in your queries.

Query functions are supported at any level of nesting, including the top-level record itself.

users.findFirst((q) => q.where((user) => user.posts.length > 0))

users.findFirst((q) =>
  q.where({
    address: {
      street: (street) => street !== 'Baker st.',
    },
  }),
)

You can build complex queries via .or() and .and() logical operators exposed through the query builder. For example, here's a query that matches all users who have posts or are editors:

users.findMany((q) =>
  q.where({ posts: (posts) => posts.length > 0 }).or({ role: 'editor' }),
)

If you prefer functional composition over method chaining, you can wrap predicates in q.or() and q.and() instead. Both syntaxes result in the same query.

users.findMany((q) =>
  q.or(
    q.where({ posts: (posts) => posts.length > 0 }),
    q.where({ role: 'editor' }),
  ),
)

This library supports offset and cursor-based pagination.

Provide the take property onto the options object of any bulk operation method, like .findMany(), .updateMany(), or .deleteMany(), to limit the number of results returned by the query.

const users = new Collection({ schema })

users.findMany(
  (q) => q.where({ email: (email) => email.includes('@google.com') }),
  {
    // Return the first 5 matching records.
    take: 5,
  },
)

You can also skip the number of first matching results by providing the skip property:

const users = new Collection({ schema })

users.findMany(
  (q) => q.where({ email: (email) => email.includes('@google.com') }),
  {
    // Skip the first 10 matching records.
    skip: 10,
    // And return the next 5.
    take: 5,
  },
)

The negative value for take is also supported to have backward pagination:

users.findMany(
  (q) => q.where({ email: (email) => email.includes('@google.com') }),
  {
    take: -5,
  },
)

Provide a reference to the record of the same collection as the cursor property for cursor-based pagination.

const users = new Collection({ schema })

const john = users.findFirst((q) => q.where({ name: 'John' }))

users.findMany((q) => q.where({ subscribed: true }), {
  cursor: john,
  take: 5,
})

You can sort the results of bulk operations, like .findMany(), .updateMany(), and .deleteMany(), by providing the orderBy property in that operation's options.

// Find all users whose name starts with "J"
// and return them sorted by their `name`.
users.findMany((q) => q.where({ name: (name) => name.startsWith('J') }), {
  orderBy: { name: 'asc' },
})

You can sort by multiple criteria by providing them in the orderBy object:

users.updateMany((q) => q.where({ name: (name) => name.startsWith('J') }), {
  data(user) {
    user.name = user.name.toUpperCase(),
  },
  orderBy: {
    name: 'asc',
    id: 'desc',
  },
})

You can define relations by calling the .defineRelations() method on the collection.

• One-to-one
• One-to-many
• One-to-many (inversed)
• Many-to-many
• Through relations
• Unique relations
• Ambiguous relations
• Polymorphic relations

Below, you can find examples of defining various types of relations, but there are a few things that apply to all of them:

• Describe relations on the schema level using your schema library. The .defineRelations() API has no effect on the model's schema/types and only operates on known properties;
• Relations are described after a collection is defined (to prevent circular references);
• Relations do not require explicit foreignKey associations and instead are bound to internal IDs of related records.

const userSchema = z.object({
  // In Zod, relational properties are best described as getters
  // so they can produce self-referencing schemas.
  get country() {
    return countrySchema
  },
})
const countrySchema = z.object({ code: z.string() })

const users = new Collection({ schema: userSchema })
const countries = new Collection({ schema: countrySchema })

// Declare the relations on the `users` collection.
users.defineRelations(({ one }) => ({
  // `user.country` is a one-of relation to the `countries` collection.
  country: one(countries),
}))

const user = await users.create({
  country: await countries.create({ code: 'usa' }),
})
user.country // { code: 'usa' }

const postSchema = z.object({
  get comments() {
    return z.array(countrySchema)
  },
})
const commentSchema = z.object({
  text: z.string(),
})

const posts = new Collection({ schema: postSchema })
const comments = new Collection({ schema: commentSchema })

posts.defineRelations(({ many }) => ({
  comments: many(comments),
}))

await posts.create({
  comments: [
    await comments.create({ text: 'First!' }),
    await comments.create({ text: 'Thanks for watching.' }),
  ],
})

Two collections may self-reference each other. For example, post.comments is a list of comments while each comment.post references to the parent post.

const postSchema = z.object({
  get comments() {
    return z.array(countrySchema)
  },
})
const commentSchema = z.object({
  text: z.string(),
  get post() {
    return postSchema
  },
})

const posts = new Collection({ schema: postSchema })
const comments = new Collection({ schema: commentSchema })

posts.defineRelations(({ many }) => ({
  comments: many(comments),
}))
comments.defineRelations(({ one }) => ({
  post: one(posts),
}))

await posts.create({
  comments: [await comments.create({ text: 'First!' })],
})

const comment = comments.findFirst((q) => q.where({ text: 'First!' }))
comment.post // { comments: [{ text: 'First', post: Circular }] }

Inversed relations are updated automatically. Whenever you add a new comment to a post, both post.comments and comment.post are updated to reference each other. The same is true when setting a new parent post on the comment.

In the next example, every user may have multiple posts while each post may have multiple authors.

const userSchema = z.object({
  get posts() {
    return z.array(postSchemas)
  },
})
const postSchema = z.object({
  get authors() {
    return z.array(userSchema)
  },
})

const users = new Collection({ schema: userSchema })
const posts = new Collection({ schema: postSchema })

users.defineRelations(({ many }) => ({
  posts: many(posts),
}))
posts.defineRelations(({ many }) => ({
  authors: many(users),
}))

Since relational properties resolve via getters, there's no need to define special "through" relations to reference one model through a relation from another.

const owners = new Collection({ schema: ownerSchema })
const cars = new Collection({ schema: carSchema })
const mechanics = new Collection({ schema: mechanicSchema })

owners.defineRelations(({ many }) => ({
  cars: many(cars),
}))
cars.defineRelations(({ one }) => ({
  owner: one(owners),
}))
mechanics.defineRelations(({ one }) => ({
  car: one(cars),
}))

const owner = await owners.create({ name: 'John' })
const car = await cars.create({ brand: 'bmw', owner })
const mechanic = await mechanics.create({ name: 'Kyle', car })

mechanic.car.owner.name // "John"

Note that although mechanics does not define an explicit relation to owners, you can get the owner of the car associated with a mechanic through the car relation.

You can mark a relation as unique by setting the unique property of the relation options to true. Unique relations cannot reference foreign records that are already associated with other owner records.

posts.defineRelations(({ one }) => ({
  author: one(users, { unique: true }),
}))

In this example, the author of each post points to a single unique user. If a post attempts to set its author to a user that's already associated with another post, an error will be thrown.

You can use the role option of the relation to disambiguate between multiple properties referencing the same foreign model.

For example, a single post may have both author and reviewer referencing the same user model. To make those properties pointing to different user records, use the role that acts as a relation identifier. This way, the library will update the corresponding relational properties for both users and posts when the referenced relation is updated.

const users = new Collection({ schema: userSchema })
const posts = new Collection({ schema: postSchema })

users.defineRelations(({ many }) => ({
  posts: many(posts, { role: 'author' }),
  reviews: many(posts, { role: 'reviewer' }),
}))

posts.defineRelations(({ one }) => ({
  author: one(user, { role: 'author' }),
  reviewer: one(user, { role: 'reviewer' }),
}))

The role property acts as a de-facto ID of a relation when synchronizing related models.

Provide an array of foreign collections to a relation to define it as polymorphic.

const posts = new Collection({ schema: postSchema })
const images = new Collection({ schema: imageSchema })
const videos = new Collection({ schema: videoSchema })

posts.defineRelations(({ many }) => ({
  // Providing a record of foreign collections allows
  // all of their records to be set as the value.
  attachments: many([images, videos]),
}))
images.defineRelations(({ one }) => ({
  post: one(posts),
}))
videos.defineRelations(({ one }) => ({
  post: one(posts),
}))

In this example, post.attachments is an array of either images or videos, where records from both collections are allowed.

Data provides multiple different error classes to help you differentiate and handle different errors.

• code <OperationErrorCode>, the error code describing the failed operation;
• cause <Error> (optional), a reference to the original thrown error.

Thrown whenever performing a record operation fails. For example:

• When creating a new record whose initial values do not match the collection's schema;
• When there are no records found for a strict query.

• code <RelationErrorCode>, the error code describing the relation operation;
• info <object>, additional error information;
  • path <PropertyPath>, path of the relational property;
  • ownerCollection <Collection>, a reference to the owner collection;
  • foreignCollection <Array<Collection>>, an array of foreign collections referenced by this relation;
  • options RelationDefinitionOptions, the options object passed upon decaring this relation.

Thrown whenever performing a relation operation fails. For example:

• When attempting to reference a foreign record that's already associated with another record in a unique relation;
• When directly assigning value to a relational property.

• options <Object>
  • schema Standard Schema A schema describing each record in this collection.
  • extensions (optional) An array of extensions to use on this collection.

Creates a new collection of data.

• initialValues Initial values for the new record.

Creates a single record with the provided initial values.

const user = await users.create({ id: 1, name: 'John' })

The .create() method returns a promise to support potential asynchronous transformations in your schema.

• count <number> A number of records to create.
• initialValuesFactory <Function> A function that returns initial values for each record.

Creates multiple records with the initial value factory.

const users = await users.createMany(5, (index) => ({
  id: index + 1,
  name: 'John',
}))

The initial value factory function accepts the index argument indicating the index of the record that's being created. Use it, as well as the function's closure, to generate unique or random values.

• query Query A query matching the record.

Returns the first record matching the query.

const users = new Collection({
  schema: z.object({
    id: z.number(),
    name: z.string(),
  }),
})

await users.create({ id: 1, name: 'John' })
await users.create({ id: 2, name: 'John' })

users.findFirst((q) => q.where({ name: 'John' }))
// { id: 1, name: 'John' }

• query Query A query matching the records.

Returns all records matching the query.

const users = new Collection({
  schema: z.object({
    id: z.number(),
    name: z.string(),
  }),
})

await users.create({ id: 1, name: 'John' })
await users.create({ id: 2, name: 'John' })

users.findFirst((q) => q.where({ name: 'John' }))
// [{ id: 1, name: 'John' }, { id: 2, name: 'John' }]

• query Query A query matching the record.
• options <Object>
  • data A function that produces changes by modifying the previous record.

Updates the first record matching the query. Returns a promise that resolves with the updated record.

// Change the name for the user with a specific `id`.
const updatedUser = await users.update((q) => q.where({ id: 123 }), {
  data(user) {
    user.name = 'Johnatan'
  },
})

Update methods return a promise in order to support potential asynchronous transformations defined in your schema.

The data function allows you to perform multiple updates upon a record by mutating that record directly. Think of it as a draft function from libraries like immer or mutative because that's precisely what it is!

You can also provide a record reference as the predicate to the .update() method to update that particular record:

const user = users.findFirst((q) => q.where({ id: 123 }))
await users.update(user, {
  //               👆👆
  data(user) {
    user.id = 456
  },
})

• query Query A query matching the records.
• options <Object>
  • data Changes to apply to each record.

Updates all records matching the query. Returns a promise that resolves with an array containing the updated records.

// Find all the users with the name "John"
// and make their name truly stand out!
const updatedUsers = await users.updateMany((q) => q.where({ name: 'John' }), {
  data(user) {
    user.name = user.name.toUpperCase()
  },
})

• query Query A query matching the record.

Deletes the first record matching the query. Returns the deleted record.

// Delete a user with a particular `id`.
const deletedUser = users.delete((q) => q.where({ id: 123 }))

You can also provide a record reference as the predicate to the .delete() method to delete that particular record:

const user = users.findFirst((q) => q.where({ id: 123 }))
users.delete(user)

• query Query A query matching the records.

Deletes all records matching the query. Returns an array containing the deleted records.

// Delete all users whose trial period has expired.
const deletedUsers = users.deleteMany((q) =>
  q.where({ trial: { expiresAt: (expiresAt) => expiresAt <= Date.now() } }),
)

• definition <Function> A function that accepts relation utilities and returns an object with relational properties.

Defines relations on the current collection.

const users = new Collection({ schema: userSchema })
const posts = new Collection({ schema: postSchema })

users.defineRelations(({ many }) => ({
  // `user.posts` is a many-of relation to `posts`.
  posts: many(posts)
}))

You can define nested relational properties by nesting them in the object returned from .defineRelations().

The following relational utilities are exposed in the argument to this method:

• one(collection[, options]), defines a one-of relation to the given collection;
• many(collection[, options]), defines a many-of relation to the given collection.

• unique <boolean>, marks this relation as unique. Foreign records referenced by this relation cannot be referenced by other models.

users.defineRelations(({ many }) => ({
  posts: many(posts)
}))
posts.defineRelations(({ one }) => ({
  author: one(users, { unique: true })
}))

const john = await users.create({
  name: 'John',
  // `john` is associated as the `author` of this post now.
  posts: [await posts.create({ title: 'First post' })]
})

await users.create({
  name: 'Katy',
  // Creating this user will error because it tries to list
  // a post which `author` already references to another user.
  posts: [john.posts[0]]
})

• role <string>, an identifier to differentiate ambiguous relations to the same foreign collection;

users.defineRelations(({ many }) => ({
  // Both `users` and `posts` reference each other in multiple keys.
  // Using `role` helps the library understand which keys are connected.
  posts: many(posts, { role: 'author' }),
  underReview: many(posts, { role: 'reviewer' })
}))

posts.defineRelations(({ one, many }) => ({
  author: one(users, { role: 'author' }),
  reviewers: many(users, { role: 'reviewer' })
}))

• onDelete "cascade" | undefined, decides how to handle referenced foreign records when the owner is deleted.

users.defineRelations(({ many }) => ({
  // If a user gets deleted, delete all of the `posts` associated with them.
  posts: many(posts, { onDelete: 'cascade' })
}))
posts.defineRelations(({ one }) => ({
  author: one(users)
}))

• predicate (optional) An object or a function that acts as a predicate for records.

Creates a new query to match records in a collection. Normally, you query records through the querying methods of the collection (see Querying). You can, however, construct a type-safe Query class to abstract common queries or query builders.

const userSchema = z.object({
  id: z.number(),
  subscribed: z.boolean().default(false),
  role: z.enum(['user', 'editor', 'admin']).default('user'),
})

// Creates a query builder for the users schema.
const query = new Query<typeof userSchema>()

• predicate A predicate for the records.
• Returns: Query.

query.where({ id: 2 })

• predicate A predicate or another Query.
• Returns: Query.

Creates a new query, merging the previous predicates with the new one under a OR relation. A record may match any predicate to be considered matching.

const unsubscribedOrEditorsQuery = query.or(
  query.where({ subscribed: false }),
  query.where({ role: 'editor' }),
)

• predicate A predicate or another Query.
• Returns: Query.

Creates a new query, merging the previous predicates with the new one under a AND relation. A record must match all predicates to be considered matching.

const exactAdminQuery = query.and(
  query.where({ id: 1 }),
  query.where({ role: 'admin' }),
)

• record <Record> A reference to a record to test.
• Returns <boolean> Indicates whether the given record matches the query.

query.test({ id: 1 })

You can extend the behavior of collections via extensions. The library comes with the following default extensions, but you can always create your own.

Synchronizes collection changes, like creating/updating/deleting records, with the same collection in another browser tab via a BroadcastChannel.

import { Collection } from '@msw/data'
import { sync } from '@msw/data/extensions/sync'

const users = new Collection({
  schema,
  extensions: [sync()],
})

Persist the records in the collection between page reloads.

import { Collection } from '@msw/data'
import { persist } from '@msw/data/extensions/persist'

const users = new Collection({
  schema,
  extensions: [persist()],
})

// my-extension.ts
import { defineExtension } from '@msw/data/extensions'

export function myExtension() {
  return defineExtension({
    name: 'my-extension',
    extend(collection) {
      // Your logic here.
    },
  })
}

import { Collection } from '@msw/data'
import { myExtension } from './my-extension.js'

new Collection({ schema, extensions: [myExtension()] })