Skip to main content

Migrate from Sequelize

This guide describes how to migrate from Sequelize to Prisma ORM. It uses an extended version of the Sequelize Express example as a sample project to demonstrate the migration steps. You can find the example used for this guide on GitHub.

This migration guide uses PostgreSQL as the example database, but it equally applies to any other relational database that's supported by Prisma.

You can learn how Prisma ORM compares to Sequelize on the Prisma ORM vs Sequelize page.

Overview of the migration process

Note that the steps for migrating from Sequelize to Prisma ORM are always the same, no matter what kind of application or API layer you're building:

  1. Install the Prisma CLI
  2. Introspect your database
  3. Create a baseline migration
  4. Install Prisma Client
  5. Gradually replace your Sequelize queries with Prisma Client

These steps apply, no matter if you're building a REST API (e.g. with Express, koa or NestJS), a GraphQL API (e.g. with Apollo Server, TypeGraphQL or Nexus) or any other kind of application that uses Sequelize for database access.

Prisma ORM lends itself really well for incremental adoption. This means, you don't have migrate your entire project from Sequelize to Prisma ORM at once, but rather you can step-by-step move your database queries from Sequelize to Prisma ORM.

Overview of the sample project

For this guide, we'll use a REST API built with Express as a sample project to migrate to Prisma ORM. It has four models/entities:

module.exports = (sequelize, DataTypes) => {
const User = sequelize.define('User', {
name: {
type: DataTypes.STRING,
},
email: {
type: DataTypes.STRING,
unique: true,
allowNull: false,
},
})

User.associate = (models) => {
User.hasMany(models.Post, {
foreignKey: 'authorId',
as: 'posts',
})
User.hasOne(models.Profile, {
onDelete: 'CASCADE',
foreignKey: 'userId',
})
}
return User
}

The models have the following relations:

  • 1-1: UserProfile
  • 1-n: UserPost
  • m-n: PostCategory

The corresponding tables have been created using a generated Sequelize migration.

In this guide, the route handlers are located in the src/controllers directory. The models are located in the src/models directory. From there, they are pulled into a central src/routes.js file which is used to set up the required routes in src/index.js:

└── blog-sequelize
├── package.json
└──src
   ├── controllers
   │   ├── post.js
   │   └── user.js
   ├── models
   │   ├── Category.js
   │   ├── Post.js
   │   ├── Profile.js
   │   └── User.js
   ├── index.js
   └── routes.js

Step 1. Install the Prisma CLI

The first step to adopt Prisma ORM is to install the Prisma CLI in your project:

npm install prisma --save-dev

Step 2. Introspect your database

2.1. Set up Prisma ORM

Before you can introspect your database, you need to set up your Prisma schema and connect Prisma ORM to your database. Run the following command in your terminal to create a basic Prisma schema file:

npx prisma init

This command created a new directory called prisma with the following files for you:

  • schema.prisma: Your Prisma schema that specifies your database connection and models
  • .env: A dotenv to configure your database connection URL as an environment variable

The Prisma schema currently looks as follows:

prisma/schema.prisma
// This is your Prisma schema file,
// learn more about it in the docs: https://pris.ly/d/prisma-schema

datasource db {
provider = "postgresql"
url = env("DATABASE_URL")
}

generator client {
provider = "prisma-client-js"
}
tip

If you're using VS Code, be sure to install the Prisma VS Code extension for syntax highlighting, formatting, auto-completion and a lot more cool features.

2.2. Connect your database

If you're not using PostgreSQL, you need to adjust the provider field on the datasource block to the database you currently use:

datasource db {
provider = "postgresql"
url = env("DATABASE_URL")
}

Once that's done, you can configure your database connection URL in the .env file. Here's how the database connection from Sequelize maps to the connection URL format used by Prisma ORM:


Assume you have the following database connection details in src/models/index.js:


src/models/index.js
const sequelize = new Sequelize('blog-sequelize', 'alice', 'myPassword42', {
host: 'localhost',
dialect: 'postgres',
})

The respective connection URL would look as follows in Prisma ORM:


.env
DATABASE_URL="postgresql://alice:myPassword42@localhost:5432/blog-sequelize"

Note that you can optionally configure the PostgreSQL schema by appending the schema argument to the connection URL:


.env
DATABASE_URL="postgresql://alice:myPassword42@localhost:5432/blog-sequelize?schema=myschema"

If not provided, the default schema called public is being used.


2.3. Introspect your database using Prisma ORM

With your connection URL in place, you can introspect your database to generate your Prisma models:

npx prisma db pull

This creates the following Prisma models:

prisma/schema.prisma
model Categories {
id Int @id @default(autoincrement())
name String
createdAt DateTime
updatedAt DateTime
PostCategories PostCategories[]
}

model PostCategories {
createdAt DateTime
updatedAt DateTime
CategoryId Int
PostId Int
Categories Categories @relation(fields: [CategoryId], references: [id])
Posts Posts @relation(fields: [PostId], references: [id])

@@id([CategoryId, PostId])
}

model Posts {
id Int @id @default(autoincrement())
title String
content String?
published Boolean? @default(false)
createdAt DateTime
updatedAt DateTime
authorId Int?
Users Users? @relation(fields: [authorId], references: [id])
PostCategories PostCategories[]
}

model Profiles {
id Int @id @default(autoincrement())
bio String
createdAt DateTime
updatedAt DateTime
userId Int? @unique
Users Users? @relation(fields: [userId], references: [id])
}

model SequelizeMeta {
name String @id
}

model Users {
id Int @id @default(autoincrement())
name String?
email String @unique
createdAt DateTime
updatedAt DateTime
Posts Posts[]
Profiles Profiles?
}

2.4. Create a baseline migration

To continue using Prisma Migrate to evolve your database schema, you will need to baseline your database.

First, create a migrations directory and add a directory inside with your preferred name for the migration. In this example, we will use 0_init as the migration name:

mkdir -p prisma/migrations/0_init

Next, generate the migration file with prisma migrate diff. Use the following arguments:

  • --from-empty: assumes the data model you're migrating from is empty
  • --to-schema-datamodel: the current database state using the URL in the datasource block
  • --script: output a SQL script
npx prisma migrate diff --from-empty --to-schema-datamodel prisma/schema.prisma --script > prisma/migrations/0_init/migration.sql

Review the generated migration to ensure everything is correct.

Next, mark the migration as applied using prisma migrate resolve with the --applied argument.

npx prisma migrate resolve --applied 0_init

The command will mark 0_init as applied by adding it to the _prisma_migrations table.

You now have a baseline for your current database schema. To make further changes to your database schema, you can update your Prisma schema and use prisma migrate dev to apply the changes to your database.

2.5. Adjust createdAt and updatedAt fields

The generated Prisma models represent your database tables and are the foundation for your programmatic Prisma Client API which allows you to send queries to your database. You'll adjust the createdAt and updatedAt fields in our models. Sequelize doesn't add the DEFAULT constraint to createdAt when creating the tables in the database. Therefore, you'll add @default(now()) and @updatedAt attributes to the createdAt and updatedAt columns respectively. To learn more how Prisma ORM does this, you can read more @default(now()) and @updatedAt here. Our updated schema will be as follows:

prisma/schema.prisma
model Categories {
id Int @id @default(autoincrement())
name String
createdAt DateTime @default(now())
updatedAt DateTime @updatedAt
PostCategories PostCategories[]
}

model PostCategories {
createdAt DateTime @default(now())
updatedAt DateTime @updatedAt
CategoryId Int
PostId Int
Categories Categories @relation(fields: [CategoryId], references: [id])
Posts Posts @relation(fields: [PostId], references: [id])

@@id([CategoryId, PostId])
}

model Posts {
id Int @id @default(autoincrement())
title String
content String?
published Boolean? @default(false)
createdAt DateTime @default(now())
updatedAt DateTime @updatedAt
authorId Int?
Users Users? @relation(fields: [authorId], references: [id])
PostCategories PostCategories[]
}

model Profiles {
id Int @id @default(autoincrement())
bio String
createdAt DateTime @default(now())
updatedAt DateTime @updatedAt
userId Int? @unique
Users Users? @relation(fields: [userId], references: [id])
}

model SequelizeMeta {
name String @id
}

model Users {
id Int @id @default(autoincrement())
name String?
email String @unique
createdAt DateTime @default(now())
updatedAt DateTime @updatedAt
Posts Posts[]
Profiles Profiles?
}

2.6. Adjust the Prisma schema (optional)

The models that were generated via introspection currently exactly map to your database tables. In this section, you'll learn how you can adjust the naming of the Prisma models to adhere to Prisma ORM's naming conventions.

All of these adjustment are entirely optional and you are free to skip to the next step already if you don't want to adjust anything for now. You can go back and make the adjustments at any later point.

As opposed to the current snake_case notation of Prisma models, Prisma ORM's naming conventions are:

  • PascalCase for model names
  • camelCase for field names

You can adjust the naming by mapping the Prisma model and field names to the existing table and column names in the underlying database using @@map and @map.

Also note that you can rename relation fields to optimize the Prisma Client API that you'll use later to send queries to your database. For example, although we are singularizing the Posts model name to Post, the posts field on the user model is a list, so it makes sense to keep that named posts to indicate that it's plural.

Sequelize generates a SequelizeMeta model that is used internally by the library that is not needed. Therefore, you'll manually delete it from the schema.

Here's an adjusted version of the Prisma schema that addresses these points:

prisma/schema.prisma
generator client {
provider = "prisma-client-js"
}

datasource db {
provider = "postgresql"
url = env("DATABASE_URL")
}

model Category {
id Int @id @default(autoincrement())
name String
createdAt DateTime @default(now())
updatedAt DateTime @updatedAt
postCategories PostToCategories[]

@@map("Categories")
}

model PostToCategories {
createdAt DateTime @default(now())
updatedAt DateTime @updatedAt
categoryId Int
postId Int
category Category @relation(fields: [categoryId], references: [id])
post Post @relation(fields: [postId], references: [id])

@@id([categoryId, postId])
@@map("PostCategories")
}

model Post {
id Int @id @default(autoincrement())
title String
content String?
published Boolean? @default(false)
createdAt DateTime @default(now())
updatedAt DateTime @updatedAt
authorId Int?
author User? @relation(fields: [authorId], references: [id])
postToCategories PostToCategories[]

@@map("Posts")
}

model Profile {
id Int @id @default(autoincrement())
bio String
createdAt DateTime @default(now())
updatedAt DateTime @updatedAt
userId Int? @unique
user User? @relation(fields: [userId], references: [id])

@@map("Profiles")
}

model User {
id Int @id @default(autoincrement())
name String?
email String @unique
createdAt DateTime @default(now())
updatedAt DateTime @updatedAt
posts Post[]
profile Profile?

@@map("Users")
}

Step 3. Install Prisma Client

As a next step, you can install Prisma Client in your project so that you can start replacing the database queries in your project that are currently made with Sequelize:

npm install @prisma/client

Step 4. Replace your Sequelize queries with Prisma Client

In this section, we'll show a few sample queries that are being migrated from Sequelize to Prisma Client based on the example routes from the sample REST API project. For a comprehensive overview of how the Prisma Client API differs from Sequelize, check out the API comparison page.

First, to set up the PrismaClient instance that you'll use to send database queries from the various route handlers. Create a new file named prisma.js in the src directory:

touch src/prisma.js

Now, instantiate PrismaClient and export it from the file so you can use it in your route handlers later:

src/prisma.js
const { PrismaClient } = require('@prisma/client')

const prisma = new PrismaClient()

module.exports = prisma

The imports in our controller files are as follows:

src/controllers/post.js
const { Post, User, Category } = require('../models')
const { Op } = require('sequelize')
src/controllers/user.js
const { User } = require('../models')

You'll update the controller imports as you migrate from Sequelize to Prisma:

src/controllers/post.js
const prisma = require('../prisma')
src/controllers/user.js
const prisma = require('../prisma')

4.1. Replacing queries in GET requests

The REST API has four routes that accept GET requests:

  • /feed: Return all published posts
  • /filterPosts?searchString=SEARCH_STRING: Filter returned posts by SEARCH_STRING
  • /post/:postId: Returns a specific post
  • /authors: Returns a list of authors

Let's dive into the route handlers that implement these requests.

/feed

The /feed handler is currently implemented as follows:

src/controllers/post.js
const feed = async (req, res) => {
try {
const feed = await Post.findAll({
where: { published: true },
include: ['author', 'categories'],
})
return res.json(feed)
} catch (error) {
return res.status(500).json(error)
}
}

Note that each returned Post object includes the relation to the author and category it's associated with. With Sequelize, including the relation is not type-safe. For example, if there was a typo in the relation that is retrieved, your database query would fail only at runtime – the JavaScript compiler does not provide any safety here.

Here is how the same route is implemented using Prisma Client:

src/controllers/post.js
const feed = async (req, res) => {
try {
const feed = await prisma.post.findMany({
where: { published: true },
include: { author: true, postToCategories: true },
})
return res.json(feed)
} catch (error) {
return res.status(500).json(error)
}
}

Note that the way how Prisma Client includes the author relation is absolutely type-safe. The JavaScript compiler would throw an error if you were trying to include a relation that does not exist on the Post model.

/filterPosts?searchString=SEARCH_STRING

The /filterPosts handler is currently implemented as follows:

src/controllers/post.js
const filterPosts = async (req, res) => {
const { searchString } = req.query

try {
const filteredPosts = await Post.findAll({
where: {
[Op.or]: [
{
title: {
[Op.like]: `%${searchString}%`,
},
},
{
content: {
[Op.like]: `%${searchString}%`,
},
},
],
},
include: 'author',
})

res.json(filteredPosts)
} catch (error) {
return res.status(500).json(error)
}
}

With Prisma ORM, the route is implemented as follows:

src/controllers/post.js
const filterPosts = async (req, res) => {
const { searchString } = req.query

try {
const filteredPosts = prisma.post.findMany({
where: {
OR: [
{
title: { contains: searchString },
},
{
content: { contains: searchString },
},
],
},
})

res.json(filteredPosts)
} catch (error) {
return res.status(500).json(error)
}
}

Note that Sequelize provides Operator symbols - Op - to be used when querying data. Prisma ORM on the other hand combines several where conditions with an implicit AND operator, so in this case the Prisma Client query needs to make the OR explicit.

/post/:postId

The /post/:postId handler is currently implemented as follows:

src/controllers/post.js
const getPostById = async (req, res) => {
const { postId } = req.params

try {
const post = await Post.findOne({
where: { id: postId },
include: 'author',
})

return res.json(post)
} catch (error) {
return res.status(500).json(error)
}
}

With Prisma ORM, the route is implemented as follows:

src/controllers/post.js
const getPostById = async (req, res) => {
const { postId } = req.params

try {
const post = await prisma.post.findUnique({
where: { id: Number(postId) },
include: { author: true },
})

return res.json(post)
} catch (error) {
return res.status(500).json(error)
}
}

4.2. Replacing queries in POST requests

The REST API has three routes that accept POST requests:

  • /user: Creates a new User record
  • /post: Creates a new User record
  • /user/:userId/profile: Creates a new Profile record for a User record with a given ID

/user

The /user handler is currently implemented as follows:

src/controllers/user.js
const createUser = async (req, res) => {
const { name, email } = req.body

try {
const user = await User.create({
name,
email,
})

return res.json(user)
} catch (error) {
return res.status(500).json(error)
}
}

With Prisma ORM, the route is implemented as follows:

src/controllers/user.js
const createUser = async (req, res) => {
const { name, email } = req.body

try {
const user = await prisma.user.create({
data: {
name,
email,
},
})

return res.json(user)
} catch (error) {
return res.status(500).json(error)
}
}

/post

The /post handler is currently implemented as follows:

src/controllers/post.js
const createDraft = async (req, res) => {
const { title, content, authorEmail } = req.body

try {
const user = await User.findOne({ email: authorEmail })

const draft = await Post.create({
title,
content,
authorId: user.id,
})

res.json(draft)
} catch (error) {
return res.status(500).json(error)
}
}

With Prisma ORM, the route is implemented as follows:

src/controllers/post.js
const createDraft = async (req, res) => {
const { title, content, authorEmail } = req.body

try {
const draft = await prisma.post.create({
data: {
title,
content,
author: {
connect: { email: authorEmail },
},
},
})

res.json(draft)
} catch (error) {
return res.status(500).json(error)
}
}

Note that Prisma Client's nested write here save an initial query where first the User record needs to be retrieved by its email. That's because, with Prisma Client you can connect records in relations using any unique property.

/user/:userId/profile

The /user/:userId/profile handler is currently implemented as follows:

src/controllers/user.js
const setUserBio = async (req, res) => {
const { userId } = req.params
const { bio } = req.body

try {
const user = await User.findOne({
where: {
id: Number(userId),
},
})

const updatedUser = await user.createProfile({ bio })

return res.json(updatedUser)
} catch (error) {
return res.status(500).json(error)
}
}

With Prisma ORM, the route is implemented as follows:

src/controllers/user.js
const setUserBio = async (req, res) => {
const { userId } = req.params
const { bio } = req.body

try {
const user = await prisma.user.update({
where: { id: Number(userId) },
data: {
profile: {
create: { bio },
},
},
})

return res.json(user)
} catch (error) {
return res.status(500).json(error)
}
}

4.3. Replacing queries in PUT requests

The REST API has one route that accept a PUT request:

  • /addPostToCategory?postId=POST_ID&categoryId=CATEGORY_ID: Adds the post with POST_ID to the category with CATEGORY_ID

Let's dive into the route handlers that implement these requests.

/addPostToCategory?postId=POST_ID&categoryId=CATEGORY_ID

The /addPostToCategory?postId=POST_ID&categoryId=CATEGORY_ID handler is currently implemented as follows:

src/controllers/post.js
const addPostToCategory = async (req, res) => {
const { postId, categoryId } = req.query

try {
const post = await Post.findOne({
where: { id: postId },
})

const category = await Category.findOne({
where: { id: categoryId },
})

const updatedPost = await post.addCategory(category)

return res.json(updatedPost)
} catch (error) {
return res.status(500).json(error)
}
}

With Prisma ORM, the route is implemented as follows:

src/controllers/post.js
const addPostToCategory = async (req, res) => {
const { postId, categoryId } = req.query

try {
const post = await prisma.post.update({
data: {
postToCategories: {
create: {
categories: {
connect: { id: Number(categoryId) },
},
},
},
},
where: {
id: Number(postId),
},
})

return res.json(post)
} catch (error) {
return res.status(500).json(error)
}
}

Note that this Prisma Client can be made less verbose by modeling the relation as an implicit many-to-many relation instead. In that case, the query would look as follows:

src/controllers/posts.js
const post = await prisma.post.update({
data: {
category: {
connect: { id: categoryId },
},
},
where: { id: postId },
})

More

Primary key column

By default, Sequelize defines a primaryKey and used id with the autoby default if not defined. This is optional. If you would like to set your own primary key, you can use the primaryKey: true and define your preferred data type in your field of choice:

// changing the primary key column
module.exports = (sequelize, DataTypes) => {
const Post = sequelize.define('Post', {
postId: {
type: DataTypes.INTEGER,
primaryKey: true,
},
})
return Post
}

// changing the id DataType
module.exports = (sequelize, DataTypes) => {
const Post = sequelize.define('Post', {
id: {
type: DataTypes.UUID, // alternative: DataTypes.STRING
primaryKey: true,
},
})
return Post
}

Table name inference

Sequelize infers table names from the model name. When the name of a table isn't provided Sequelize automatically pluralizes the model name and uses that as the table name using a library called inflection. Prisma ORM on the other hand maps the model name to the table name in your database modelling your data. If you wish to change this default behaviour in Sequelize, you can either enforce the table name to be equal to the model name or provide the table name directly:

// enforcing table name to be equal to model name
module.exports = (sequelize, DataTypes) => {
const Post = sequelize.define(
'Post',
{
// ... attributes
},
{
freezeTableName: true,
}
)
return Post
}
// providing the table name directly
module.exports = (sequelize, DataTypes) => {
const Post = sequelize.define(
'Post',
{
// ... attributes
},
{
tableName: 'Post',
}
)
return Post
}

Timestamps

Sequelize automatically adds the fields createdAt and updatedAt to every model using the data type DataTypes.DATE, by default. You can disable this for a model with the timestamps: false option:

sequelize.define(
'User',
{
// ... (attributes)
},
{
timestamps: false,
}
)

Prisma ORM offers you the flexibility to define these fields in your model. You add the createdAt and updatedAt fields by defining them explicitly in your model. To set the createdAt field in your model, add the default(now()) attribute to the column. In order to set the updatedAt column, update your model by adding the @updatedAt attribute to the column.

model User {
id Int @id @default(autoincrement())
name String?
email String @unique
createdAt DateTime @default(now())
updatedAt DateTime @updatedAt
}

Implicit many-to-many relations

Similar to the belongsToMany() association method in Sequelize, Prisma ORM allows you to model many-to-many relations implicitly. That is, a many-to-many relation where you do not have to manage the relation table (also sometimes called JOIN table) explicitly in your schema. Here is an example with Sequelize:

module.exports = (sequelize, DataTypes) => {
const Post = sequelize.define('Post', {
title: {
type: DataTypes.STRING,
allowNull: false,
},
content: {
type: DataTypes.STRING,
},
published: {
type: DataTypes.BOOLEAN,
defaultValue: false,
},
})
Post.associate = (models) => {
Post.belongsTo(models.User, {
foreignKey: 'authorId',
as: 'author',
})
Post.belongsToMany(models.Category, {
through: 'PostCategories',
as: 'categories',
})
}
return Post
}
module.exports = (sequelize, DataTypes) => {
const Category = sequelize.define('Category', {
name: {
type: DataTypes.STRING,
allowNull: false,
},
})
Category.associate = (models) => {
Category.belongsToMany(models.Post, {
through: 'PostCategories',
as: 'posts',
})
}
return Category
}

When you start your application, Sequelize will create the the tables for you - based on these models:

Executing (default): CREATE TABLE IF NOT EXISTS "PostCategories"
("createdAt" TIMESTAMP WITH TIME ZONE NOT NULL, "updatedAt" TIMESTAMP WITH TIME ZONE NOT NULL,
"CategoryId" INTEGER REFERENCES "Categories" ("id") ON DELETE CASCADE ON UPDATE CASCADE,
"PostId" INTEGER REFERENCES "Posts" ("id") ON DELETE CASCADE ON UPDATE CASCADE, PRIMARY KEY ("CategoryId","PostId"));

If you introspect the database with Prisma ORM, you'll get the following result in the Prisma schema (note that some relation field names have been adjusted to look friendlier compared to the raw version from introspection):

model Categories {
id Int @id @default(autoincrement())
name String
createdAt DateTime
updatedAt DateTime
PostCategories PostCategories[]

@@map("category")
}

model PostCategories {
createdAt DateTime
updatedAt DateTime
CategoryId Int
PostId Int
Categories Categories @relation(fields: [CategoryId], references: [id])
Posts Posts @relation(fields: [PostId], references: [id])

@@id([CategoryId, PostId])
@@map("PostCategories")
}

model Posts {
id Int @id @default(autoincrement())
title String
content String?
published Boolean? @default(false)
createdAt DateTime
updatedAt DateTime
authorId Int?
Users Users? @relation(fields: [authorId], references: [id])
PostCategories PostCategories[]

@@map("post")
}

In this Prisma schema, the many-to-many relation is modeled explicitly via the relation table PostCategories

By adhering to the conventions for Prisma relation tables, the relation could look as follows:

model Categories {
id Int @id @default(autoincrement())
name String
posts Posts[]

@@map("category")
}

model Posts {
id Int @id @default(autoincrement())
title String
content String?
published Boolean @default(false)
authorId Int?
author User? @relation(fields: [authorId], references: [id])
categories Categories[]

@@map("post")
}

This would also result in a more ergonomic and less verbose Prisma Client API to modify the records in this relation, because you have a direct path from Post to Category (and the other way around) instead of needing to traverse the PostCategories model first.