Introspection

JavaScript

CockroachDB

Introspect your database with Prisma ORM

For the purpose of this guide, we'll use a demo SQL schema with three tables:

CREATE TABLE "User" (
  id INT8 PRIMARY KEY DEFAULT unique_rowid(),
  name STRING(255),
  email STRING(255) UNIQUE NOT NULL
);

CREATE TABLE "Post" (
  id INT8 PRIMARY KEY DEFAULT unique_rowid(),
  title STRING(255) UNIQUE NOT NULL,
  "createdAt" TIMESTAMP NOT NULL DEFAULT now(),
  content STRING,
  published BOOLEAN NOT NULL DEFAULT false,
  "authorId" INT8 NOT NULL,
  FOREIGN KEY ("authorId") REFERENCES "User"(id)
);

CREATE TABLE "Profile" (
  id INT8 PRIMARY KEY DEFAULT unique_rowid(),
  bio STRING,
  "userId" INT8 UNIQUE NOT NULL,
  FOREIGN KEY ("userId") REFERENCES "User"(id)
);

Note: Some fields are written in double quotes to ensure CockroachDB uses proper casing. If no double-quotes were used, CockroachDB would just read everything as lowercase characters.

Expand for a graphical overview of the tables

User

Column name	Type	Primary key	Foreign key	Required	Default
id	INT8	✔️	No	✔️	autoincrementing
name	STRING(255)	No	No	No	-
email	STRING(255)	No	No	✔️	-

Post

Column name	Type	Primary key	Foreign key	Required	Default
id	INT8	✔️	No	✔️	autoincrementing
createdAt	TIMESTAMP	No	No	✔️	now()
title	STRING(255)	No	No	✔️	-
content	STRING	No	No	No	-
published	BOOLEAN	No	No	✔️	false
authorId	INT8	No	✔️	✔️	-

Profile

Column name	Type	Primary key	Foreign key	Required	Default
id	INT8	✔️	No	✔️	autoincrementing
bio	STRING	No	No	No	-
userId	INT8	No	✔️	✔️	-

As a next step, you will introspect your database. The result of the introspection will be a data model inside your Prisma schema.

Run the following command to introspect your database:

npx prisma db pull

This commands reads the environment variable used to define the url in your schema.prisma, DATABASE_URL, that in our case is set in .env and connects to your database. Once the connection is established, it introspects the database (i.e. it reads the database schema). It then translates the database schema from SQL into a Prisma data model.

After the introspection is complete, your Prisma schema file was updated:

The data model now looks similar to this:

prisma/schema.prisma

model Post {
  id        BigInt   @id @default(autoincrement())
  title     String   @unique @db.String(255)
  createdAt DateTime @default(now()) @db.Timestamp(6)
  content   String?
  published Boolean  @default(false)
  authorId  BigInt
  User      User     @relation(fields: [authorId], references: [id], onDelete: NoAction, onUpdate: NoAction)
}

model Profile {
  id     BigInt  @id @default(autoincrement())
  bio    String?
  userId BigInt  @unique
  User   User    @relation(fields: [userId], references: [id], onDelete: NoAction, onUpdate: NoAction)
}

model User {
  id      BigInt   @id @default(autoincrement())
  name    String?  @db.String(255)
  email   String   @unique @db.String(255)
  Post    Post[]
  Profile Profile?
}

Prisma ORM's data model is a declarative representation of your database schema and serves as the foundation for the generated Prisma Client library. Your Prisma Client instance will expose queries that are tailored to these models.

Right now, there's a few minor "issues" with the data model:

• The User relation field is uppercased and therefore doesn't adhere to Prisma's naming conventions . To express more "semantics", it would also be nice if this field was called author to describe the relationship between User and Post better.
• The Post and Profile relation fields on User as well as the User relation field on Profile are all uppercased. To adhere to Prisma's naming conventions , both fields should be lowercased to post, profile and user.
• Even after lowercasing, the post field on User is still slightly misnamed. That's because it actually refers to a list of posts – a better name therefore would be the plural form: posts.

These changes are relevant for the generated Prisma Client API where using lowercased relation fields author, posts, profile and user will feel more natural and idiomatic to JavaScript/TypeScript developers. You can therefore configure your Prisma Client API.

Because relation fields are virtual (i.e. they do not directly manifest in the database), you can manually rename them in your Prisma schema without touching the database:

prisma/schema.prisma

model Post {
  id        BigInt   @id @default(autoincrement())
  title     String   @unique @db.String(255)
  createdAt DateTime @default(now()) @db.Timestamp(6)
  content   String?
  published Boolean  @default(false)
  authorId  BigInt
  author    User     @relation(fields: [authorId], references: [id], onDelete: NoAction, onUpdate: NoAction)
}

model Profile {
  id     BigInt  @id @default(autoincrement())
  bio    String?
  userId BigInt  @unique
  user   User    @relation(fields: [userId], references: [id], onDelete: NoAction, onUpdate: NoAction)
}

model User {
  id      BigInt   @id @default(autoincrement())
  name    String?  @db.String(255)
  email   String   @unique @db.String(255)
  posts   Post[]
  profile Profile?
}

In this example, the database schema did follow the naming conventions for Prisma ORM models (only the virtual relation fields that were generated from introspection did not adhere to them and needed adjustment). This optimizes the ergonomics of the generated Prisma Client API.

Using custom model and field names

Sometimes though, you may want to make additional changes to the names of the columns and tables that are exposed in the Prisma Client API. A common example is to translate snake_case notation which is often used in database schemas into PascalCase and camelCase notations which feel more natural for JavaScript/TypeScript developers.

Assume you obtained the following model from introspection that's based on snake_case notation:

model my_user {
  user_id    Int     @id @default(sequence())
  first_name String?
  last_name  String  @unique
}

If you generated a Prisma Client API for this model, it would pick up the snake_case notation in its API:

const user = await prisma.my_user.create({
    data: {
        first_name: 'Alice',
        last_name: 'Smith',
    },
})

If you don't want to use the table and column names from your database in your Prisma Client API, you can configure them with @map and @@map:

model MyUser {
  userId    Int     @id @default(sequence()) @map("user_id")
  firstName String? @map("first_name")
  lastName  String  @unique @map("last_name")

  @@map("my_user")
}

With this approach, you can name your model and its fields whatever you like and use the @map (for field names) and @@map (for models names) to point to the underlying tables and columns. Your Prisma Client API now looks as follows:

const user = await prisma.myUser.create({
    data: {
        firstName: 'Alice',
        lastName: 'Smith',
    },
})

Learn more about this on the Configuring your Prisma Client API page.