Data Modelling & GraphQL API

Prisma introduces a few changes to the way how your data model is written as well as to the generated GraphQL API.

Data modelling

Remove @model directive

The @model directive that was previously required to denote your model types is removed.

Before

type User @model {
  id: ID! @isUnique
  name: String!
}

After

type User {
  id: ID! @unique
  name: String!
}

@relation directive becomes optional on unambiguous relations

When a relation in your data model is unambiguous, you can omit the @relation directive.

Before

type User @model {
  id: ID! @isUnique
  name: String!
  posts: [Post!]! @relation(name: "UsersPosts")
}

type Post @model {
  id: ID! @isUnique
  title: String!
  author: User! @relation(name: "UsersPosts")
}

After

type User @model {
  id: ID! @unique
  name: String!
  posts: [Post!]!
}

type Post @model {
  id: ID! @unique
  title: String!
  author: User!
}

id field is optional

The id field is now optional on the model types in your data model (similar to createdAt and updatedAt), you can remove it if it's not needed on a type.

@isUnique is renamed to @unique

The @isUnique directive is renamed to @unique.

Before

type User {
  id: ID! @isUnique
  email: String! @isUnique
}

After

type User {
  id: ID! @unique
  email: String! @unique
}

@defaultValue is renamed to @default

The @defaultValue directive is renamed to @default.

Before

type User {
  id: ID! @isUnique
  name: String! @defaultValue(value: "Unknown")
}

After

type User {
  id: ID! @unique
  name: String! @default(value: "Unknown")
}

GraphQL API

Unified API: Merging Simple & Relay APIs

Most notably of all API changes, Prisma merges the previous Simple and Relay APIs. The resulting API is compatible with all GraphQL clients. Consequently, each Prisma service only provides a single HTTP endpoint.

Wrapped input arguments

The Simple API of the Graphcool Framework followed the approach of passing single values to mutations. In the new Prisma API, all input arguments for mutations are wrapped in one data argument.

Before

mutation {
  createPost(title: "GraphQL is great" text: "It really is") {
    id
  }
}

After

mutation {
  createPost(data: {
    title: "GraphQL is great"
    text: "It really is"
  }) {
    id
  }
}

Removing all-prefix from query root fields

The generated queries to return lists of nodes have the all-prefix removed.

Before

query {
  allUsers {
    id
    name
  }
}

After

query {
  users {
    id
    name
  }
}

Lowercasing query root fields asking for single nodes

When asking for a single node, the corresponding query is now lowercased.

query {
  User(id: "cjd5pqjuzpbuy0171tiuj098t") {
    id
    name
  }
}

After

query {
  user(id: "cjd5pqjuzpbuy0171tiuj098t") {
    id
    name
  }
}

filter renamed to where

The filter argument has been renamed to where in the Prisma GraphQL API. \

Before

query {
  allUsers(filter: {
    name_contains: "Karl"
  }) {
    id
    name
  }
}

After

query {
  users(where: {
    name_contains: "Karl"
  }) {
    id
    name
  }
}

Selecting nodes by any @unique field

In the Graphcool Framework GraphQL API, it was only possible to update and delete nodes by selecting them via their id field. With Prisma, you can use any field that's annotated with the @unique directive for that.

Consider this data model:

type User {
  id: ID! @unique
  email: String! @unique
}

With Prisma, you can now send the following mutation to delete a User node:

mutation {
  deleteUser(by: {
    email: "[email protected]"
  }) {
    id
  }
}

New API features

New API features introduced in Prisma include batch operations, improved nested mutations, transactional mutations and more.

This means that for these use cases, you can now use these new primitives instead of following a more complex setup as before.