This library uses an abstract syntax tree to parse schema.prisma files into an object in JavaScript. It also allows you to update your Prisma schema files using a Builder object pattern that is fully implemented in TypeScript.

It is similar to @prisma/sdk except that it preserves comments and model attributes. It also doesn't attempt to validate the correctness of the schema at all; the focus is instead on the ability to parse the schema into an object, manipulate it using JavaScript, and re-print the schema back to a file without losing information that isn't captured by other parsers.

It is probable that a future version of @prisma/sdk will render this library obsolete.

npm install @mrleebo/prisma-ast

produceSchema(source: string, (builder: PrismaSchemaBuilder) => void, printOptions?: PrintOptions): string

produceSchema is the simplest way to interact with prisma-ast; you input your schema source and a producer function to produce modifications to it, and it will output the schema source with your modifications applied.

import { produceSchema } from '@mrleebo/prisma-ast'

const input = `
model User {
  id   Int    @id @default(autoincrement())
  name String @unique
}
`

const output = produceSchema(source, (builder) => {
  builder
    .model("AppSetting")
    .field('key', 'String', [{ name: 'id' }])
    .field('value', 'Json')
})

model User {
  id   Int    @id @default(autoincrement())
  name String @unique
}

model AppSetting {
  key   String @id
  value Json
}

For more information about what the builder can do, check out the PrismaSchemaBuilder class.

The produceSchema() utility will construct a builder for you, but you can also create your own instance, which may be useful for more interactive use-cases.

import { createPrismaSchemaBuilder } from '@mrleebo/prisma-ast'

const builder = createPrismaSchemaBuilder()

builder.model('User')
  .field('id', 'Int')
  .attribute('id')
  .attribute('default', [{ function: 'autoincrement' }])
  .field('name', 'String')
  .attribute('unique')
  .break()
  .comment("this is a comment")
  .blockAttribute('index', ['name'])

const output = builder.print()

model User {
  id   Int @id @default(autoincrement())
  name String @unique

  // this is a comment
  @@index([name])
}

prisma-ast can sort the schema for you. The default sort order is ['generator', 'datasource', 'model', 'enum'] and will sort objects of the same type alphabetically.

print(options?: { 
  sort: boolean,
  locales?: string | string[],
  sortOrder?: Array<'generator' | 'datasource' | 'model' | 'enum'>
})

You can optionally set your own sort order, or change the locale used by the sort.

// sort with default parameters
builder.print({ sort: true })

// sort with options
builder.print({ 
  sort: true, locales: 'en-US', sortOrder: ['datasource', 'generator', 'model', 'enum']
});

Since a schema can only have one datasource, calling this command will override the existing datasource if the schema already has one, or create a datasource block if it doesn't.

datasource(provider: string, url: string | { env: string })

You can set a datasource by passing in the provider and url parameters.

builder.datasource('postgresql', { env: 'DATABASE_URL' })

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

If you want to perform a custom action that there isn't a Builder method for, you can access the underlying schema object programmatically.

import { Datasource } from '@mrleebo/prisma-ast'

// rename the datasource programmatically
builder.datasource('postgresql', { env: 'DATABASE_URL' }).then<Datasource>((datasource) => {
  datasource.name = "DS"
})

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

generator(name: string, provider: string)

If the schema already has a generator with the given name, it will be updated. Otherwise, a new generator will be created.

builder.generator('nexusPrisma', 'nexus-prisma')

generator nexusPrisma {
  provider = "nexus-prisma"
}

assignment(key: string, value: string)

If your generator accepts additional assignments, they can be added by chaining .assignment() calls to your generator.

builder.generator('client', 'prisma-client-js').assignment('output', 'db.js')

generator client {
  provider = "prisma-client-js"
  output   = "db.js"
}

If you want to perform a custom action that there isn't a Builder method for, you can access the underlying schema object programmatically.

generator client {
  provider = "prisma-client-js"
  output   = "db.js"
}

import { Generator } from '@mrleebo/prisma-ast'

// rename the generator programmatically
builder.generator('client').then<Generator>((generator) => {
  generator.name = "foo"
})

generator foo {
  provider = "prisma-client-js"
  output   = "db.js"
}

If the model with that name already exists in the schema, it will be selected and any fields that follow will be appended to the model. Otherwise, the model will be created and added to the schema.

builder.model('Project').field('name', 'String')

model Project {
  name String
}

If you want to perform a custom action that there isn't a Builder method for, you can access the underlying schema object programmatically.

model Project {
  name String
}

import { Model } from '@mrleebo/prisma-ast'

// rename the datasource programmatically
builder.model('Project').then<Model>((model) => {
  model.name = "Task"
})

model Task {
  name String
}

If the entered model name already exists, that model will be used as the subject for any field and attribute calls that follow.

model Project {
  name        String
}

builder.model('Project').field('projectCode', 'String').attribute('unique')

model Project {
  name        String
  projectCode String @unique
}

If the field already exists, you can add new attributes to it by making calls to .attribute().

model Project {
  name        String
  projectCode String @unique
}

builder.model('Project').field('name').attribute('unique')

model Project {
  name        String @unique
  projectCode String @unique
}

You can remove an existing field with .removeField().

model Project {
  name        String
  projectCode String @unique
}

builder.model('Project').removeField('projectCode')

model Project {
  name String
}

You can remove an attribute from a field with .removeAttribute().

model Project {
  name        String
  projectCode String @unique
}

builder.model('Project').field('projectCode').removeAttribute('unique')

model Project {
  name        String
  projectCode String
}

If you want to perform a custom action that there isn't a Builder method for, you can access the underlying schema object programmatically.

model TaskMessage {
  createdAt DateTime? @db.Timestamptz(6)
}

import { Field, Attribute } from '@mrleebo/prisma-ast'

// Replace the @db.Timestamptz(6) attribute with @default(now())
builder
  .model('TaskMessage')
  .field('createdAt')
  .then<Field>(field => {
    const attribute: Attribute = {
      type: 'attribute',
      kind: 'field',
      name: 'default',
      args: [{ type: 'attributeArgument', value: 'now()' }],
    };
    field.attributes = [attribute];
  });

model TaskMessage {
  createdAt DateTime? @default(now())
}

model Project {
  name        String
  projectCode String @unique
}

builder.model('Project').blockAttribute('index', ['name'])

model Project {
  name        String
  projectCode String @unique
  @@index([name])
}

builder.enum('Role', ['USER', 'ADMIN'])

enum Role {
  USER
  ADMIN
}

Additional enumerators can also be added to an existing Enum

builder.enum('Role').break().comment("New role added for feature #12").enumerator('ORGANIZATION')

enum Role {
  USER
  ADMIN

  // New role added for feature #12
  ORGANIZATION
}

builder
  .model("Project")
  .break()
  .comment("I wish I could add a color to your rainbow")

model Project {
  name        String
  projectCode String @unique
  @@index([name])

  // I wish I could add a color to your rainbow
}

The produceSchema and createPrismaSchemaBuilder functions are intended to be your interface for interacting with the prisma schema, but you can also get direct access to the AST representation if you need to edit the schema for more advanced usages that aren't covered by the methods above.

The shape of the AST is not fully documented, and it is more likely to change than the builder API.

import { getSchema } from '@mrleebo/prisma-ast'

const source = `
model User {
  id   Int    @id @default(autoincrement())
  name String @unique
}
`

const schema = getSchema(source)

This is what builder.print() calls internally, and is what you'd use to print if you called getSchema().

import { printSchema } from '@mrleebo/prisma-ast'

const source = printSchema(schema)

You can optionally re-sort the schema. The default sort order is ['generator', 'datasource', 'model', 'enum'], and objects with the same type are sorted alphabetically, but the sort order can be overridden.

const source = printSchema(schema, { sort: true, locales: "en-US", sortOrder: ["datasource", "generator", "model", "enum"] })