kysely-zod-codegen v0.17.7

kysely-zod-codegen generates Kysely type definitions and zod schemas from your database.

This project is based on kysely-codegen. It can be used the same way kysely-codegen is used and it provides one additional argument:

--generate-zod-schemas which is used to generate zod schemas instead of typescript types.

Table of contents

• Installation
• Generating type definitions
• Using the type definitions
• CLI arguments
• Issue funding

Installation

npm install --save-dev kysely-zod-codegen

You will also need to install Kysely with your driver of choice:

# PostgreSQL
npm install kysely pg

# MySQL
npm install kysely mysql2

# SQLite
npm install kysely better-sqlite3

# MSSQL
npm install kysely tedious tarn @tediousjs/connection-string

# LibSQL
npm install @libsql/kysely-libsql

Generating type definitions

The most convenient way to get started is to create an .env file with your database connection string:

# PostgreSQL
DATABASE_URL=postgres://username:password@yourdomain.com/database

# MySQL
DATABASE_URL=mysql://username:password@yourdomain.com/database

# SQLite
DATABASE_URL=C:/Program Files/sqlite3/db

# MSSQL
DATABASE_URL=Server=mssql;Database=database;User Id=user;Password=password

# LibSQL
DATABASE_URL=libsql://token@host:port/database

If your URL contains a password with special characters, those characters may need to be percent-encoded.

If you are using PlanetScale, make sure your URL contains this SSL query string parameter: ssl={"rejectUnauthorized":true}

Then run the following command, or add it to the scripts section in your package.json file:

kysely-zod-codegen

This command will generate a .d.ts file from your database, for example:

import { ColumnType } from 'kysely';

export type Generated<T> = T extends ColumnType<infer S, infer I, infer U>
  ? ColumnType<S, I | undefined, U>
  : ColumnType<T, T | undefined, T>;

export type Timestamp = ColumnType<Date, Date | string, Date | string>;

export interface Company {
  id: Generated<number>;
  name: string;
}

export interface User {
  company_id: number | null;
  created_at: Generated<Timestamp>;
  email: string;
  id: Generated<number>;
  is_active: boolean;
  name: string;
  updated_at: Timestamp;
}

export interface DB {
  company: Company;
  user: User;
}

To specify a different output file:

kysely-zod-codegen --out-file ./src/db/db.d.ts

Using the type definitions

Import DB into new Kysely<DB>, and you're done!

import { Kysely, PostgresDialect } from 'kysely';
import { DB } from 'kysely-zod-codegen';
import { Pool } from 'pg';

const db = new Kysely<DB>({
  dialect: new PostgresDialect({
    pool: new Pool({
      connectionString: process.env.DATABASE_URL,
    }),
  }),
});

const rows = await db.selectFrom('users').selectAll().execute();
//    ^ { created_at: Date; email: string; id: number; ... }[]

If you need to use the generated types in e.g. function parameters and type definitions, you may need to use the Kysely Insertable, Selectable, Updateable types. Note that you don't need to explicitly annotate query return values, as it's recommended to let Kysely infer the types for you.

import { Insertable, Updateable } from 'kysely';
import { DB } from 'kysely-zod-codegen';
import { db } from './db';

async function insertUser(user: Insertable<User>) {
  return await db
    .insertInto('users')
    .values(user)
    .returningAll()
    .executeTakeFirstOrThrow();
  // ^ Selectable<User>
}

async function updateUser(user: Updateable<User>) {
  return await db
    .updateTable('users')
    .set(user)
    .where({ id: user.id })
    .returning(['email', 'id'])
    .executeTakeFirstOrThrow();
  // ^ { email: string; id: number; }
}

Read the Kysely documentation for more information.

CLI arguments

--generate-zod-schemas

Generates zod schemas instead of typescript types.

Example:

export const profilesTableSchema = z.object({
  id: z.string().uuid(),
  avatar_url: z.string().nullable(),
  updated_at: timestampSchema.nullable(),
  username: z.string().nullable(),
  website: z.string().nullable(),
});

--camel-case

Use the Kysely CamelCasePlugin for generated table column names.

Example:

export interface User {
  companyId: number | null;
  createdAt: Generated<Timestamp>;
  email: string;
  id: Generated<number>;
  isActive: boolean;
  name: string;
  updatedAt: Timestamp;
}

--date-parser

Specify which parser to use for PostgreSQL date values. (values: string, timestamp, default: timestamp)

--dialect value

Set the SQL dialect. (values: postgres, mysql, sqlite, mssql, libsql, bun-sqlite, kysely-bun-sqlite, worker-bun-sqlite)

--env-file value

Specify the path to an environment file to use.

--help, -h

Print all command line options.

--include-pattern value, --exclude-pattern value

You can choose which tables should be included during code generation by providing a glob pattern to the --include-pattern and --exclude-pattern flags. We use micromatch under the hood, which provides advanced glob support. For instance, if you only want to include your public tables:

kysely-zod-codegen --include-pattern="public.*"

You can also include only certain tables within a schema:

kysely-zod-codegen --include-pattern="public.+(user|post)"

Or exclude an entire class of tables:

kysely-zod-codegen --exclude-pattern="documents.*"

--log-level value

Set the terminal log level. (values: debug, info, warn, error, silent, default: warn)

--no-domains

Skip generating types for PostgreSQL domains. (default: false)

--numeric-parser

Specify which parser to use for PostgreSQL numeric values. (values: string, number, number-or-string, default: string)

--overrides

Specify type overrides for specific table columns in JSON format.

Example:

kysely-zod-codegen --overrides='{"columns":{"table_name.column_name":"{foo:\"bar\"}"}}'

--out-file value

Set the file build path. (default: ./node_modules/kysely-zod-codegen/dist/db.d.ts)

--partitions

Include partitions of PostgreSQL tables in the generated code.

--print

Print the generated output to the terminal instead of a file.

--runtime-enums, --runtime-enums-style

The PostgreSQL --runtime-enums option generates runtime enums instead of string unions.

The option --runtime-enums-style specifies which naming convention to use for runtime enum keys. (values: pascal-case, screaming-snake-case, default: pascal-case)

Examples:

--runtime-enums=false

export type Status = 'CONFIRMED' | 'UNCONFIRMED';

--runtime-enums

export enum Status {
  CONFIRMED = 'CONFIRMED',
  UNCONFIRMED = 'UNCONFIRMED',
}

--runtime-enums --runtime-enums-style=pascal-case

export enum Status {
  Confirmed = 'CONFIRMED',
  Unconfirmed = 'UNCONFIRMED',
}

--schema value

Set the default schema(s) for the database connection.

Multiple schemas can be specified:

kysely-zod-codegen --schema=public --schema=hidden

--singular

Singularize generated table names, e.g. BlogPost instead of BlogPosts. We use the pluralize package for singularization.

--type-only-imports

Generate code using the TypeScript 3.8+ import type syntax. (default: true)

--url value

Set the database connection string URL. This may point to an environment variable. (default: env(DATABASE_URL))

--verify

Verify that the generated types are up-to-date. (default: false)

Issue funding

We use Polar.sh to upvote and promote specific features that you would like to see implemented. Check the backlog and help out: