snowflake-kysely-dialect

A Kysely dialect for Snowflake.

Installation

npm install snowflake-kysely-dialect kysely snowflake-sdk
# or
yarn add snowflake-kysely-dialect kysely snowflake-sdk

Usage

import { Kysely } from 'kysely'
import { SnowflakeDialect } from 'snowflake-kysely-dialect'

interface Database {
  users: {
    id: number
    name: string
    email: string
  }
}

const db = new Kysely<Database>({
  dialect: new SnowflakeDialect({
    connection: {
      account: 'your-account',
      username: 'your-username',
      password: 'your-password',
      database: 'your-database',
      schema: 'your-schema',
      warehouse: 'your-warehouse',
    },
  }),
})

// Select
const users = await db.selectFrom('users').selectAll().execute()

// Insert
await db.insertInto('users').values({ id: 1, name: 'Alice', email: '[email protected]' }).execute()

// Update
await db.updateTable('users').set({ name: 'Bob' }).where('id', '=', 1).execute()

// Delete
await db.deleteFrom('users').where('id', '=', 1).execute()

Transactions

Transactions are fully supported:

await db.transaction().execute(async (trx) => {
  await trx.insertInto('users').values({ id: 2, name: 'Carol', email: '[email protected]' }).execute()
  await trx.updateTable('users').set({ name: 'Dave' }).where('id', '=', 1).execute()
})

Streaming

const stream = await db.selectFrom('users').selectAll().stream()

for await (const row of stream) {
  console.log(row)
}

Introspection

const tables = await db.introspection.getTables()

Notes

• Uses PostgresQueryCompiler under the hood — Snowflake's SQL dialect (double-quoted identifiers, window functions, standard string functions) is closer to Postgres than MySQL.
• Uses a connection pool backed by snowflake-sdk's built-in pool. Pool size and timeouts are configurable via the poolOptions config key.
• RETURNING is not supported by Snowflake and is disabled in the adapter.
• Snowflake folds unquoted identifiers to uppercase. Use quoted identifiers (via sql.id(...)) when you need case-sensitive names.
• VARIANT / OBJECT columns (JSON), TIMESTAMP_NTZ / TIMESTAMP_LTZ / TIMESTAMP_TZ, and NUMBER precision are the most common sources of type round-tripping surprises — test these explicitly in your application.

License

MIT