Arrow Flight Client

Unified Arrow Flight and Flight SQL client for JavaScript and TypeScript runtimes. Provides complete gRPC client support for Apache Arrow Flight and Flight SQL protocols.

Features

• Flight RPC — All core methods: listFlights, getFlightInfo, doGet, doPut, doAction, etc.
• Flight SQL — SQL queries, prepared statements, transactions, and database metadata
• Authentication — Bearer tokens, mTLS, Flight Handshake (BasicAuth), and rotating credentials
• Arrow IPC — Encode/decode Apache Arrow data for streaming transfers
• Multi-runtime — Works with Bun, Node.js 20+, and Deno

Installation

bun add @qualithm/arrow-flight-client
# or
npm install @qualithm/arrow-flight-client

Quick Start

import { createFlightClient, decodeFlightDataToTable } from "@qualithm/arrow-flight-client"

const client = createFlightClient({
  url: "http://localhost:50051",
  auth: { type: "bearer", token: "your-token" }
})

// List available flights
for await (const info of client.listFlights()) {
  console.log(info.flightDescriptor?.path)
}

// Fetch data
const flightInfo = await client.getFlightInfo({ type: "path", path: ["my", "dataset"] })
const ticket = flightInfo.endpoint[0]?.ticket
if (ticket === undefined) throw new Error("flight has no ticket")
const table = await decodeFlightDataToTable(client.doGet(ticket))

console.log(`Received ${table.numRows} rows`)

client.close()

Usage

FlightSqlClient (SQL Operations)

import { createFlightSqlClient } from "@qualithm/arrow-flight-client"

const client = createFlightSqlClient({
  url: "http://localhost:50051",
  auth: { type: "bearer", token: "your-token" }
})

// Execute a query
const table = await client.query("SELECT * FROM users LIMIT 10")
console.log(`Got ${table.numRows} rows`)

// Stream large results
for await (const batch of client.queryBatches("SELECT * FROM large_table")) {
  console.log(`Batch: ${batch.numRows} rows`)
}

// Prepared statements
const stmt = await client.prepare("SELECT * FROM users WHERE id = ?")
const result = await client.executePrepared(stmt)
await client.closePreparedStatement(stmt)

// Transactions
const txn = await client.beginTransaction()
try {
  await client.executeUpdate("INSERT INTO users (name) VALUES ('Alice')", {
    transactionId: txn.id
  })
  await client.commit(txn)
} catch {
  await client.rollback(txn)
}

// Database metadata
const catalogs = await client.getCatalogs()
const schemas = await client.getDbSchemas({})
const tables = await client.getTables({})

client.close()

Authentication

Bearer Token

const client = createFlightClient({
  url: "https://flight-server:50051",
  auth: { type: "bearer", token: "your-api-token" }
})

Basic Auth (Flight Handshake)

const client = createFlightClient({
  url: "https://flight-server:50051",
  auth: {
    type: "basic",
    credentials: { username: "user", password: "pass" }
  }
})

// Authenticate (performs Flight Handshake RPC)
await client.authenticate()

Rotating Credentials

For credentials that expire or rotate, supply an authProvider instead of a static auth. The client resolves credentials through it before the first request and again whenever the server rejects a call as unauthenticated, so a long-lived client picks up a rotation without being reconstructed.

const client = createFlightClient({
  url: "https://flight-server:50051",
  authProvider: async () => ({
    type: "bearer",
    token: await fetchFreshToken()
  })
})

The resolved credential is cached, so the provider is not consulted per request. A rejected call is retried once with fresh credentials; if it is rejected again the error is surfaced. Concurrent rejections share a single refresh.

Two calls are not retried: doPut, because its input stream has already been partly consumed and cannot be replayed, and any server stream that had already delivered a message before the rejection, which would otherwise repeat its opening messages. Both surface FlightAuthError for the caller to retry.

mTLS

import { readFileSync } from "node:fs"

const client = createFlightClient({
  url: "https://flight-server:50051",
  tls: {
    ca: readFileSync("ca.pem"),
    cert: readFileSync("client.pem"),
    key: readFileSync("client-key.pem")
  }
})

Arrow IPC Utilities

import {
  encodeTableToFlightData,
  decodeFlightDataToTable,
  decodeFlightDataStream
} from "@qualithm/arrow-flight-client"
import { tableFromArrays } from "apache-arrow"

// Create Arrow data
const table = tableFromArrays({
  id: Int32Array.from([1, 2, 3]),
  name: ["Alice", "Bob", "Charlie"]
})

// Encode for upload
const flightData = encodeTableToFlightData(table)

// Decode a stream
const receivedTable = await decodeFlightDataToTable(flightData)

// Stream batches
for await (const batch of decodeFlightDataStream(encodeTableToFlightData(table))) {
  console.log(`Batch: ${batch.numRows} rows`)
}

This is the beginning of the README. Read the rest at the original. Read the full README ↗