Python, ADBC (Arrow Database Connectivity)

ADBC is a database connectivity standard on Apache Arrow. The adbc-driver-flightsql package connects to any Arrow Flight SQL server, also to Beacon. It returns the results as Arrow record batches, with zero copy. This fits data science work.

Install

pip install adbc-driver-flightsql adbc-driver-manager pyarrow

Connect

The Arrow Flight SQL server of Beacon listens on port 32011 by default. The HTTP API uses port 5001. Make that port reachable. See Expose the Flight SQL port below.

DBAPI 2.0 interface (recommended)

The adbc_driver_flightsql.dbapi module gives a standard PEP 249 interface. It works with pandas.read_sql and similar helpers.

import adbc_driver_flightsql.dbapi as flight_sql

conn = flight_sql.connect(
    "grpc://localhost:32011",
    db_kwargs={
        "username": "admin",        # BEACON_ADMIN_USERNAME
        "password": "securepassword", # BEACON_ADMIN_PASSWORD
    },
)

Low-level AdbcDatabase interface

Use adbc_driver_manager.AdbcDatabase for direct control over the connection. It also helps with a library that takes an ADBC database handle.

import adbc_driver_manager as mgr
import adbc_driver_flightsql as flightsql

db = mgr.AdbcDatabase(
    driver=flightsql.DRIVER_PATH,
    uri="grpc://localhost:32011",
    **{flightsql.DatabaseOptions.USERNAME.value: "admin"},
    **{flightsql.DatabaseOptions.PASSWORD.value: "securepassword"},
)
conn = db.connect()

Run queries

Fetch rows with a cursor

with flight_sql.connect("grpc://localhost:32011", db_kwargs={...}) as conn:
    with conn.cursor() as cur:
        cur.execute("SELECT * FROM default LIMIT 10")
        rows = cur.fetchall()
        print(rows)

Fetch as an Arrow table

with flight_sql.connect("grpc://localhost:32011", db_kwargs={...}) as conn:
    with conn.cursor() as cur:
        cur.execute("SELECT time, latitude, longitude, temp FROM default LIMIT 10000")
        arrow_table = cur.fetch_arrow_table()  # pyarrow.Table, zero-copy
        print(arrow_table.schema)

Read into a pandas DataFrame

import pandas as pd

with flight_sql.connect("grpc://localhost:32011", db_kwargs={...}) as conn:
    df = pd.read_sql("SELECT * FROM default LIMIT 1000", conn)
    print(df.head())

Expose the Flight SQL port

With Docker Compose, publish port 32011 next to the HTTP API:

services:
    beacon:
        image: ghcr.io/maris-development/beacon:v2.0.0
        ports:
            - "5001:5001"   # HTTP API
            - "32011:32011" # Arrow Flight SQL  #

TLS connections

Does your Beacon server use TLS? Then use the grpc+tls:// URI scheme:

conn = flight_sql.connect(
    "grpc+tls://your-beacon-host:32011",
    db_kwargs={
        "username": "admin",
        "password": "securepassword",
    },
)

For a self-signed certificate, switch the certificate check off:

import adbc_driver_flightsql as flightsql

conn = flight_sql.connect(
    "grpc+tls://your-beacon-host:32011",
    db_kwargs={
        "username": "admin",
        "password": "securepassword",
        flightsql.DatabaseOptions.TLS_SKIP_VERIFY.value: "true",
    },
)

Configuration reference

Variable	Default	Description
BEACON_FLIGHT_SQL_ENABLE	true	Switch the Flight SQL server on or off
BEACON_FLIGHT_SQL_HOST	0.0.0.0	The IP address of the listener
BEACON_FLIGHT_SQL_PORT	32011	The port of the Flight SQL gRPC server
BEACON_FLIGHT_SQL_ALLOW_ANONYMOUS	false	Allow a connection without credentials
BEACON_FLIGHT_SQL_TOKEN_TTL_SECS	3600	The lifetime of an auth token, in seconds