Getting started

A family of pure-native, high-performance ADBC drivers for SQL Server, PostgreSQL and Oracle. Each one speaks its database's wire protocol directly — the bytes the server itself expects, rather than a call into ODBC or a vendor client library.

Apache Arrow RecordBatches travel both ways: out of the database on the read path, and back into it on the ingest path. Extraction and loading go through the same driver, with no conversion to rows in between.

Which driver do I need?

Your database	Driver	Load name	Also covers	Status
Microsoft SQL Server	ArrowTDS	arrowtds	Azure SQL, Microsoft Fabric	✅ Published
PostgreSQL	ArrowFEBE	arrowfebe	PostgreSQL-compatible services	✅ Published
Oracle	ArrowTTC	arrowttc	Oracle Cloud Autonomous Database, OCI IAM	✅ Published
IBM Db2	ArrowDRDA	arrowdrda	—	🚧 Coming soon

The driver binaries are published as public GitHub Releases and are free to download. They are licence-gated: a driver needs a valid Arpeio licence at runtime — there is no trial build. Contact sales@arpe.io for a licence.

The drivers

Microsoft SQL ServerArrowTDSNative MS-TDS 7.4 over TCP/TLS. Azure SQL, Microsoft Fabric, SQL & Entra ID auth, geospatial & VECTOR types.v0.5.25 Linux · Windows Azure SQL · FabricPostgreSQLArrowFEBENative Frontend/Backend v3 over TCP/TLS — no libpq. NUMERIC → decimal with no string detour, SCRAM & Kerberos auth.v0.3.8 Linux · Windows PG-compatibleOracleArrowTTCNative TNS + TTC over TCP/TLS — no OCI, no Instant Client. Oracle Cloud Autonomous Database, OCI IAM & Entra ID tokens.v0.2.8 Linux · Windows Oracle Cloud ADB

Install in one line

Linux / macOS

curl -fsSL https://raw.githubusercontent.com/arpe-io/adbc-drivers/main/install.sh \
  | sh -s -- arrowtds --license /path/to/your.lic

Windows (PowerShell)

& ([scriptblock]::Create((irm https://raw.githubusercontent.com/arpe-io/adbc-drivers/main/install.ps1))) `
  arrowtds -License C:\path\to\your.lic

Then load it by name from any ADBC client:

Python

import adbc_driver_manager.dbapi as dbapi
with dbapi.connect(driver="arrowtds",
                   db_kwargs={"uri": "sqlserver://sa:<pw>@host:1433/?database=db&encrypt=true"}) as conn:
    cur = conn.cursor()
    cur.execute("SELECT * FROM dbo.orders")
    table = cur.fetch_arrow_table()   # ready as Arrow

C / C++

#include <arrow-adbc/adbc.h>
// (error checks elided for brevity)
AdbcError e = {}; AdbcDatabase db = {}; AdbcConnection conn = {}; AdbcStatement st = {};
AdbcDatabaseNew(&db, &e);
AdbcDatabaseSetOption(&db, "driver", "arrowtds", &e);
AdbcDatabaseSetOption(&db, "uri", "sqlserver://sa:<pw>@host:1433/?database=db&encrypt=true", &e);
AdbcDatabaseInit(&db, &e);
AdbcConnectionNew(&conn, &e); AdbcConnectionInit(&conn, &db, &e);
AdbcStatementNew(&conn, &st, &e);
AdbcStatementSetSqlQuery(&st, "SELECT * FROM dbo.orders", &e);
ArrowArrayStream stream = {}; int64_t n = -1;
AdbcStatementExecuteQuery(&st, &stream, &n, &e);   // consume `stream`

C#

using Apache.Arrow.Adbc;
using Apache.Arrow.Adbc.DriverManager;

using var driver = AdbcDriverManager.FindLoadDriver("arrowtds", loadOptions: AdbcLoadFlags.Default);
using var database = driver.Open(new Dictionary<string, string> {
    ["uri"] = "sqlserver://sa:<pw>@host:1433/?database=db&encrypt=true" });
using var connection = database.Connect(null);
using var statement = connection.CreateStatement();
statement.SqlQuery = "SELECT * FROM dbo.orders";
using var stream = statement.ExecuteQuery().Stream!;   // IArrowArrayStream

Go

var drv drivermgr.Driver
db, _ := drv.NewDatabase(map[string]string{
	"driver":          "arrowtds",
	adbc.OptionKeyURI: "sqlserver://sa:<pw>@host:1433/?database=db&encrypt=true",
})
conn, _ := db.Open(ctx)
stmt, _ := conn.NewStatement()
_ = stmt.SetSqlQuery("SELECT * FROM dbo.orders")
reader, _, _ := stmt.ExecuteQuery(ctx)   // array.RecordReader
defer reader.Release()

Java

Map<String, Object> params = new HashMap<>();
JniDriver.PARAM_DRIVER.set(params, "arrowtds");
AdbcDriver.PARAM_URI.set(params, "sqlserver://sa:<pw>@host:1433/?database=db&encrypt=true");
try (BufferAllocator a = new RootAllocator();
     AdbcDatabase db = new JniDriver(a).open(params);
     AdbcConnection conn = db.connect();
     AdbcStatement st = conn.createStatement()) {
  st.setSqlQuery("SELECT * FROM dbo.orders");
  try (AdbcStatement.QueryResult r = st.executeQuery()) {
    ArrowReader reader = r.getReader();
  }
}

R

library(adbcdrivermanager)
db  <- adbc_database_init(adbc_driver("arrowtds"),
                          uri = "sqlserver://sa:<pw>@host:1433/?database=db&encrypt=true")
con <- adbc_connection_init(db)
table <- read_adbc(con, "SELECT * FROM dbo.orders") |> arrow::as_arrow_table()

Rust

use adbc_core::{Connection, Database, Driver, Statement, LOAD_FLAG_DEFAULT};
use adbc_core::options::{AdbcVersion, OptionDatabase, OptionValue};
use adbc_driver_manager::ManagedDriver;

let mut driver = ManagedDriver::load_from_name("arrowtds", None, AdbcVersion::default(), LOAD_FLAG_DEFAULT, None)?;
let mut db = driver.new_database_with_opts([(OptionDatabase::Uri,
    OptionValue::String("sqlserver://sa:<pw>@host:1433/?database=db&encrypt=true".into()))])?;
let mut conn = db.new_connection()?;
let mut stmt = conn.new_statement()?;
stmt.set_sql_query("SELECT * FROM dbo.orders")?;
let reader = stmt.execute()?;   // impl RecordBatchReader

Swap arrowtds for arrowfebe or arrowttc. Full options, licence handling and troubleshooting are on the Installation page.

note

Platforms: the drivers ship for Linux (x64) and Windows (x64). macOS binaries are staged but not yet published.

Find your way around

InstallationThe one-line installer, options, licence handling, download-only, and troubleshooting.VersioningHow the drivers are versioned, current versions, and how docs track releases.