Other JDBC applications¶
All integration guides. Applications that accept a custom JDBC driver can try Flight SQL. The JDBC driver is validated; each application's metadata handling, generated SQL and write behavior still needs acceptance testing.
Configure a custom driver¶
- Start a server and create the sample SQL table.
- Run
make bin/flight-sql-jdbc-driver-19.0.0.jarfrom the repository root. Add that JAR to the application's driver libraries, or use Maven coordinateorg.apache.arrow:flight-sql-jdbc-driver:19.0.0. - Set the driver class to
org.apache.arrow.driver.jdbc.ArrowFlightJdbcDriver. - Set URL to
jdbc:arrow-flight-sql://localhost:8815/?useEncryption=falsefor the local unauthenticated endpoint. Use the actual host from the client. - Configure the application's JVM with
--add-opens=java.base/java.nio=ALL-UNNAMED. On Java 23+ also add--sun-misc-unsafe-memory-access=allow, then restart the application. - Use autocommit. Browse catalog
redis, schemapublic, and tableflight_tool_sample.
For TLS, credentials, session behavior and long-result timeouts, follow the JDBC connection reference and official driver options.
Check application behavior¶
Run these statements separately in the application's SQL editor:
SELECT id, label, amount FROM flight_tool_sample ORDER BY id;
SELECT COUNT(*) AS n, SUM(amount) AS total FROM flight_tool_sample;
Expect ids 1–3, a NULL label for id 2, and aggregate values 3 and 7.5.
Check table/column discovery and generated filters separately. If writes are
part of the workflow, insert, update and delete a disposable row through that
application and verify each operation with a SELECT. Fixture creation through
another client does not prove application write support.
Run make jdbc-smoke to distinguish a driver/server issue from an application
issue. The driver's empty getTypeInfo() can affect application type discovery;
see known caveats.
SQL tools see SQL tables; named datasets require a native Flight client.
Finish with the sample cleanup.
Write example with read-back¶
Use only the disposable shared sample and ensure id 99 is unused. Execute
one statement at a time in the application SQL editor. Expect a new row with amount 9.5,
then amount 10.5, then an empty result after deletion:
INSERT INTO flight_tool_sample (id, label, amount) VALUES (99, 'guide', 9.5);
SELECT * FROM flight_tool_sample WHERE id = 99;
UPDATE flight_tool_sample SET amount = 10.5 WHERE id = 99;
SELECT amount FROM flight_tool_sample WHERE id = 99;
DELETE FROM flight_tool_sample WHERE id = 99;
SELECT * FROM flight_tool_sample WHERE id = 99;
The live guide run passed these SQL statements through the Flight SQL JDBC driver. That driver-level result does not validate arbitrary application UIs; record each application separately. SQL uses autocommit.