GitHub user cshuo edited a comment on the discussion: Spark query and write serving endpoint for hudi-lakehouse
@vinothchandar, I tested several OSS BI tools against **Spark Connect 4.1.3 + Hudi 1.2.0**, using [Foundry ADBC 0.2.1](https://adbc-drivers.org/drivers/spark/) and the [official Spark Connect JDBC driver](https://github.com/apache/spark/tree/v4.1.3/sql/connect/client/jdbc): | Tool | ADBC | JDBC | |---|---|---| | **[Superset](https://github.com/apache/superset)** (Star 74,760) | ✅ Works with a custom SQLAlchemy dialect | ✅ Works with the dialect plus a custom Statement-only Python/JVM bridge | | **[Redash](https://github.com/getredash/redash)** (Star 28,794) | ✅ Works through its built-in Python runner | ✅ Works through the Python runner using JPype + `Statement` | | **[Metabase](https://github.com/metabase/metabase)** (Star 49,255) | ❌ No usable integration entry point identified | ❌ Blocked by unsupported `prepareStatement`, even with a routing plugin | | **[Zeppelin](https://github.com/apache/zeppelin)** (Star 6,656) | ✅ Works through `%python` | ❌ Native `%jdbc` blocked by unsupported `setFetchSize` | These are **working integrations, not out-of-the-box Connect connectors**. No upstream driver binaries were modified. For Superset and Redash, validation covered metadata discovery, aggregation, region filtering, date grouping, and rendered dashboards. INSERT/UPDATE/DELETE and read-after-write also passed: * Superset + ADBC: all three operations passed directly in the SQL Lab UI with **Allow DML** enabled. * Redash executed writes inside its Python runner; * Zeppelin’s ADBC path passed aggregation and CRUD with notebook result rendering. I'd suggest **Redash + ADBC for the simplest initial dashboard example**, or **Superset + ADBC for a fuller BI integration** if we’re comfortable maintaining the SQLAlchemy adapter. Both provide an end-to-end path without Thrift Server. **Improving JDBC compatibility—particularly `prepareStatement` and `setFetchSize` support—could be follow-up work** to enable broader BI tool integration. These were local functional tests; production concerns such as authentication, concurrency, cancellation, and large results still need validation. See https://docs.google.com/document/d/1JwCbTByB7vUC02lcQ2HtefraPSFvO8RGYZd9VMn7XJo/edit?usp=sharing for more test details. GitHub link: https://github.com/apache/hudi/discussions/19843#discussioncomment-18442667 ---- This is an automatically sent email for [email protected]. To unsubscribe, please send an email to: [email protected]
