jiangxt2 opened a new issue, #13323: URL: https://github.com/apache/gravitino/issues/13323
## What would you like to be improved? Gravitino documents integration with Ray Data through the Iceberg REST endpoint and `catalog_kwargs`, but the Python client integration suite does not currently provide a real end-to-end regression test for this contract. This leaves potential regressions in the following areas undetected: - Iceberg REST auxiliary service configuration; - PyIceberg/Ray catalog loading; - reading an existing Iceberg table through Ray Data; - writing to an existing Iceberg table through Ray Data; - Gravitino metadata and temporary warehouse cleanup. The existing Lance Ray integration test provides a useful lifecycle baseline, but there is no equivalent Iceberg REST test. ## How should we improve? Add a Gravitino-owned Python integration test under `clients/client-python/tests/integration/` with an independent Gradle entry point. The test should: - use Ray 2.55.1 and a compatible PyIceberg version; - start Gravitino with the Iceberg REST auxiliary service; - configure the dynamic Iceberg catalog provider; - use PyIceberg to create the namespace and existing Iceberg table fixture; - verify an empty table can be read by `ray.data.read_iceberg`; - append data with `Dataset.write_iceberg`; - read the data again through Ray Data and verify row counts, field order, and filtering; - clean up the Iceberg table, Gravitino catalog, metalake, temporary warehouse, and configuration; - run in a dedicated Python environment without changing the default client-python test dependencies. The implementation should remain limited to the Gravitino Python client integration test and should not modify Java Iceberg REST code, Ray source code, or introduce Ray `Catalog.resolve()` support. Related documentation: - [Connect Ray to Iceberg REST](https://github.com/apache/gravitino/blob/main/docs/iceberg-rest-engine/ray.md) - [Iceberg REST Catalog Service](https://github.com/apache/gravitino/blob/main/docs/iceberg-rest-service.md) -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
