zhangshenghang commented on issue #10227: URL: https://github.com/apache/seatunnel/issues/10227#issuecomment-3742242750
<!-- seatunnel-pr-reviewer --> Thanks for the interest in Hudi Source support. The original concern was that Hudi's source/snapshot APIs depend on compute engines (Spark/Flink). Recent Hudi releases have improved with Hudi's Java client and more decoupled table services, but there are still challenges: 1. **Engine coupling**: The primary read APIs (`HoodieTableFileSystemView`) still pull in Spark/Hadoop dependencies transitively 2. **Incremental query support**: Requires accessing Hudi's timeline metadata, which may need Spark/Flink classes 3. **Format complexity**: Handling MOR tables, log formats, and compaction without an engine is non-trivial We'd welcome community contributions here. For anyone interested: - Could we leverage Hudi's `HoodieTableMetadata` or newer Java APIs without pulling in Spark/Flink? - Is batch-only mode (no incremental) acceptable as a first step? - Would you accept Hadoop/Hive dependencies as a baseline requirement? If you'd like to work on this, please propose your approach in a design doc or PR so we can review dependency constraints. -- 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]
