Hello Fluss Community, I propose initiating discussion on FIP-10: Support Log RecordBatch Filter Pushdown (FIP Page <https://github.com/platinumhamburg/fluss >). This optimization aims to improve the performance of Log table queries and is now ready for community feedback. Core Motivation Currently, filtering on non-PK/non-partition keys requires:
* Transferring full RecordBatches from storage, * Transmitting irrelevant records over the network, * Decompressing non-matching Arrow data. This results in unnecessary network/memory overhead, especially for low-selectivity queries. FIP-10 introduces RecordBatch-level filter pushdown to enable early filtering at the storage layer, reducing: * Network transfer by skipping non-matching batches, * Memory pressure via pre-deserialization filtering, * CPU cost from decompression of discarded data. Implementation Status A proof-of-concept (PoC) has been implemented in the logfilter branch <https://github.com/platinumhamburg/fluss > and is ready for testing and preview. Any feedback and suggestions on this proposal are welcome! Looking forward to your insights. Best regards, Yang Wang
