DanielLeens commented on issue #12266: URL: https://github.com/apache/seatunnel/issues/12266#issuecomment-5855680632
Classification: I / Zeta file MapStore durability benchmark. The storage-first split is the right direction, but it is still design-only. The parent implementation #12173 is open, BLOCKED, and has a failed Build, so please do not stack a benchmark or storage PR on it yet. For Phase A, an additive keyed-load method with a default full-load-and-filter fallback can preserve SPI compatibility, but that fallback cannot satisfy the memory acceptance criterion. Therefore Phase B may use the keyed path only after the file implementation is explicitly proven selective; do not make the benchmark rely on a generic interface default or a heap estimate. The file implementation must preserve MapLoader semantics for requested keys across overwrites and tombstones, return no unrelated key, and retain no whole-map result. Cover that through FileMapStore delegation plus WAL reader/storage tests with many filler keys and selected updated/deleted keys; a constrained-heap diagnostic run is useful evidence, but not a substitute for the functional contract. Please split this into two reviewable changes after #12173 is healthy: first the keyed storage contract and its deterministic tests, then the benchmark widening outside the measured SingleShot window. The second change may claim full-batch durability only after the first is merged and demonstrated. No label or assignment changes here. -- 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]
