prashantwason commented on PR #19898: URL: https://github.com/apache/hudi/pull/19898#issuecomment-6024201953
Thanks @yihua, I traced it against `BaseHoodieTableFileIndex.listPartitionPaths` and your analysis is right. The "always" in the description was wrong. - **Latency claim:** the "minutes of latency" number was not measured in a specific scenario, so I've removed it. The description now names the two cases where `listFileSplits()` lists the whole table: (1) the start is `earliest` or older than the active timeline, and (2) a partition predicate with no bound leading prefix on a hive-style, URL-encoded table. It also says the gain is small when the start is inside the active timeline. The prune-rule skip is described as saving redundant work in every case, and as a full-table cost only in those two cases. - **Scaladoc:** in 7863548, `listMatchingPartitionPaths` now spells out when it lists the table, using the cases above. It points callers that already hold candidate partitions to `filterPartitionPaths`. I also reworded its "listing full table" log line, since the no-predicate path doesn't always do that. -- 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]
