foxtail463 opened a new pull request, #67426: URL: https://github.com/apache/doris/pull/67426
Problem Summary: When column statistics were missing or unknown, RuntimeFilterPruner could remove runtime filters that were still able to eliminate entire partitions or hash buckets. Remote TYPE_DATE IN filters were also deserialized as DATETIME values, producing a distribution hash different from the original DATE value and potentially selecting incorrect tablets. Solution: Introduce a unified scan-range pruning capability covering both partition and bucket pruning. When row-level selectivity cannot be estimated, retain joins containing runtime filters with this capability; decisions based on valid statistics remain unchanged. Deserialize TYPE_DATE exact-set values with DATE semantics so their type and distribution hash survive remote runtime-filter serialization. Enable runtime-filter pruning in the existing bucket-pruning regression case so it exercises the real optimizer path. -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
