XiaoHongbo-Hope opened a new pull request, #9031:
URL: https://github.com/apache/paimon/pull/9031

   ### Purpose
   
   Follow-up to #8930 and 
https://github.com/apache/paimon/pull/8930#discussion_r3712090800.
   
   Full-text scans select index files at one snapshot, but data-evolution reads 
previously evaluated deletion-vector live rows and raw fallback data against 
the latest snapshot. A materialized-deletion compaction between planning and 
reading could therefore combine old index row IDs with rewritten row IDs and 
return incomplete or incorrect top-K results.
   
   This change carries the planned snapshot through FullTextScan.Plan, uses it 
for live-row filtering and raw fallback reads, and forwards the snapshot 
already pinned by hybrid search into data-evolution full-text scans. Raw 
fallback uses copyWithoutTimeTravel so the resolved read schema is preserved.
   
   ### Tests
   
   - FullTextSearchBuilderTest, including plan -> DV commit -> read
   - FullTextSearchBuilderTest, including plan -> materialized-deletion 
compaction -> raw fallback read
   - PrimaryKeyFullTextReadTest
   - PrimaryKeyFullTextScanTest
   
   All 34 tests passed with checkstyle, Spotless, and Maven enforcer enabled.


-- 
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]

Reply via email to