zhangyue19921010 opened a new pull request, #11151:
URL: https://github.com/apache/paimon/pull/11151

   ### Purpose
   
   Fixes #11150.
   
   When a Java batch commit collects normal-only files before files with 
vectors, the vector row-ID cursor still points at the earlier rows. Reading IDs 
0–3 can then return vectors `[22, 33, NULL, NULL]` instead of `[NULL, NULL, 22, 
33]`.
   
   Advance the vector cursor to each new normal file's start when it lags 
behind, keeping dedicated vector files aligned with their rows without 
rewinding an already-ahead copied-file cursor.
   
   ### Tests
   
   Added a table regression using two writers and one commit, with no manually 
constructed metadata. It covers JSON and Parquet vector files and normal-file 
rolling. An existing copied-file fixture also verifies that an already-ahead 
vector cursor is preserved. Both table cases fail on unchanged production at 
`57233dbc1` with vectors attached to the wrong IDs and pass with the change.
   
   55 related tests passed on JDK 11, including dedicated-file rolling, vector 
updates/compaction and copied-file row tracking. Checkstyle, Spotless, Enforcer 
and RAT passed without `fast-build`. The Maven `verify` lifecycle ran the 
selected suite twice.
   
   ```sh
   mvn -pl paimon-core -am -DwildcardSuites=none -DfailIfNoTests=false \
     
'-Dtest=VectorStoreTableTest,DedicatedFormatRollingFileWriterVectorTest,RowTrackingCopiedFilesTest,CopiedDataFilesTest,DataEvolutionTableTest#testPartitionGroupedRowIdsWithBlobFiles+testDataEvolutionWriteRollsByRows+testCompactPreservesConcurrentPartialUpdateWithinCandidateRange'
 verify
   ```
   
   AI-generated with OpenAI Codex; independently reviewed by an AI agent. No 
human review is claimed.
   


-- 
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