JingsongLi opened a new pull request, #8669:
URL: https://github.com/apache/paimon/pull/8669

   ## What changed
   
   - pass the bucket deletion-vector factory through the write and primary-key 
index maintainer chain
   - snapshot deletion vectors when an asynchronous ANN build is scheduled
   - exclude snapshot-deleted physical row positions while preserving 
source-relative row ids
   - publish a zero-row marker segment when every source row is excluded
   - keep query-time deletion-vector filtering for deletions created after the 
build snapshot
   
   ## Why
   
   Primary-key ANN construction previously read compact source files without 
considering deletion
   vectors. Deleted rows were still stored in newly built indexes, so 
query-time filtering preserved
   correctness but left avoidable holes in every rebuild.
   
   ## Impact
   
   New and rebuilt primary-key vector indexes no longer contain rows already 
deleted at scheduling
   time. Existing published indexes remain valid, and later deletions continue 
to be filtered during
   search until normal compaction replaces their source files.
   
   ## Validation
   
   - 39 focused core tests covering ANN maintenance, empty segments, writer 
wiring, and vector search
   - `mvn -pl paimon-core -DskipTests compile`
   - Checkstyle, Spotless, Maven Enforcer, and `git diff --check`
   


-- 
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