JingsongLi opened a new pull request, #11330:
URL: https://github.com/apache/paimon/pull/11330

   ### Purpose
   
   After an indexed Data Evolution column is widened from INT to BIGINT, the 
old BTree still contains INT-encoded keys. Decoding those keys with the new 
type can silently prune matching rows and makes the same index unsafe for Top-N.
   
   Check BTree key compatibility before planning eager or reader-side filters, 
search prefilters, and Top-N. Reuse the build-schema filename convention 
already understood by Paimon Rust. For a legacy filename, use the index only 
when the available schema history proves the indexed field type stayed 
compatible.
   
   An incremental rebuild retires incompatible files in the same commit as 
their replacements, so the new index can serve the latest schema while the 
older snapshot remains readable.
   
   ### Tests
   
   - The baseline returned no rows for `f0 = 7` after widening the indexed 
column. The regression verifies eager and reader-side filtering, Top-N 
fallback, historical index reads, and successful incremental replacement.
   - Added compatibility checks for legacy and schema-tagged filenames.
   - JDK 8 Maven `verify` passed 195 focused tests.
   


-- 
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