jiwen624 commented on PR #18806: URL: https://github.com/apache/iceberg/pull/18806#issuecomment-6702586638
Hi @mxm @pvary, could you review this? It documents #18748 and a workaround validated through stop-with-savepoint → spec change → rewrite → rebuild and restore, on main and a non-main branch. The prototype fix in #18747 writes deletes across historical specs, but requires a serializer-version change. In addition, it only supports partitioning based on equality fields and still misses Dynamic Sink changes within one checkpoint. Given that complexity, limited coverage, and v4’s prohibition on new equality-delete files, I’m leaning toward landing these docs rather than pursuing #18747. As a follow-up, I’d suggest a startup check that rejects upsert/CDC writers using a partitioned spec when the target branch contains data files under another spec. That would turn existing mismatches into an explicit error instead of silent incorrect results. Runtime spec changes would need separate handling. Does that direction make sense? -- 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]
