Gabriel39 commented on PR #65851: URL: https://github.com/apache/doris/pull/65851#issuecomment-5111774199
Current-head review summary (`6c6ac29105d`) The recent updates address the earlier cross-partition equality-delete preflight, shared cached-table refresh, commit-replay validation, overwrite spec drift, DEFAULT handling, and metadata-only COUNT(*) concerns. I rechecked the current implementation and still see four merge-blocking/follow-up items: 1. **P1 — Preserve the ancestor tree for nested equality-delete keys.** The FE carrier and both BE lookup paths currently lose or fail to resolve nested field identity: https://github.com/apache/doris/pull/65851#discussion_r3669378588 2. **P1 — Convert legacy UUID text before writing ORC binary.** With `enable.mapping.varbinary=false`, explicit UUID values are written as 32/36-byte text into an Iceberg UUID ORC binary column; FIXED length also needs validation: https://github.com/apache/doris/pull/65851#discussion_r3669378596 3. **P2 — Scope the rolling-upgrade name-collision fence to resolved paths.** An unrelated, unprojected nested collision currently rejects the whole scan: https://github.com/apache/doris/pull/65851#discussion_r3669378602 4. **P1 — Avoid exhaustive/synchronous file planning in the equality-delete preflight.** Non-batch scans plan files twice, while batch mode loses its asynchronous planning benefit: https://github.com/apache/doris/pull/65851#discussion_r3670362669 The first two are data correctness/interoperability issues. The third is a rolling-upgrade availability regression. The fourth can materially regress planning latency and metadata work on large tables. I recommend resolving these and adding the focused nested equality-key, legacy ORC UUID/FIXED, projection-scoped collision, and `planFiles()` invocation-count tests before merge. -- 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]
