LuciferYang opened a new issue, #10406: URL: https://github.com/apache/paimon/issues/10406
### Search before asking - [X] I searched in the [issues](https://github.com/apache/paimon/issues) and found nothing similar. ### Paimon version master (6b29174a82) ### Compute Engine Java API (`ReadBuilder.newStreamScan()`) ### Minimal reproduce step 1. Create a data-evolution table `(f0 INT, f1 STRING, f2 BLOB, f3 BLOB)` with `row-tracking.enabled` and `data-evolution.enabled`. 2. Write `f0, f1` for two rows, row ids [0, 1]. 3. Open a stream scan and consume the initial plan. 4. Commit an update that writes only `f2, f3` for row ids [0, 1]. Its delta holds only two `.blob` files. 5. Plan the next delta, either projecting `f0`, or projecting `f2` with `deletion-vectors.enabled = true`. ### What doesn't meet your expectations? Both plans throw: ``` IllegalStateException: A data-evolution row range group must contain a normal (non-blob, non-vector-store) anchor file. ``` The normal file of this row range is in an earlier snapshot, so the delta group has no normal file, but `retrieveAnchorFile` is still called on the whole group: in `DataEvolutionFileStoreScan.pruneByReadType` when deletion vectors are on, and when no file in the group matches the projection; and in `DataEvolutionSplitRead.readDeletionVector`. Expected: the delta reads like a delta that updates only `f1`, which already returns both rows with the unprojected columns NULL, with or without deletion vectors. ### Anything else? Returning the group unchanged when nothing matches is not enough: a blob bunch whose read type has no blob field fails with "Blob bunch read type should contain a blob field." Found while reviewing #9966, which keeps these groups unchanged but does not fix this. ### Are you willing to submit a PR? - [X] I'm willing to submit a PR! -- 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]
