[HIVE-27356] Hive should write name of blob type instead of table name in Puffin

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 4.1.0
• Assignee: Simhadri Govindappa
• Reporter: Zoltán Borók-Nagy
• Created: 18/May/23 08:15
• Updated: 02/Jun/24 13:39
• Resolved: 02/Jun/24 13:39

Description

Currently Hive writes the name of the table plus snapshot id as blob type:

https://github.com/apache/hive/blob/aa1e067033ef0b5468f725cfd3776810800af96d/iceberg/iceberg-handler/src/main/java/org/apache/iceberg/mr/hive/HiveIcebergStorageHandler.java#L422

Instead, it should write the name of the blog it writes. Table name and snapshot id are redundant information anyway, as they can be inferred from the location and filename of the puffin file.

Currently it writes a non-standard blob (Standard blob types are listed here). I think it would be better to write standard blobs for interoperability. But if Hive wants to write non-standard blobs anyway, it should still come up with a descriptive name for them, e.g. 'hive-column-statistics-v1'.

Issue Links

• is a child of: HIVE-27158 Store hive columns stats in puffin files for iceberg tables