deniskuzZ commented on code in PR #6515:
URL: https://github.com/apache/hive/pull/6515#discussion_r3337070347
##########
iceberg/iceberg-handler/src/main/java/org/apache/iceberg/mr/hive/IcebergTableUtil.java:
##########
@@ -452,6 +452,25 @@ public static void performMetadataDelete(Table
icebergTable, String branchName,
deleteFiles.deleteFromRowFilter(exp).commit();
}
+ /**
+ * Parses an Iceberg partition path into a Hive-compatible spec map.
Review Comment:
The issue is in how the partition path string is generated and then parsed
back. `spec.partitionToPath(partitionData)` uses Iceberg's toHumanString() to
convert partition values to path segments. For `NULL` values, toHumanString()
produces the string "null". So the path looks like col=null. When
Warehouse.makeSpecFromName parses this path back into a spec map, it just sees
the literal string "null" as the value — it has no way to know whether this
came from an actual NULL or from a legitimate string value "null". It puts
"null" (the string) into the spec map rather than `NULL`.
The fix uses `partitionData` to check whether the value is actually `NULL`
at the Iceberg level, and if so, puts null into the spec map directly. We need
this because toHumanString() is lossy — it conflates NULL and the string
"null" into the same path representation.
--
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]
