[SPARK-40804] Missing handling a catalog name in destination tables in `RenameTableExec`

• Type: Bug
• Status: In Progress
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 3.1.3, 3.3.0, 3.2.2
• Fix Version/s: 3.1.3, 3.3.0, 3.2.2
• Component/s: SQL

Environment

This depends on Spark versions, since the version in which ALTER TABLE RENAME TO is added.

• Flags: Patch
• External issue URL: https://github.com/apache/spark/pull/38268

Language

scala

• Assignee: Unassigned
• Reporter: Tomohiro Tanaka
• Created: 15/Oct/22 14:43
• Updated: 19/Oct/22 02:00

Description

Currently `RenameTableExec` only handles an empty namespace for destination tables as the following current spec:

(3.3.0 is picked up) https://github.com/apache/spark/blob/v3.3.0/sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/v2/RenameTableExec.scala#L48

 

# L48
    val qualifiedNewIdent = if (newIdent.namespace.isEmpty) {
      Identifier.of(oldIdent.namespace, newIdent.name)
    } else newIdent

This part doesn't handle the case where a destination table is specified catalog.db.table. For example, Apache Iceberg is expected to handle the destination table as catalog.db.table in the document; https://iceberg.apache.org/docs/latest/spark-ddl/#alter-table--rename-to

 

If catalog.db.table is passed to "ALTER TABLE <src> RENAME TO <dst>" query, there's a difference of handling namespaces between source and destination tables. Specifically, source tables can be correctly handled as [db] for its namespace, but destination tables are handled as [catalog, db] for its namespace.

Issue Links

• links to: [Github] Pull Request #38268 (tomtongue)