[SPARK-49352] Avoid redundant array transform for identical expression

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 4.0.0, 3.3.4, 3.5.2, 3.4.3
• Fix Version/s: 4.0.0, 3.4.4, 3.5.3
• Component/s: SQL
• Labels: pull-request-available
• Assignee: L. C. Hsieh
• Reporter: L. C. Hsieh
• Created: 22/Aug/24 06:41
• Updated: 11/Oct/24 10:12
• Resolved: 23/Aug/24 22:43

Description

Our customer encounters significant performance regression when migrating from Spark 3.2 to Spark 3.4 on a `Insert Into` query which is analyzed as a `AppendData` on an Iceberg table.

We found that the root cause is in Spark 3.4, `TableOutputResolver` resolves the query with additional `ArrayTransform` on an `ArrayType` field. The `ArrayTransform`'s lambda function is actually an identical function, i.e., the transformation is redundant.

Issue Links

• links to: GitHub Pull Request #47843
• links to: GitHub Pull Request #47862
• links to: GitHub Pull Request #47863