[SPARK-43876] Enable fast hashmap for distinct queries

• Type: Improvement
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 3.5.0
• Fix Version/s: 3.5.0
• Component/s: SQL
• Assignee: Wan Kun
• Reporter: Wan Kun
• Created: 30/May/23 03:41
• Updated: 20/Jun/23 17:14
• Resolved: 20/Jun/23 17:14

Description

Spark will enable fast hash map for primitive data types in HashAggregateExec.

Could we also enable this for distinct queries which bufferSchema is empty.

For example, we can also build a fast hash map with the key a + b for query

 SELECT distinct a, b from tab