[HIVE-21267] Extend HiveRelColumnsAlignment to reorder group-by and join keys on decreasing NDV automatically

• Type: Improvement
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: CBO
• Labels: newbie
• Assignee: John Sherman
• Reporter: Jesús Camacho Rodríguez
• Created: 14/Feb/19 00:10
• Updated: 27/Feb/24 22:13

Description

HiveRelColumnsAlignment was introduced to align the order of columns in join, group-by, and order-by operators in the plan pipeline, trying to increase the effect of ReduceDeduplication and thus reducing data shuffle.

The optimization could be extended to reorder group-by and join keys on decreasing NDV, which would accelerate comparison runtime.