[IMPALA-4289] Mark slots produced by NDV() and NDV_NO_FINALIZE() as non-nullable.

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: Impala 2.7.0
• Fix Version/s: Impala 2.8.0
• Component/s: Frontend
• Labels: compute-stats performance
• Target Version: Impala 2.8.0
• Assignee: Alexander Behm
• Reporter: Alexander Behm
• Created: 12/Oct/16 23:56
• Updated: 13/Oct/16 05:36
• Resolved: 13/Oct/16 05:36

Description

COMPUTE STATS and COMPUTE INCREMENTAL STATS might get a slight speedup if we marked the aggregation slots as non-nullable.

Of course, regular uses of NDV() and NDV_NO_FINALIZE() would also benefit.

Not sure what speedup might be (simple experiments were inconclusive), but making this change seems strictly better than leaving the slots nullable.