[HIVE-17465] Statistics: Drill-down filters don't reduce row-counts progressively

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 3.0.0
• Component/s: Physical Optimizer, Statistics
• Assignee: Vineet Garg
• Reporter: Gopal Vijayaraghavan
• Created: 06/Sep/17 18:57
• Updated: 22/May/18 23:13
• Resolved: 18/Sep/17 20:21

Description

explain select count(d_date_sk) from date_dim where d_year=2001 ;
explain select count(d_date_sk) from date_dim where d_year=2001  and d_moy = 9;
explain select count(d_date_sk) from date_dim where d_year=2001 and d_moy = 9 and d_dom = 21;

All 3 queries end up with the same row-count estimates after the filter.

            Map Operator Tree:
                TableScan
                  alias: date_dim
                  filterExpr: (d_year = 2001) (type: boolean)
                  Statistics: Num rows: 73049 Data size: 82034027 Basic stats: COMPLETE Column stats: COMPLETE
                  Filter Operator
                    predicate: (d_year = 2001) (type: boolean)
                    Statistics: Num rows: 363 Data size: 4356 Basic stats: COMPLETE Column stats: COMPLETE
 
        Map 1 
            Map Operator Tree:
                TableScan
                  alias: date_dim
                  filterExpr: ((d_year = 2001) and (d_moy = 9)) (type: boolean)
                  Statistics: Num rows: 73049 Data size: 82034027 Basic stats: COMPLETE Column stats: COMPLETE
                  Filter Operator
                    predicate: ((d_year = 2001) and (d_moy = 9)) (type: boolean)
                    Statistics: Num rows: 363 Data size: 5808 Basic stats: COMPLETE Column stats: COMPLETE
        Map 1 
            Map Operator Tree:
                TableScan
                  alias: date_dim
                  filterExpr: ((d_year = 2001) and (d_moy = 9) and (d_dom = 21)) (type: boolean)
                  Statistics: Num rows: 73049 Data size: 82034027 Basic stats: COMPLETE Column stats: COMPLETE
                  Filter Operator
                    predicate: ((d_year = 2001) and (d_moy = 9) and (d_dom = 21)) (type: boolean)
                    Statistics: Num rows: 363 Data size: 7260 Basic stats: COMPLETE Column stats: COMPLETE

Issue Links

• links to: ReviewBoard