[HIVE-18108] in case basic stats are missing; rowcount estimation depends on the selected columns size

• Type: Sub-task
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Assignee: Zoltan Haindrich
• Reporter: Zoltan Haindrich
• Created: 20/Nov/17 15:26
• Updated: 05/Jan/18 10:49
• Resolved: 05/Jan/18 10:49

Description

in case basicstats are not available (especially rowcount):

set hive.stats.autogather=false;
create table t (a integer, b string);

insert into t values (1,'asd1');
insert into t values (2,'asd2');
insert into t values (3,'asd3');
insert into t values (4,'asd4');
insert into t values (5,'asd5');

explain select a,count(1) from t group by a;
-- estimated to read 8 rows from table t
explain select b,count(1) from t group by b;
-- estimated: 1 rows
explain select a,b,count(1) from t group by a,b;
-- estimated: 1 rows

it may not depend on the actually selected column set.

Issue Links

• blocks: HIVE-18140 Partitioned tables statistics can go wrong in basic stats mixed case
• is blocked by: HIVE-18149 Stats: rownum estimation from datasize underestimates in most cases
• links to: re