[HIVE-7060] Column stats give incorrect min and distinct_count

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Duplicate
• Affects Version/s: 0.13.0
• Fix Version/s: 0.14.0
• Component/s: Statistics
• Assignee: Unassigned
• Reporter: Xuefu Zhang
• Created: 14/May/14 17:15
• Updated: 13/Nov/14 19:40
• Resolved: 03/Jun/14 05:28

Description

It seems that the result from column statistics isn't correct on two measures for numeric columns: min (which is always 0) and distinct count. Here is an example:

select count(distinct avgTimeOnSite), min(avgTimeOnSite) from UserVisits_web_text_none;
...
OK
9	1
Time taken: 9.747 seconds, Fetched: 1 row(s)

The statisitics for the column:

desc formatted UserVisits_web_text_none avgTimeOnSite
...
# col_name              data_type               min                     max                     num_nulls               distinct_count          avg_col_len             max_col_len             num_trues               num_falses              comment

avgTimeOnSite           int                     0                       9                       0                       11                      null                    null                    null               

Issue Links

• duplicates: HIVE-4561 Column stats : LOW_VALUE (or HIGH_VALUE) will always be 0.0000 ,if all the column values larger than 0.0 (or if all column values smaller than 0.0)
• is related to: HIVE-4561 Column stats : LOW_VALUE (or HIGH_VALUE) will always be 0.0000 ,if all the column values larger than 0.0 (or if all column values smaller than 0.0)