Zhihua Deng created HIVE-24575:
----------------------------------
Summary: VectorGroupByOperator reusing keys can lead to wrong
results
Key: HIVE-24575
URL: https://issues.apache.org/jira/browse/HIVE-24575
Project: Hive
Issue Type: Bug
Components: Vectorization
Reporter: Zhihua Deng
Assignee: Zhihua Deng
A common sql like
{code:java}
select category as category, count(distinct maskdid) as uv from
dwd_internal_inc_d group by category{code}
can have a wrong result on the trunk, the result of column category can be
confused and
aggregate of distinct maskdid is also wrong.
After some debugging, We find that the problem is caused by wrong byteStarts[i]
when using it to copy the current keys to the reusable keys:
[https://github.com/apache/hive/blob/master/ql/src/java/org/apache/hadoop/hive/ql/exec/vector/wrapper/VectorHashKeyWrapperGeneral.java#L351-L362]
The byteStarts[i] is always 0 due to Arrays.fill(byteStarts, 0); so it copies
the range from 0 other then the real start index to len of the current keys to
the reusable keys when clone.byteValues[i].length >= byteValues[i].length met,
which results to the problem.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
