[HIVE-14741] Incorrect results on boolean col when vectorization is ON

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 2.0.0, 2.1.0, 2.3.2
• Labels: orc vectorization
• Assignee: Unassigned
• Reporter: Amruth Sampath
• Created: 13/Sep/16 12:02
• Updated: 30/Jul/19 06:10

Description

I have attached the ORC part file on which the issue is manifesting. It has just one boolean column (lot of nulls, 231=trues : verified using orc file dump utility)

1) Create external table on the part file attached

CREATE EXTERNAL TABLE bool_vect_issue (
`bool_col` BOOLEAN)
ROW FORMAT SERDE
'org.apache.hadoop.hive.ql.io.orc.OrcSerde'
STORED AS INPUTFORMAT
'org.apache.hadoop.hive.ql.io.orc.OrcInputFormat'
OUTPUTFORMAT
'org.apache.hadoop.hive.ql.io.orc.OrcOutputFormat'
LOCATION
'<loc to which the part file is copied>';

2)
set hive.vectorized.execution.enabled = true;
select sum(if((bool_col) , 1, 0)) from bool_vect_issue;
gives
708206

3)
set hive.vectorized.execution.enabled = false;
select sum(if((bool_col) , 1, 0)) from bool_vect_issue;
gives
231

The issue seem to have the same impact as HIVE-12435