[jira] [Comment Edited] (HIVE-21503) Vectorization: query with regex gives incorrect results with vectorization

["Rajesh Balamohan (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HIVE-21503?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16801325#comment-16801325 ] 

Rajesh Balamohan edited comment on HIVE-21503 at 3/26/19 2:50 AM:
------------------------------------------------------------------

It reproduces consistently with case-when. Here is the simplest query which was reproducing the issue.
{noformat}
with test as (
  select (CASE      
      WHEN name like '%radio%' THEN 'radio'
      ELSE name
  END) AS name from tmp.delete_vect_test
)
select name, count(*) as c from test group by name order by c desc limit 10;{noformat}


was (Author: rajesh.balamohan):
It reproduces consistently with case-when. Here is the simplest query which was reproducing the issue.

```
with test as (
  select (CASE      
      WHEN name like '%radio%' THEN 'radio'
      ELSE name
  END) AS name from tmp.delete_vect_test
)
select name, count(*) as c from test group by name order by c desc limit 100;
```

> Vectorization: query with regex gives incorrect results with vectorization
> --------------------------------------------------------------------------
>
>                 Key: HIVE-21503
>                 URL: https://issues.apache.org/jira/browse/HIVE-21503
>             Project: Hive
>          Issue Type: Bug
>          Components: Vectorization
>            Reporter: Rajesh Balamohan
>            Assignee: Laszlo Bodor
>            Priority: Major
>
> i see wrong results with vectorization. Without vectorization, it works fine. Suspecting minor issue in {{StringGroupColConcatCharScalar}}
> {noformat}
> e.g 
> WHEN x like '%radio%' THEN 'radio' 
> WHEN x like '%tv%' THEN 'tv'
> {noformat}



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)