You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@drill.apache.org by "Victoria Markman (JIRA)" <ji...@apache.org> on 2015/02/26 01:15:06 UTC
[jira] [Commented] (DRILL-2148) Wrong result with grouping on a
column of date type with streaming aggregation
[ https://issues.apache.org/jira/browse/DRILL-2148?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14337541#comment-14337541 ]
Victoria Markman commented on DRILL-2148:
-----------------------------------------
This bug was fixed by commit 5efc7e6 (DRILL-1062 - NULLS FIRST/NULLS LAST)
Verified fixed in:
{code}
#Mon Feb 23 16:54:54 EST 2015
git.commit.id.abbrev=c8d2fe1
{code}
Note for QA: I'm adding tests for this under: Functional/Passing/aggregation/bugs/drill-2148*.sql
> Wrong result with grouping on a column of date type with streaming aggregation
> ------------------------------------------------------------------------------
>
> Key: DRILL-2148
> URL: https://issues.apache.org/jira/browse/DRILL-2148
> Project: Apache Drill
> Issue Type: Bug
> Components: Execution - Relational Operators
> Affects Versions: 0.8.0
> Reporter: Victoria Markman
> Assignee: Chris Westin
> Priority: Critical
> Fix For: 0.8.0
>
> Attachments: 0_0_0.parquet, t1.out, t1.res
>
>
> Disable hash aggregation and run query below:
> {code}
> alter system set `planner.enable_hashagg` = false;
> select
> c_date,
> COUNT(*)
> from t1
> group by
> c_date
> order by
> c_date;
> {code}
> You will get wrong result. Because NULLs are sorted in the middle ( see DRILL-2084 ) they are folded in one of the non related groups.
> We might have the same problem with the merge join on date, time and timestamp columns.
> Attached is a parquet file that was used in this query.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)