hive-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Alan Gates (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HIVE-8740) Sorted dynamic partition does not work correctly with constant folding
Date Wed, 05 Nov 2014 19:24:35 GMT

    [ https://issues.apache.org/jira/browse/HIVE-8740?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14198913#comment-14198913
] 

Alan Gates commented on HIVE-8740:
----------------------------------

Is this the set of queries you're referring to?
{code}
+insert into table acid partition(ds)  select key,value,ds from srcpart;
+select count(*) from acid where ds='2008-04-08';
+
+insert into table acid partition(ds='2008-04-08') values("foo", "bar");
+select count(*) from acid where ds='2008-04-08';
+
+update acid set key = 'foo' where ds='2008-04-08';
+select count(*) from acid where ds='2008-04-08';
+
+update acid set key = 'foo' where ds in ('2008-04-08');
+select count(*) from acid where ds in ('2008-04-08');
+
+delete from acid where key = 'foo' and ds='2008-04-08';
+select count(*) from acid where ds='2008-04-08';
{code}
I don't see an issue here.  You're setting every row in 2008-04-08 to have a key of foo, and
then deleting all records with a key of foo, so that partition will be empty.

> Sorted dynamic partition does not work correctly with constant folding
> ----------------------------------------------------------------------
>
>                 Key: HIVE-8740
>                 URL: https://issues.apache.org/jira/browse/HIVE-8740
>             Project: Hive
>          Issue Type: Bug
>    Affects Versions: 0.14.0
>            Reporter: Prasanth J
>            Assignee: Prasanth J
>         Attachments: HIVE-8740.1.patch
>
>
> Sorted dynamic partition optimization looks for partition columns from the operator above
FileSinkOperator. As per hive convention it expects partition columns at the last. But with
HIVE-8585 equality filters on partition columns gets folded to constant. The column pruner
then prunes the constant expression as they don't reference any columns. This in some cases
will yield unexpected results (throw ArrayIndexOutOfBounds exception) with sorted dynamic
partition insert optimization. In such we don't really need sorted dynamic partition optimization.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message