hive-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Dayue Gao (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HIVE-10792) PPD leads to wrong answer when mapper scans the same table with multiple aliases
Date Fri, 22 May 2015 07:16:17 GMT

    [ https://issues.apache.org/jira/browse/HIVE-10792?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14555728#comment-14555728
] 

Dayue Gao commented on HIVE-10792:
----------------------------------

[~gopalv], yes, I think it's related to ORC PPD.

Your query on "test_txt" returns one row [0, 1], which is the correct answer. However, same
query on "test_orc" will return no rows.

> PPD leads to wrong answer when mapper scans the same table with multiple aliases
> --------------------------------------------------------------------------------
>
>                 Key: HIVE-10792
>                 URL: https://issues.apache.org/jira/browse/HIVE-10792
>             Project: Hive
>          Issue Type: Bug
>          Components: File Formats, Query Processor
>    Affects Versions: 0.13.0, 0.14.0, 0.13.1, 1.0.0, 1.2.0, 1.1.0
>            Reporter: Dayue Gao
>            Assignee: Dayue Gao
>            Priority: Critical
>         Attachments: HIVE-10792.test.sql
>
>
> Here's the steps to reproduce the bug.
> First of all, prepare a simple ORC table with one row
> {code}
> create table test_orc (c0 int, c1 int) stored as ORC;
> {code}
> Table: test_orc
> ||c0||c1||
> |0|1|
> The following SQL gets empty result which is not expected
> {code}
> select * from test_orc t1
> union all
> select * from test_orc t2
> where t2.c0 = 1
> {code}
> Self join is also broken
> {code}
> set hive.auto.convert.join=false; -- force common join
> select * from test_orc t1
> left outer join test_orc t2 on (t1.c0=t2.c0 and t2.c1=0);
> {code}
> It gets empty result while the expected answer is
> ||t1.c0||t1.c1||t2.c0||t2.c1||
> |0|1|NULL|NULL|
> In these cases, we pushdown predicates into OrcInputFormat. As a result, TableScanOperator
for "t1" can't receive its rows.



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

Mime
View raw message