sqoop-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jarek Jarcec Cecho (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (SQOOP-489) Cannot define partition keys for Hive tables created through Sqoop
Date Sat, 26 May 2012 06:13:23 GMT

    [ https://issues.apache.org/jira/browse/SQOOP-489?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13283912#comment-13283912
] 

Jarek Jarcec Cecho commented on SQOOP-489:
------------------------------------------

I also do have one concern with automatic fix. This time, let me start with example :-)

Let's consider table with three columns A, B and C with two rows - "1, 2, 3" and "4, 5, 6".
Also let say that we applied suggested fix and also solve the issue with importing all columns.
Sqoop command would contain --hive-partition A --hive-value 666. Then file on HDFS  would
probably contain two rows - "2, 3" and "5, 6". View on hive would be:

B  C  A
2  3  666
5  6  666

I do not see an issue with changing column ordering, however I'm concerned that we silently
ignored one column in the table and replaced it with constant value from command line -- both
values "1" and "4" from original table were lost.

I do not have easy solution to address this problem and therefore I would suggest to fix this
JIRA by gracefully exiting in case that user will specify partition column as one of the exported
columns.

Jarcec
                
> Cannot define partition keys for Hive tables created through Sqoop
> ------------------------------------------------------------------
>
>                 Key: SQOOP-489
>                 URL: https://issues.apache.org/jira/browse/SQOOP-489
>             Project: Sqoop
>          Issue Type: Bug
>    Affects Versions: 1.4.1-incubating
>            Reporter: Kathleen Ting
>            Assignee: Cheolsoo Park
>         Attachments: SQOOP-489.patch
>
>
> By enabling the table option, Sqoop includes every column in the table in the create
table query, and by enabling the hive-partition-key option, Sqoop blindly appends the "partitioned
by" clause. Now if you specify one of columns in the table in the hive-partition-key, this
will cause a syntax error in Hive.
> For example, if we have a table 'FOO' that has columns 'I' and 'J':
> sqoop create-hive-table --table FOO ...
> will generate the following Hive query:
> CREATE TABLE IF NOT EXISTS `FOO` ( `I` STRING, `J` STRING)
> Now if we add "--hive-partition-key I" to the command, Sqoop generates the following
query:
> CREATE TABLE IF NOT EXISTS `FOO` ( `I` STRING, `J` STRING) PARTITIONED BY (I STRING)
> The problem is that since 'I' is defined twice (once in CRATE TABLE and once in PARTITIONED
BY), this is a syntax error in Hive.
> This correct query would be something like:
> CREATE TABLE IF NOT EXISTS `FOO` (`J` STRING) PARTITIONED BY (I STRING)

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Mime
View raw message