[
https://issues.apache.org/jira/browse/HIVE-8720?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14196623#comment-14196623
]
Prasanth J commented on HIVE-8720:
----------------------------------
[~hagleitn] Can we have this for 0.14? These are just test file diffs to make
the qfile results consistent across platforms.
> Update orc_merge tests to make it consistent across OS'es
> ---------------------------------------------------------
>
> Key: HIVE-8720
> URL: https://issues.apache.org/jira/browse/HIVE-8720
> Project: Hive
> Issue Type: Bug
> Affects Versions: 0.14.0
> Reporter: Prasanth J
> Assignee: Prasanth J
> Attachments: HIVE-8720.1.patch, orc_merge5_filedump_macosx.txt,
> orc_merge5_filedump_opensuse.txt
>
>
> orc_merge*.q test cases fails with qfile diffs related to file size on
> different OSes. I have seen failures with Open SUSE and CentOS. The order of
> insertion of rows into ORC table impacts the file size because of run length
> encoding. Since the order of rows is not guaranteed during insertion into
> table we may get different file sizes. We cannot add ORDER BY to insert
> queries as it will force insertion through single reducer which will disable
> orc merge file optimization. Since these test cases test if the files are
> merged or not it is sufficient to know the number of files after merging.
> Instead of DESCRIBE FORMATTED (which shows the numFiles and fileSize) we can
> use "dfs -ls" to know the number of files.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
