[
https://issues.apache.org/jira/browse/HIVE-16974?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16065288#comment-16065288
]
Naveen Gangam commented on HIVE-16974:
--------------------------------------
[~asherman] Thank you for the suggestion. I have taken a quick look at using
"nulls first/last" clause as part of the query but then it is still a problem
when there are multiple rows with null names. Their order is
un-deterministic/random. So thats when I was thinking about doing away with
sorting on null columns.
Does this help? Thanks
> Change the sort key for the schema tool validator to be <ID>
> ------------------------------------------------------------
>
> Key: HIVE-16974
> URL: https://issues.apache.org/jira/browse/HIVE-16974
> Project: Hive
> Issue Type: Sub-task
> Components: HiveServer2
> Affects Versions: 3.0.0
> Reporter: Naveen Gangam
> Assignee: Naveen Gangam
>
> In HIVE-16729, we introduced ordering of results/failures returned by
> schematool's validators. This allows fault injection testing to expect
> results that can be verified. However, they were sorted on NAME values which
> in the HMS schema can be NULL. So if the introduced fault has a NULL/BLANK
> name column value, the result could be different depending on the backend
> database(if they sort NULLs first or last).
> So I think it is better to sort on a non-null column value.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
