[
https://issues.apache.org/jira/browse/HIVE-24861?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17768482#comment-17768482
]
xiegang commented on HIVE-24861:
--------------------------------
We encountered the similiar issue while connect to impala with hive jdbc. It
seems that the behavior is quit different between impala and hive. Empty result
is not enough to tell if there is no rows when use impala server.
So, do we consider have this fixed? Meanwhile，I wonder why we leave this issue
opened and closed the pull request. Is that because the hive server have the
bug to set hasMoreRows, if we use this in jdbc, it could be incompatiable with
server?
> Hive JDBC driver doesn't consider the value of 'hasMoreRows'
> ------------------------------------------------------------
>
> Key: HIVE-24861
> URL: https://issues.apache.org/jira/browse/HIVE-24861
> Project: Hive
> Issue Type: Bug
> Components: JDBC
> Reporter: Zoltán Borók-Nagy
> Priority: Major
> Labels: pull-request-available
> Time Spent: 1h
> Remaining Estimate: 0h
>
> TCLIService's FetchResults might return an empty result set, but with
> hasMoreRows=true. In that case the driver ignores the flag hasMoreRows and
> thinks it is the end of the result stream, causing data loss.
> I've seen this when the Hive JDBC driver was used to connect to Impala.
> IMPALA-7312 introduced a timeout on FetchResults(). If Impala cannot produce
> rows in the given timeout then it returns an empty result set, but setting
> hasMoreRows=true. However, the Hive JDBC driver interprets it as the end of
> the result stream and closes the operation.
> I think if hasMoreRows=true then the Hive JDBC driver should issue
> FetchResults() again.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
