[
https://issues.apache.org/jira/browse/HIVE-11849?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14791221#comment-14791221
]
Enis Soztutar commented on HIVE-11849:
--------------------------------------
Offline conversation with Jason, we have noted a couple of things:
- HiveHBaseTableSnapshotInputFormat.java uses the mapred API while
HiveHBaseTableInputFormat uses the mapreduce API. [~ndimiduk] I remember you
were talking about specifically that all Hive IFs use mapred. Is that changed?
- mapred version of the HBase's TableMapreduceUtil does not have the utility
methods to pass the Scan serialized inside the job configuration. The only
supported way is to set the Scan through now-deprecated
{{TableInputFormat.COLUMN_LIST}}.
- Although the {{mapred.TableMapreduceUtil}} does not support setting the
serialized scan, we can still manually set it and have the TSIF work correctly
since the mapred and mapreduce versions use the same underlying implementation
({{TableSnapshotInputFormatImpl}}).
> NPE in HiveHBaseTableShapshotInputFormat in query with just count(*)
> --------------------------------------------------------------------
>
> Key: HIVE-11849
> URL: https://issues.apache.org/jira/browse/HIVE-11849
> Project: Hive
> Issue Type: Bug
> Components: HBase Handler
> Affects Versions: 1.3.0
> Reporter: Jason Dere
>
> Adding the following example as a qfile test in hbase-handler fails. Looks
> like this may have been introduced by HIVE-5277.
> {noformat}
> SET hive.hbase.snapshot.name=src_hbase_snapshot;
> SET hive.hbase.snapshot.restoredir=/tmp;
> select count(*) from src_hbase;
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
