[
https://issues.apache.org/jira/browse/HIVE-1229?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12885689#action_12885689
]
HBase Review Board commented on HIVE-1229:
------------------------------------------
Message from: bkm.had...@gmail.com
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://review.hbase.org/r/239/#review309
-----------------------------------------------------------
http://svn.apache.org/repos/asf/hadoop/hive/trunk/hbase-handler/src/java/org/apache/hadoop/hive/hbase/HBaseSerDe.java
<http://review.hbase.org/r/239/#comment1362>
I have added a HashMap to HBaseSerDe to cache the String to byte []
mapping. The code looks it up for the serialization path. The map is also
passed to LazyHBaseRow and LazyHBaseCellMap for lookup during deserialization.
In addition I have moved some function calls to serdeParams and saved their
return values as instance variables to reduce the per row calls.
http://svn.apache.org/repos/asf/hadoop/hive/trunk/hbase-handler/src/java/org/apache/hadoop/hive/hbase/HBaseSplit.java
<http://review.hbase.org/r/239/#comment1363>
I have left this in the 3rd patch. Thanks for explaining this. I don't
think we have a test case which exposes this or even whether the serde instance
and the record reader instance could get out of sync, but good idea to leave it
in.
The failing tests were due to an improperly initialized Scan instance.
http://svn.apache.org/repos/asf/hadoop/hive/trunk/hbase-handler/src/java/org/apache/hadoop/hive/hbase/HiveHBaseTableOutputFormat.java
<http://review.hbase.org/r/239/#comment1366>
This OutputFormat is from the deprecated mapred package - this is needed to
keep it compatible with the storage handler which needs it to be compatible
with this -- see the getOutputFormat() method.
http://svn.apache.org/repos/asf/hadoop/hive/trunk/hbase-handler/src/java/org/apache/hadoop/hive/hbase/LazyHBaseCellMap.java
<http://review.hbase.org/r/239/#comment1365>
Actually, parse() should also be called only once. I have added the missing
call to set parsed to true. In addition the cached values are now passed in
from the serde to lazy row to the lazy cell map in the deserialization path.
http://svn.apache.org/repos/asf/hadoop/hive/trunk/hbase-handler/src/java/org/apache/hadoop/hive/hbase/LazyHBaseRow.java
<http://review.hbase.org/r/239/#comment1364>
Done, reverted these changes.
- bkm
> replace dependencies on HBase deprecated API
> --------------------------------------------
>
> Key: HIVE-1229
> URL: https://issues.apache.org/jira/browse/HIVE-1229
> Project: Hadoop Hive
> Issue Type: Improvement
> Components: HBase Handler
> Affects Versions: 0.6.0
> Reporter: John Sichi
> Assignee: Basab Maulik
> Fix For: 0.7.0
>
> Attachments: HIVE-1229.1.patch, HIVE-1229.2.patch, HIVE-1229.3.patch
>
>
> Some of these dependencies are on the old Hadoop mapred packages; others are
> HBase-specific. The former have to wait until the rest of Hive moves over to
> the new Hadoop mapreduce package, but the HBase-specific ones don't have to
> wait.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
