[
https://issues.apache.org/jira/browse/HBASE-11137?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13998881#comment-13998881
]
Nick Dimiduk commented on HBASE-11137:
--------------------------------------
bq. Forwarding my +1 from RB.
Thanks [~enis]. Replied on RB as well (I'm not receiving notification email for
some reason).
In addition, I'll s/TableSnapshotInputFormatBase/TableSnapshotInputFormatImpl/
so use as a delegate (instead of a base-class) isn't confused.
bq. Did you figure out how Hive does filter pushdown without scan support in
mapred?
Yes, I looked into that. Hive works around the lack of feature by consuming our
internal API, mapreduce.TableInputFormatBase. I opened HBASE-11163 and
HIVE-7058 to get them back onto our public APIs on all fronts.
> Add mapred.TableSnapshotInputFormat
> -----------------------------------
>
> Key: HBASE-11137
> URL: https://issues.apache.org/jira/browse/HBASE-11137
> Project: HBase
> Issue Type: Improvement
> Components: mapreduce, Performance
> Affects Versions: 0.98.0, 0.96.2
> Reporter: Nick Dimiduk
> Assignee: Nick Dimiduk
> Fix For: 0.99.0, 0.98.3
>
> Attachments: HBASE-11137.00-0.98.patch, HBASE-11137.00.patch,
> HBASE-11137.01.patch, HBASE-11137.01_rerun.patch, HBASE-11137.02-0.98.patch,
> HBASE-11137.02.patch
>
>
> We should have feature parity between mapreduce and mapred implementations.
> This is important for Hive.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
