[
https://issues.apache.org/jira/browse/PHOENIX-946?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13977633#comment-13977633
]
James Taylor commented on PHOENIX-946:
--------------------------------------
bq. (from [~ndimiduk] on PHOENIX-918) I think this could be a very good
approach, benefiting both users across projects. Can we dig into this in more
detail? Basically, Hive's HBaseStorageHandler converts the relevant portions of
the hive query into a configured (meaning, scan plus hadoop.mapred namespace)
InputFormat for the execution engine to consume. So for this to work with hive,
we'd need a Phoenix-planned query that can be partitioned according to the Hive
runtime semantics.
Would this imply that the query is then executed as a map-reduce job? If that
is the case, you'd lose a lot of the benefit of using Phoenix. Would another
approach be for Phoenix to support HiveQL parsing? We'd need to do a gap
analysis between the two to see what's missing. For example, we'd need to beef
up the built-in functions supported by Phoenix. We could then just execute the
query and hand back a regular ResultSet (which could potentially be wrapped by
whatever interface you'd require).
> Use Phoenix to service Hive queries over HBase data
> ---------------------------------------------------
>
> Key: PHOENIX-946
> URL: https://issues.apache.org/jira/browse/PHOENIX-946
> Project: Phoenix
> Issue Type: Bug
> Reporter: James Taylor
>
--
This message was sent by Atlassian JIRA
(v6.2#6252)
