[
https://issues.apache.org/jira/browse/DRILL-3534?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16138297#comment-16138297
]
John Omernik commented on DRILL-3534:
-------------------------------------
[~kbotzum] I agree whole heartedly with your statement except in
implementation...
I.e. I think we should be doing the inserts into the HBASE/MapRDB 100%, I think
many folks would benefit from this, and in talking with other Drill members,
this actually may be another area where it may be good to discuss how Drill can
implement an interface for INSERT for storage plugins. I.e. beyond just HBASE,
what about Mongo DB? Instead of handling all functionality of insert in the
plugin itself, we could instead have an interface that storage plugins could
extend if they want to do inserts. This would help make it easier to implement
on backends that would easily support.
Where I disagree though is I think Filesystem inserts, especially as it relates
to Parquet files, is a must have in the MVP for Drill Inserts. Drill is often
compared to Hive on Spark, Hive on Tez, and Impala, and all three of those
support some variance of Insert into Parquet table... this is one area that can
be a deal breaker for those looking to stack up Drill against other SQL on
Hadoop Products. Thus, starting with the MapRDB/Hbase stuff, while it may be
easier, I think at a minimum should done in parallel with the Parquet INSERTS
so we can achieve some feature parity.
> Insert into table support
> -------------------------
>
> Key: DRILL-3534
> URL: https://issues.apache.org/jira/browse/DRILL-3534
> Project: Apache Drill
> Issue Type: New Feature
> Reporter: Mehant Baid
> Fix For: Future
>
>
> Umbrella JIRA to track the "Insert into table" feature. More details
> regarding the scope, design etc will follow as things start to materialize.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
