[
https://issues.apache.org/jira/browse/HIVE-7513?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14088642#comment-14088642
]
Eugene Koifman commented on HIVE-7513:
--------------------------------------
If we modify MapOperator.process(Writable value) to process(Writable key,
Writable value), we could pass the key ExecMapper.map() which means we don't
need to use IOContext to pass the RecordIdentifier from IO layer into the
operator pipeline.
Note that Tez also calls MapOperator.process() - need to see if 'key' is
available in that context.
> Add ROW__ID VirtualColumn
> -------------------------
>
> Key: HIVE-7513
> URL: https://issues.apache.org/jira/browse/HIVE-7513
> Project: Hive
> Issue Type: Sub-task
> Components: Query Processor
> Affects Versions: 0.13.1
> Reporter: Eugene Koifman
> Assignee: Eugene Koifman
> Attachments: HIVE-7513.3.patch, HIVE-7513.4.patch, HIVE-7513.5.patch
>
>
> In order to support Update/Delete we need to read rowId from AcidInputFormat
> and pass that along through the operator pipeline (built from the WHERE
> clause of the SQL Statement) so that it can be written to the delta file by
> the update/delete (sink) operators.
> The parser will add this column to the projection list to make sure it's
> passed along.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
