hive-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Alan Gates (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (HIVE-7571) RecordUpdater should read virtual columns from row
Date Sat, 23 Aug 2014 20:36:11 GMT

     [ https://issues.apache.org/jira/browse/HIVE-7571?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Alan Gates updated HIVE-7571:
-----------------------------

    Status: Open  (was: Patch Available)

This patch assumes it can find the ROW__ID column via the column name, but that isn't always
passed to the table.  Changing to require the constructor of the RecordUpdater to tell which
column it is in.

> RecordUpdater should read virtual columns from row
> --------------------------------------------------
>
>                 Key: HIVE-7571
>                 URL: https://issues.apache.org/jira/browse/HIVE-7571
>             Project: Hive
>          Issue Type: Sub-task
>          Components: Transactions
>    Affects Versions: 0.13.0
>            Reporter: Alan Gates
>            Assignee: Alan Gates
>         Attachments: HIVE-7571.WIP.patch, HIVE-7571.patch
>
>
> Currently RecordUpdater.update and delete take rowid and original transaction as parameters.
 These values are already present in the row as part of the new ROW__ID virtual column in
HIVE-7513, and thus can be read by the writer from there.  And the writer will already have
to handle skipping ROW__ID when writing, so it needs to be aware of that column anyone.
> We could instead read the values from ROW__ID and then remove it from the object inspector
in FileSinkOperator, but this will be hard in the vectorization case where rows are being
dealt with 10k at a time.
> For these reasons it makes more sense to do this work in the writer.



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Mime
View raw message