[
https://issues.apache.org/jira/browse/CASSANDRA-3264?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13141200#comment-13141200
]
Jonathan Ellis commented on CASSANDRA-3264:
-------------------------------------------
Right. Which makes no sense at all to me, an interface to return
{noformat}
RowA | col1
RowA | col2
RowA | col3
...
RowA | colN
{noformat}
gives you the same information but in a more consistent fashion.
Put another way: an interface that gives you all the columns for one row makes
sense (the existing CFIF). An interface that gives you one column per row also
makes sense (transposition). An interface that "pages" one row into multiple
rows does not make sense, it's a workaround for an implementation problem with
no real justification otherwise. So let's finish transposition instead since
that solves the same problem with more coherence to the rest of the system.
> Add wide row paging for ColumnFamilyInputFormat and ColumnFamilyOutputFormat
> ----------------------------------------------------------------------------
>
> Key: CASSANDRA-3264
> URL: https://issues.apache.org/jira/browse/CASSANDRA-3264
> Project: Cassandra
> Issue Type: Improvement
> Components: Hadoop
> Reporter: T Jake Luciani
> Labels: lhf
> Fix For: 1.0.2
>
>
> Hadoop input/output formats currently can OOM on wide rows.
> We can add a new option to the ConfigHelper like columnPagingSize with a
> default of Integer.MAX_VALUE.
> The input format would page the row internally rather than pull it over at
> once.
> The output format could also use this to avoid sending huge rows over at once.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
