[
https://issues.apache.org/jira/browse/CASSANDRA-1315?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12898441#action_12898441
]
Jonathan Ellis commented on CASSANDRA-1315:
-------------------------------------------
bq. In order to use Thrift, you would need to generate code for your dynamic
language, and then distribute it to all of the nodes in your Hadoop cluster:
either as a library that you update for each Cassandra version, or as a JAR'd
script dependency. Not the end of the world, I suppose, but more difficult then
distributing only the protocol file.
How is that different from having to distribute the Avro library for whatever
streaming processor language you are using, other than the codegen step?
> ColumnFamilyOutputFormat should use client API objects
> ------------------------------------------------------
>
> Key: CASSANDRA-1315
> URL: https://issues.apache.org/jira/browse/CASSANDRA-1315
> Project: Cassandra
> Issue Type: Bug
> Components: Hadoop
> Reporter: Stu Hood
> Assignee: Stu Hood
> Fix For: 0.7 beta 2
>
> Attachments: 0001-Use-Avro-objects-as-input-to-CFOutputFormat.patch,
> 0002-Allow-multiple-mutations-per-key-to-arrive-during-in.patch
>
>
> ColumnFamilyOutputFormat currently takes IColumns as its input, meaning that
> users need to understand Cassandra's internals reasonably well in order to
> use it, and need to hardcode things like the comparator type and clock type
> into their MapReduce jobs.
> Instead, CFOutputFormat should take either Thrift or Avro objects, which are
> familiar interfaces for users.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
