[
https://issues.apache.org/jira/browse/HBASE-13028?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14320594#comment-14320594
]
Sean Busbey commented on HBASE-13028:
-------------------------------------
it looks like there's a bunch of drift between the mapreduce and mapred APIs.
Unless someone objects, I'm going to limit this ticket's scope to just the
initialization stuff. I'll file a follow on with pointers on the rest of the
things that need to be ported over to the mapred implementation.
> Cleanup mapreduce API changes
> -----------------------------
>
> Key: HBASE-13028
> URL: https://issues.apache.org/jira/browse/HBASE-13028
> Project: HBase
> Issue Type: Bug
> Components: mapreduce
> Affects Versions: 1.0.0
> Reporter: Sean Busbey
> Assignee: Sean Busbey
> Priority: Blocker
> Labels: api
> Fix For: 1.0.0, 2.0.0
>
>
> The API changes made in HBASE-12798 need to be cleaned up so their common use
> makes more sense.
> * initialize should take some kind of configuration object that can be used
> in setting up a Connection (e.g. JobConf)
> * initialize should be called at the start of getSplits and getRecordReader
> rather than across the class when a member is needed
> * the same changes should be present in the mapred package since it is not
> longer deprecated
> * examples should rely on idiomatic MapReduce (specifically they should not
> be relying on JobConfigurable to get a JobConf instance)
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
