[
https://issues.apache.org/jira/browse/HIVE-10227?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14505401#comment-14505401
]
Sushanth Sowmyan commented on HIVE-10227:
-----------------------------------------
After some sleeping on this, I feel like I should be stricter still about
erroring out whenever .create is called, so that no events get seen as getting
processed, but I don't think adding one more Factory is a good way of doing
that.
Here's what I now think:
a) We get rid of NoopFactory as a default - that should move to tests, and not
stay here - we require that that config parameter be set to some instantiatable
factory to use this class.
b) We get rid of the explicit InvalidStateFactory I mention above, but
instantiate it as an inline anonymous Factory instantiation if we fail to load
whatever Factory class the user provides.
Thoughts?
Also, is it possible for us to do any factory refactoring in another jira? This
jira is huge enough that the longer we leave it uncommitted, the more it'll be
exposed to rebasing needs. Also, a couple of other jiras like HIVE-9674 are
awaiting this landing before they can be made patch-available.
> Concrete implementation of Export/Import based ReplicationTaskFactory
> ---------------------------------------------------------------------
>
> Key: HIVE-10227
> URL: https://issues.apache.org/jira/browse/HIVE-10227
> Project: Hive
> Issue Type: Sub-task
> Components: Import/Export
> Affects Versions: 1.2.0
> Reporter: Sushanth Sowmyan
> Assignee: Sushanth Sowmyan
> Attachments: HIVE-10227.2.patch, HIVE-10227.3.patch,
> HIVE-10227.4.patch, HIVE-10227.patch
>
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
