You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hive.apache.org by "Prasanth Jayachandran (JIRA)" <ji...@apache.org> on 2015/11/17 01:46:11 UTC
[jira] [Commented] (HIVE-12424) Make use of Kryo's Object-to-Object
deep copy
[ https://issues.apache.org/jira/browse/HIVE-12424?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15007710#comment-15007710 ]
Prasanth Jayachandran commented on HIVE-12424:
----------------------------------------------
This issue is currently blocked by https://github.com/EsotericSoftware/kryo/issues/148 which copies transient fields as well. maven jars has not reflected the changes yet.
> Make use of Kryo's Object-to-Object deep copy
> ---------------------------------------------
>
> Key: HIVE-12424
> URL: https://issues.apache.org/jira/browse/HIVE-12424
> Project: Hive
> Issue Type: Sub-task
> Components: Serializers/Deserializers
> Affects Versions: 2.0.0
> Reporter: Prasanth Jayachandran
> Assignee: Prasanth Jayachandran
>
> Currently, plan serialization and operator tree serialization uses Object -> bytes -> Object approach for deep copy. It also uses ByteArrayOutputStream as intermediate buffer whose write method is synchronized. Similarly read from ByteArrayInputStream is also synchronized. Also Utilities.clonePlan() creates a new HiveConf object that scans through conf directories and adds site.xml which is an expensive operation. All these can be avoided using Kryo's Object -> Object deep copy.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)