[jira] [Resolved] (HIVE-13595) HiveServer2: Evaluate if ThriftJDBCBinarySerde should implement VectorizedSerde

["Vaibhav Gumashta (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/HIVE-13595?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Vaibhav Gumashta resolved HIVE-13595.
-------------------------------------
    Resolution: Later

We've investigated the idea of implementing VectorizedSerde interface for writing final results, but as of now there are some issues:
1. Not all datatypes are vectorized.
2. For the ones that are, we'll also need to avoid the vector --> row --> columnar translation we're doing.
3. For final write, we'll need to use VectorFileSinkOperator instead of FileSinkOperator.

We'll take up these as separate jiras in HIVE-14549.

Thanks [~ziyangz] for the investigation.

> HiveServer2: Evaluate if ThriftJDBCBinarySerde should implement VectorizedSerde
> -------------------------------------------------------------------------------
>
>                 Key: HIVE-13595
>                 URL: https://issues.apache.org/jira/browse/HIVE-13595
>             Project: Hive
>          Issue Type: Sub-task
>          Components: HiveServer2
>    Affects Versions: 2.1.0
>            Reporter: Vaibhav Gumashta
>            Assignee: Ziyang Zhao
>
> As part of HIVE-12049, ThriftJDBCBinarySerde was introduced which buffers rows and writes thrift converted columnar row batches as part of the final task output. Hive has VectorizedSerde which is used during vectorized operations. We should explore if ThriftJDBCBinarySerde should implement that.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)