You are viewing a plain text version of this content. The canonical link for it is here.

Posted to dev@sqoop.apache.org by "Ryan Blue (JIRA)" <ji...@apache.org> on 2015/04/29 23:20:06 UTC

[jira] [Commented] (SQOOP-2340) Map TIMESTAMP to INT96 in Parquet format by default

    [ https://issues.apache.org/jira/browse/SQOOP-2340?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14520274#comment-14520274 ] 

Ryan Blue commented on SQOOP-2340:
----------------------------------

I think Sqoop's mapping is probably correct if it is going to an INT64. The [specification for timestamp|https://github.com/apache/parquet-format/blob/master/LogicalTypes.md] (with millisecond precision) is to use an int64 like java.util.Date. If you need higher precision, we're also working on a microsecond-precision spec. While Hive implements a timestamp stored in an int96, it isn't actually defined anywhere and we are moving away from it.

> Map TIMESTAMP to INT96 in Parquet format by default
> ---------------------------------------------------
>
>                 Key: SQOOP-2340
>                 URL: https://issues.apache.org/jira/browse/SQOOP-2340
>             Project: Sqoop
>          Issue Type: Bug
>    Affects Versions: 1.4.5
>            Reporter: Mariano Dominguez
>
> When importing to HDFS, Sqoop maps TIMESTAMP columns to Parquet INT64 (<> BIGINT).
> It would be great if the default mapping were INT96 (<> TIMESTAMP) instead.



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