[jira] [Commented] (FLINK-8169) Document timestamp and timezone behavior for Table API & SQL

["Timo Walther (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/FLINK-8169?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16309481#comment-16309481 ] 

Timo Walther commented on FLINK-8169:
-------------------------------------

Hi [~sunjincheng121], this issue should actually only improve the documentation about timestamps and timezones for now. But I agree that this topic has to be addressed in the near future. I will open an issue for it where we can discuss the design.

I don't know if just a global query configuration is enough for all use cases. Usually company work with different timezones simultaneously. We could add support for the new time classes introduced with Java 8 and enable our scalar functions to also work with those (or some custom time class implementations like those from Calcite). 

> Document timestamp and timezone behavior for Table API & SQL
> ------------------------------------------------------------
>
>                 Key: FLINK-8169
>                 URL: https://issues.apache.org/jira/browse/FLINK-8169
>             Project: Flink
>          Issue Type: Improvement
>          Components: Table API & SQL
>            Reporter: Timo Walther
>
> {{java.sql.Date}}, {{java.sql.Time}}, and {{java.sql.Timestamp}} are timezone dependent. However, in Flink's Table & SQL API we remove this dependency and work with UTC timestamps internally. It should be documented how windows use the timestamps in different timezones for event-time. The behavior for processing-time should be explained as well. What is the result if I cast both timestamps to Long, String, or output them to a DataSet/DataStream?



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)