[
https://issues.apache.org/jira/browse/FLINK-40421?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18110138#comment-18110138
]
Hirson Zhang commented on FLINK-40421:
--------------------------------------
Hi [~dianfu] ,
I have completed the implementation work for FLINK-40422 and would like to
take FLINK-40421 as my next contribution.
I reviewed the sorting section in FLIP-591. I plan to implement
`DataFrame.sort()` as a lazy transformation backed by the existing
`Table.order_by()` and `Expression.asc`/`desc` APIs. The implementation will
support multiple columns or expressions, per-column `descending` and
`nulls_first` options, and explicit NULL placement using existing expression
APIs such as `is_null`, without modifying the planner or runtime.
One semantic detail is how to handle an Expression that already has `.asc` or
`.desc`: should its ordering always be overridden according to the
`descending` option, even when the option is omitted？
Could you please confirm the intended behavior and assign the issue to me if
appropriate?
Thank you!
> Add sorting support in DataFrame API
> ------------------------------------
>
> Key: FLINK-40421
> URL: https://issues.apache.org/jira/browse/FLINK-40421
> Project: Flink
> Issue Type: Sub-task
> Components: API / Python
> Reporter: Liu Liu
> Priority: Major
>
> Add ordering transformations:
> {code:java}
> def sort(
> self,
> by,
> *,
> descending=False,
> nulls_first=None,
> ) -> DataFrame {code}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
