[
https://issues.apache.org/jira/browse/BEAM-11715?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17275460#comment-17275460
]
Yifan Mai commented on BEAM-11715:
----------------------------------
>From offline sync:
* I will implement "rewrite translations.pack_combiners so that the packed
combine is expanded to GroupByKey and CombineValues" as a short term fix. This
will remove the need for PTransformOverride and thus the from_runner_api()
call. This should also fix portable Dataflow. This will be a blocker for 2.28.
* I will remove the from_runner_api() call. This means that the original
non-proto pipeline will be unmodified, so non-portable Dataflow will _not_ get
pipeline optimizations. This is OK since portable Dataflow will be the
preferred way of doing things going forward.
> Combiner packing creates an incorrect proto
> -------------------------------------------
>
> Key: BEAM-11715
> URL: https://issues.apache.org/jira/browse/BEAM-11715
> Project: Beam
> Issue Type: Bug
> Components: sdk-py-core
> Reporter: Chamikara Madhusanka Jayalath
> Assignee: Yifan Mai
> Priority: P0
> Fix For: 2.28.0
>
> Time Spent: 1h
> Remaining Estimate: 0h
>
> Seems like optimization creates a CombinePerKey transform that does not have
> any sub-transforms. We should fix this by preserving the old structure for
> CombinePerKey (GBK+CombineValues composite).
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
