Remove unnecessary reshuffle for stateful ParDo after keyed operation
Open
flink
improvement
P3
runners
- Dominant language
- Java
- Stars
- 8.7k
- Forks
- 4.7k
- Avg merge
- 1d 20h
- Merged PRs (30d)
- 196
Description
When we have stateful pardo after GBK / Combine, we can safely assume that partitioning remains consistent and we can safe an extra shuffle.
The use case for this are user defined timers / datadriven triggers. This code path is stressed for example by org.apache.beam.sdk.transforms.ParDoTest.TimerTests#testGbkFollowedByUserTimers.
Imported from Jira [BEAM-11267](https://issues.apache.org/jira/browse/BEAM-11267). Original Jira may contain additional context.
Reported by: dmvk.
Contributor guide
Assessment
This issue has not been assessed yet.