[BEAM-4689] Reverts change of SDF key type#5834
Conversation
|
Run Dataflow ValidatesRunner |
| shadow library.java.joda_time | ||
| shadow library.java.slf4j_api | ||
| shadow project(path: ":beam-runners-direct-java", configuration: "shadow") | ||
| shadow project(path: ":beam-runners-google-cloud-dataflow-java", configuration: "shadow") |
There was a problem hiding this comment.
I don't think you intended to make this change.
| PipelineOptions options = PipelineOptionsFactory.create(); | ||
|
|
||
| // In order to run your pipeline, you need to make following runner specific changes: | ||
| // |
There was a problem hiding this comment.
I don't think you intended to make this change.
|
Reverted the unrelated changes, PTAL. |
|
Run Dataflow ValidatesRunner |
|
VR test didn't propagate to Github but I found it and it's running here https://builds.apache.org/job/beam_PostCommit_Java_ValidatesRunner_Dataflow_Gradle_PR/5/ |
|
Run Java PreCommit |
|
Run Flink ValidatesRunner |
|
Dataflow SDF tests passed (looking at console output in the link above), and the only error in the Java precommit was a compile error in Apex runner - I'm going to merge this after the Java precommit turns green. |
|
Java precommit is stable. Merging |
|
FWIW I think you may have reverted the unrelated parts locally and not pushed them. I get compile failures in the example locally; not sure how precommit was/is stable. |
|
Curiously enough |
… type [BEAM-4689] Reverts change of SDF key type
3 participants
This stops the bleeding https://issues.apache.org/jira/browse/BEAM-4689
With this change, Dataflow SDFs work, but ULR SDF test also works.
That said, this will not work with SDF implementations in other language SDK harnesses because String is not a well-known type (which was the reason for changing it to byte[]). We'll need to resolve that later.
R: @lukecwik