guoweiM edited a comment on pull request #12132:
URL: https://github.com/apache/flink/pull/12132#issuecomment-629650094
Hi, @aljoscha thanks for your review and suggestion.
1. The reason I used Operator to generate snapshots at the time is that we
could use a set of code to generate both v1 and v2 snapshots of `BucketState`.
1. Otherwise, I would have to write twice for how to generate
`BucketStatev1` and `BucketStateV2`.
1. There is a special process also needs to be written twice. The path
in the snapshot cannot have an absolute path and needs to be specially
processed when the snapshot is generated.
1. Another issue is that the `BucketStateSerializerTest` would change the
file in the resource directory. For example, the BucketStateSerializerTest
would recover and commit the pending files. This would change the output files
in the resources directory. This means only the first time the test would check
the recovery and commit. If we want to test the recovery and commit every time
we would have to copy the scenario directory every test.
All these problems have already resolved in the `BucketStateMigrationTest`.
Do you think the following action is ok?
1. Change `BucketStateMigrationTest` to a update test. Let it could read
both v1&v2 snapshots.
2. Remove all the v1 state in the `BucketStateSerializerTest`.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
