Postpone-bucket getWriteId misparses the write id when data-file.prefix or the commit user contains the -s- separator

[LuciferYang]

Search before asking

• I searched in the issues and found no similar issues.

Paimon version

master (1.5-SNAPSHOT)

Compute Engine

Flink (postpone-bucket write path).

Minimal reproduce step

1. Create a postpone-bucket table and set data-file.prefix to a value that contains the -s- separator (for example a-s-b).
2. Write data so postpone-bucket data files are produced. Each file name is {data-file.prefix}-u-{commitUser}-s-{writeId}-w-{uuid}-{count}.
3. Read or compact the table, which routes splits by getWriteId(fileName) % parallelism.

What doesn't meet your expectations?

PostponeBucketFileStoreWrite.getWriteId recovered the write id by splitting on the first -s- in the file name. When data-file.prefix (or a custom commit user) already contains -s-, the first -s- is not the writer's separator, so the parse either throws Data file name ... does not match the pattern or returns a wrong write id. A wrong id then routes the split to the wrong reader under getWriteId(...) % parallelism.

Anything else?

The correct anchor is the -w- writer marker that always terminates the interpolated prefix, so the write id is the segment between the last -s- before -w- and -w- itself.

Are you willing to submit a PR?

• I'm willing to submit a PR!