Zongwen Li

There is currently no particularly good way to support metadata columns. If metadata columns are required, you may only add one option for each metadata column; solution proposal for metadata...

MetadataColumn and SupportReadingMetadata are indeed designed to solve this problem, but it has not been completed yet, and some pre-problems need to be solved； #2490

I agree with you, because some of the data integration features are not supported in Spark and Flink engines. Of course, this is a challenging thing, and I look forward...

> > > Can it be like this? > > No, it's different. The above writer has multiple input subtasks, which is to reduce the resource usage on the sink...

@EricJoy2048 Now how do we deal with the union of multiple upstream operators? That's the same thing as the following Users need `union (all)` post-processing, and we cannot separate them

Currently, each new version of engine support needs to add a `translation` and `starter` module. The `starter` module is more coupled with the old API and may need to be...

Job fault-tolerant(Checkpoint): https://github.com/apache/incubator-seatunnel/issues/2274

> The design proposals above are detailed designs. As a rookie, I am really interested in the following questions: > > 1. What's the background of the new engine? What's...

We can provide simple soure and sink, masking some advanced features.

I can't understand what `PartitionTransform` does in the diagram. If there is a mechanism for `Transform4` to distribute data to `PartitionTransform`, then why don't we distribute to `Sink` directly? Wouldn't...