Connect: Remove redundant JSON formats and migrate logic to Transformation Script

[tenthe]

Body

Following the introduction of the output collector in the preprocessing script (#4097), maintaining multiple specialized JSON formats is no longer necessary. All array-based formatting can now be handled via the preprocessing script.

Goals

• Remove the following JSON formats from AdapterDescription:

  • Json Array
  • Json Array Field
  • GeoJson

• Implement a migration script to move existing logic from the Format Parser to the Transformation/Preprocessing script.

StreamPipes Committer

I acknowledge that I am a maintainer/committer of the Apache StreamPipes project.

[tenthe]

Upon further review, we need to retain the Json Array format.

The transform method requires events to be injected specifically as a Map. By keeping this format, we ensure users can explicitly define whether the adapter should process standard JSON objects or arrays of objects. This distinction is necessary to maintain compatibility with the transformation logic, which expects a Map<String, Object> structure for each processed event.

[tenthe]

I implemented the feature in PR #4109.

The original plan was to remove both the GeoJson and JSONArrayField parsers. However, we discovered that JSONArrayField is still required for the FileReplayAdapter to function correctly.

If JSONArrayField were removed, FileReplayAdapter would need to handle two different input scenarios in

protected void processEvent(
    IEventCollector collector,
    Map<String, Object> event
)

This becomes especially problematic when replaying events based on the original timestamp, as the timestamp field would need to be extracted and the replay initialized accordingly.

Additionally, removing JSONArrayField could cause the Schema Editor in the Connect transformation step to become excessively large for events with large payloads.

For these reasons, we decided to keep the JSONArrayField option for the JSON parser.