Handling Empty Events with latest_by_offset() in Transition Detection

ksqlDB
1

Hi all,

I’m encountering an issue using latest_by_offset() where empty events are sometimes generated in my stream. Here’s the scenario:

{‘key’: ‘0’, ‘x’: ‘A’}
{‘key’: ‘0’, ‘x’: ‘A’}
{‘key’: ‘0’, ‘x’: ‘B’}
{‘key’: ‘0’, ‘x’: ‘A’}
{‘key’: ‘0’, ‘x’: ‘A’}

I want to capture events only when there is a transition between values, like when ‘A’ changes to ‘B’ and vice versa. For the example above, I expect only two events:

{‘key’: ‘0’, ‘curr_x’: ‘B’}
{‘key’: ‘0’, ‘curr_x’: ‘A’}

my current ksqlDB setup looks like:

CREATE STREAM input_s (
    key VARCHAR KEY,
    x VARCHAR)
WITH (kafka_topic='input');

CREATE TABLE change_t
	WITH (kafka_topic='changed') AS
	SELECT
		s.key AS id,
		latest_by_offset(s.x) AS curr_x
	FROM input_s AS s
	GROUP BY s.key
	HAVING latest_by_offset(s.x) <> latest_by_offset(s.x, 2)[1];

After a value transition (e.g., the last event {'key': '0', 'x': 'A'}), an empty event is created in the changed topic. I would like to prevent this behavior and ensure only valid transitions are emitted.
(SELECT * FROM aggregated_t EMIT CHANGES; shows a record with ‘<TOMBSTONE>’ as value for ‘xcurr_’.)

Does anyone have suggestions on how to fix this?

Thanks in advance for your help!

Best,
Oliver

2

This topic was automatically closed after 30 days. New replies are no longer allowed.