Open panasenco opened 4 years ago
As a workaround I could obviously just not do that and either keep the circular insert as the last piece in a topology or create a new intermediate stream:
CREATE STREAM countdown (i INT) WITH (kafka_topic='countdown', value_format='json', partitions=1);
CREATE STREAM countdown2 (i INT) WITH (kafka_topic='countdown2', value_format='json', partitions=1);
INSERT INTO countdown2 SELECT i FROM countdown EMIT CHANGES;
INSERT INTO countdown2 SELECT i-1 AS i FROM countdown2 WHERE i > 0 EMIT CHANGES;
I'm not sure we even want to be allowing an INSERT INTO
itself. This is just going to create a circular workload that would overwhelm ksqlDB. I think the bug here is actually not failing on that statement.
Describe the bug All circular inserts in one stream don't get picked up in a different stream.
To Reproduce ksqlDB version 0.9.0
SQL Statements:
Input JSON:
Output JSON:
Expected behavior Test passes.
Actual behaviour
Additional context Removing the "countdown2" topic outputs makes the test pass.
Running the statements manually in ksqlDB results in the same rows in both streams.
Full output: