Update nullable columns to NULL is not reported correctly

[Oli180369]

If a nullable column is updated with NULL the new value is not reported as NULL, much more the old value is reported as the new value.

[thake]

Can you provide a PR with a reproducable failing unit test? As a starting point you can have a look at the existing unit tests. Oli180369 <[email protected]> schrieb am Do., 4. März 2021, 09:02:

…

If a nullable column is updated with NULL the new value is not reported as NULL, much more the old value is reported as the new value. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#14>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AC7XPEKHXDOVKNTPVFOS4ATTB45A7ANCNFSM4YSX47QQ> .

[Oli180369]

Seems that we have an issue with the logminer itself

[Oli180369]

We checked the tables on which the logminer-kafka-connect component is working. These tables were filled correctly. There has to be an issue with the component. We try to improve the test

[Oli180369]

It seems that we have a problem within the ConnectSchemaFactrory at line 99 where the after struct is enriched by before values for absent keys. At this point the component can not distinguish an absence from an NULL-Update because the Hashmap will return NULL in both cases. So the after struct is always enriched by the before struct for columns which are updated to NULL.
I was not able to provide a test for this issue so far, but I hope you have understood.

[thake]

@Oli180369 Thanks for this patch! A tricky bug.. ;)

I'll prepare a new release just now.