[ 
https://issues.apache.org/jira/browse/FLINK-40847?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Martijn Visser reassigned FLINK-40847:
--------------------------------------

    Assignee: Aleksandr Savonin

> KafkaDynamicSink#equals/hashCode ignore transactionNamingStrategy
> -----------------------------------------------------------------
>
>                 Key: FLINK-40847
>                 URL: https://issues.apache.org/jira/browse/FLINK-40847
>             Project: Flink
>          Issue Type: Bug
>          Components: Connectors / Kafka
>            Reporter: Aleksandr Savonin
>            Assignee: Aleksandr Savonin
>            Priority: Minor
>              Labels: pull-request-available
>
> KafkaDynamicSink.equals() and .hashCode() do not include the field 
> transactionNamingStrategy.
> The field, together with the table option sink.transaction-naming-strategy, 
> was introduced before. That change added the field to the constructor, to 
> copy() and to the KafkaSinkBuilder call, but not to equals/hashCode, which 
> still compare every other configuration field of the sink (delivery 
> guarantee, transactional id prefix, topics, partitioner, flush mode, 
> parallelism, ...).
> KafkaDynamicSource compares all of its configuration fields.
> Impact:
>  * Two KafkaDynamicSink instances that differ only in the naming strategy 
> (INCREMENTING vs POOLING) compare equal.
>  * KafkaDynamicTableFactoryTest#testTableSinkStrategyTranslation and 
> UpsertKafkaDynamicTableFactoryTest#testTableSinkStrategyTranslation assert 
> actualSink.isEqualTo(expectedSink). Because equals ignores the strategy, 
> these tests pass for any value, so a regression in the option translation 
> (e.g. the factory passing TransactionNamingStrategy.DEFAULT instead of the 
> configured value) would not be detected.
>  * No runtime impact: the planner does not compare DynamicTableSink 
> instances, and the runtime sink is built from the field directly in 
> getSinkRuntimeProvider. The sink is @Internal.
>  
> Fix: add transactionNamingStrategy to equals and hashCode. Optionally 
> strengthen the two translation tests by asserting that the actual sink is not 
> equal to an expected sink built with a different strategy, which fails 
> without the fix and passes with it.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to