About Jackson

[bromine0x23]

Since flink-runtime includes a org.apache.flink:flink-shaded-jackson dependency, JavaNetHttpPollingClient introduces and use the original Jackson dependency (in com.fasterxml.jackson).

Is this necessary? Introducing original version may cause class loading issues in some situations.
Should JavaNetHttpPollingClient switch to the classes provided by flink-shaded-jackson? Other classes like ObjectMapperAdapter and GenericJsonAndUrlQueryCreator is using shaded version.

[davidradl]

@bromine0x23 I looked at the Flink Kafka connector it does https://github.com/apache/flink-connector-kafka/blob/3c6991d86389450fe2dafac1d685c87e6ca82756/pom.xml#L437

It sets the jackson bom level https://github.com/apache/flink-connector-kafka/blob/3c6991d86389450fe2dafac1d685c87e6ca82756/pom.xml#L69C3-L69C52

So in the Kafka case the parent pom brings in a compiled jackson dependancy which is then imported into the kafka submodule.

Also JDBC connector
https://github.com/apache/flink-connector-jdbc/blob/669314d46d8900c6490083854dea1b71f482a67d/pom.xml#L271
also uses this import style .

If we were to change the http connector to have a provided dependancy on the Flink shaded jackson, then we would need to
change all our imports to refer to this shaded version. Anything else in the dependancy tree would not pick up the Jackson dependancy, I wonder if this might be the reason for this approach.

I will raise this on the Flink connector Slack channel. We should agree on how we should do this across the connectors and would like to understand the thinking behind the current approach. Note this connector has been donated to Flink as flink-connectors-http - we are close to releasing it.

[bromine0x23]

JDBC connector actually not uses Jackson, while CDC is using both:

• https://github.com/apache/flink-cdc/blob/release-3.5.0/flink-cdc-connect/flink-cdc-pipeline-connectors/flink-cdc-pipeline-connector-doris/src/main/java/org/apache/flink/cdc/connectors/doris/sink/DorisRowConverter.java#L32
• https://github.com/apache/flink-cdc/blob/release-3.5.0/flink-cdc-connect/flink-cdc-pipeline-connectors/flink-cdc-pipeline-connector-oceanbase/src/main/java/org/apache/flink/cdc/connectors/oceanbase/sink/OceanBaseRowConvert.java#L32
• https://github.com/apache/flink-cdc/blob/release-3.5.0/flink-cdc-connect/flink-cdc-source-connectors/flink-connector-postgres-cdc/src/main/java/org/apache/flink/cdc/connectors/postgres/table/PostgreSQLDeserializationConverterFactory.java#L25
• https://github.com/apache/flink-cdc/blob/release-3.5.0/flink-cdc-connect/flink-cdc-source-connectors/flink-connector-mysql-cdc/src/main/java/org/apache/flink/cdc/connectors/mysql/source/utils/RecordUtils.java#L29

This seems to lack best practice guidance.
Anyway, I think it should at least be using only one of them for consistency.