[
https://issues.apache.org/jira/browse/NIFI-5794?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16679648#comment-16679648
]
ASF GitHub Bot commented on NIFI-5794:
--------------------------------------
Github user pvillard31 commented on the issue:
https://github.com/apache/nifi/pull/3159
In the end, I only changed the ConsumeKafka processors since it does not
make sense in PublishKafka (one would just have to not set any demarcator to
get the same result: all messages in the flow files are sent as one single
message in Kafka). The main interest of this PR is to increase efficiency when
we consume Avro data from Kafka: there is no need to set a demarcator and we do
want to have multiples messages in the same flow file. I've used this
[template](https://gist.github.com/pvillard31/57e627d206e7965cfb87cc27c09caf44)
to test the change and to confirm the behavior: when pulling Avro data from
Kafka with "empty string" as demarcator to have multiple messages in the same
flow file, we are able to process the records in the next processor using a
ConvertRecord for instance.
> ConsumeKafka and PublishKafka should allow empty string demarcator
> ------------------------------------------------------------------
>
> Key: NIFI-5794
> URL: https://issues.apache.org/jira/browse/NIFI-5794
> Project: Apache NiFi
> Issue Type: Improvement
> Components: Extensions
> Reporter: Pierre Villard
> Assignee: Pierre Villard
> Priority: Major
>
> ConsumeKafka(_*) and PublishKafka(_*) processors should allow "empty string"
> as a message demarcator. This would allow consuming Avro data without the
> serialization/de-serialization cost while still allowing the use of Record
> processors once the data is in NiFi.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
