[
https://issues.apache.org/jira/browse/NIFI-548?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14519852#comment-14519852
]
Mark Payne commented on NIFI-548:
---------------------------------
[~bbende]: not sure. I am not very familiar with this processor at all. It
appears to be far more complex than I would have thought.
I do remember discussing with [~slyouts] about how to design it. The design
that I would have gone with is very different than this, but I'm guessing he
has a good reason for doing this the way he did. It's also very possible that
the API constructs that are there now didn't exist when this processor was
written.
Personally the approach that I would have taken would be to create a
BlockingQueue<ByteBuffer> maybe with a max capacity of 1000 messages or so and
a background thread that just shoves messages onto the queue until it fills.
OnTrigger would simply pull from the queue and create a FlowFile from the
ByteBuffer and move on.
Steve, any insight as to why we used this design? All I remember was that it
was a lot more complicated than I would have thought but I agreed with the
design that you laid out at the time.
> Listen UDP should support generation of a flowfile per datagram
> ---------------------------------------------------------------
>
> Key: NIFI-548
> URL: https://issues.apache.org/jira/browse/NIFI-548
> Project: Apache NiFi
> Issue Type: Improvement
> Components: Extensions
> Reporter: Bryan Bende
> Priority: Minor
>
> Listen UDP presently places udp datagrams into a buffer and then once the
> buffer reaches some trigger size/age it turns that into a flowfile. There
> should be an option to turn each received datagram into a flowfile. This is
> helpful for cases where each datagram is a complete object by itself but that
> isnt necessarily trivially demarcated if it were concatenated with other
> datagrams.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
