You are viewing a plain text version of this content. The canonical link for it is here.
Posted to jira@kafka.apache.org by "Valentino Proietti (JIRA)" <ji...@apache.org> on 2018/01/25 15:48:00 UTC

[jira] [Created] (KAFKA-6486) TimeWindows causes unordered calls to windowed aggregation functions

Valentino Proietti created KAFKA-6486:
-----------------------------------------

             Summary: TimeWindows causes unordered calls to windowed aggregation functions
                 Key: KAFKA-6486
                 URL: https://issues.apache.org/jira/browse/KAFKA-6486
             Project: Kafka
          Issue Type: Bug
          Components: streams
    Affects Versions: 1.0.0
            Reporter: Valentino Proietti


This is not a real bug but it causes some weird behaviour, at least in my opinion.

The TimeWindows has a method called windowsFor() that uses and returns an HashMap:

    @Override

    *public* Map<Long, TimeWindow> windowsFor(*final* *long* timestamp) {

        *long* windowStart = (Math._max_(0, timestamp - sizeMs + advanceMs) / advanceMs) * advanceMs;

        *final* Map<Long, TimeWindow> windows = *new* HashMap<>();

        ....

the HashMap does not preserve the order of insertion and this ends up later in calls to any streams windowed aggregation functions that are not ordered by window time as I would expect.

A simple solution is to replace the HashMap with a LinkedHashMap and that's what I did.

Anyway replacing it directly in your code can save hours of debugging to understand what's happening.

Thank you 

 



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)