flink-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From dyanarose <...@git.apache.org>
Subject [GitHub] flink issue #5295: [FLINK-8384] [streaming] Session Window Assigner with Dyn...
Date Wed, 31 Jan 2018 15:06:29 GMT
Github user dyanarose commented on the issue:

    https://github.com/apache/flink/pull/5295
  
    I'm not the biggest fan of unchecked casts, but testing this in our POC environment casting
the existing EventTimeTrigger to a typed Trigger<T, W> is working. So if the unchecked
cast is acceptable, that would get rid of all changes required to Event/ProcessingTimeTrigger
    
    On your second point. (sorry if I mis-attribute this based on github profile name) I believe
you had mentioned breaking this out into new classes when I first brought this up on the mailing
list https://lists.apache.org/thread.html/6ceb094460bca8e9731e9e1dc0bb479f407f769458bff30c412adf78@%3Cdev.flink.apache.org%3E
    
    Now that you see it in action, do you feel it would be better off as an addition to the
existing session window classes?
    
    The way I see it is, if I put withGapExtractor() on the existing classes, without adding
type information to them, then the extract() method on SessionWindowGapExtractor will need
to have the signature of extract(Object input) leaving the implementer to have to cast to
the input type.
    
    I have to admit it feels strange to me that the Window assigners all drop input type information.
But that would mean that these new typed assigners would be the odd ones out.


---

Mime
View raw message