Github user koeninger commented on the issue:

    https://github.com/apache/spark/pull/15102
  
    This is pretty much the fundamental issue.  Kafka offsets alone aren't
    capable of meeting the SQL Offset interface as defined.  I think that means
    the Offset interface needs to be reconsidered or eliminated.  I think that
    work needs to be done now, before even more work gets put into a
    fundamentally unworkable interface.  I don't see what the rush is, because
    this isn't going to make it into 2.0.1 anyway.
    
    
    On Thu, Sep 22, 2016 at 12:29 PM, Shixiong Zhu <notificati...@github.com>
    wrote:
    
    > PR with failing test indicating at least one reason why it's wrong from an
    > end-user perspective:
    >
    > @koeninger <https://github.com/koeninger> Thanks for writing the test.
    > Yes, we are aware of this issue. However, it's unlikely that we can 
support
    > deleting topics using the current Source API. You can take a look at how
    > StreamExecution checks the new data here: https://github.com/apache/
    > spark/blob/976f3b1227c1a9e0b878e010531285fdba57b6a7/sql/core/src/main/
    > scala/org/apache/spark/sql/execution/streaming/StreamExecution.scala#L320
    >
    > Using the hash code to compare offsets has a potential issue, it may make
    > the latest offset be smaller than the old offset, then StreamExecution
    > won't process the new data.
    >
    > I think one possible solution is StreamExecution doesn't compare the
    > offsets, instead, it just assumes getOffset will always return the latest
    > offset, and it never rollback to an old offset. This needs more discussion
    > anyway. Hence I suggest we don't block this PR for this. Deleting topics
    > can be supported in a later PR when we make an agreement on how to resolve
    > the issue.
    >
    > —
    > You are receiving this because you were mentioned.
    > Reply to this email directly, view it on GitHub
    > <https://github.com/apache/spark/pull/15102#issuecomment-248971360>, or 
mute
    > the thread
    > 
<https://github.com/notifications/unsubscribe-auth/AAGAB0a4SZx4Zjq4hMkI6tSLIDwJLW8Bks5qsrsUgaJpZM4J9QvR>
    > .
    >



---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---

---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org

Reply via email to