[
https://issues.apache.org/jira/browse/SPARK-3098?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14110183#comment-14110183
]
Matei Zaharia commented on SPARK-3098:
--------------------------------------
Sorry, I don't understand -- what exactly is the bug here? There's no guarantee
about the ordering of elements in distinct(). If you're relying on zipWithIndex
creating specific values, that's a wrong assumption to make. The question is
just whether the *set* of elements returned by zipWithIndex is correct.
I don't think we should change our randomize() to be more deterministic here
just because you want zipWithIndex. We have to allow shuffle fetches to occur
in a random order, or else we can get inefficiency when there are hotspots. If
you'd like to make sure values land in specific partitions and in a specific
order in each partition, you can partition the data with your own Partitioner,
and run a mapPartitions that sorts them within each one.
> In some cases, operation zipWithIndex get a wrong results
> ----------------------------------------------------------
>
> Key: SPARK-3098
> URL: https://issues.apache.org/jira/browse/SPARK-3098
> Project: Spark
> Issue Type: Bug
> Components: Spark Core
> Affects Versions: 1.0.1
> Reporter: Guoqiang Li
> Priority: Critical
>
> The reproduce code:
> {code}
> val c = sc.parallelize(1 to 7899).flatMap { i =>
> (1 to 10000).toSeq.map(p => i * 6000 + p)
> }.distinct().zipWithIndex()
> c.join(c).filter(t => t._2._1 != t._2._2).take(3)
> {code}
> =>
> {code}
> Array[(Int, (Long, Long))] = Array((1732608,(11,12)), (45515264,(12,13)),
> (36579712,(13,14)))
> {code}
--
This message was sent by Atlassian JIRA
(v6.2#6252)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
