[
https://issues.apache.org/jira/browse/SPARK-20106?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15943133#comment-15943133
]
Herman van Hovell commented on SPARK-20106:
-------------------------------------------
Caching requires use the backing RDD. That requires we also know the backing
partitions, and this is somewhat special for a global order: it triggers a job
(scan) because we need to determine the partition bounds.
I am closing this as not a problem.
> Nonlazy caching of DataFrame after orderBy/sort
> -----------------------------------------------
>
> Key: SPARK-20106
> URL: https://issues.apache.org/jira/browse/SPARK-20106
> Project: Spark
> Issue Type: Bug
> Components: PySpark, SQL
> Affects Versions: 2.0.1, 2.1.0
> Reporter: Richard Liebscher
> Priority: Minor
>
> Calling {{cache}} or {{persist}} after a call to {{orderBy}} or {{sortBy}} on
> a DataFrame runs not lazy and creates a Spark job:
> {code}spark.range(1, 1000).orderBy("id").cache(){code}
> Other operations do not generate a job when cached:
> {code}spark.range(1, 1000).repartition(2).cache()
> spark.range(1, 1000).groupBy("id").agg(fn.min("id")).cache()
> spark.range(1, 1000).cache(){code}
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
