You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@spark.apache.org by "eugen yushin (JIRA)" <ji...@apache.org> on 2019/08/16 13:39:00 UTC

[jira] [Commented] (SPARK-28742) StackOverflowError when using otherwise(col()) in a loop

    [ https://issues.apache.org/jira/browse/SPARK-28742?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16909052#comment-16909052 ] 

eugen yushin commented on SPARK-28742:
--------------------------------------

Looks like the issue is diff in logic between LocalRelation (used for data frames) and LogicalRDD (used for DF created from RDD)

```

val df2 = Seq("1").toDF("c1")

df.explain(true)

df2.explain(true)

```

 

> StackOverflowError when using otherwise(col()) in a loop
> --------------------------------------------------------
>
>                 Key: SPARK-28742
>                 URL: https://issues.apache.org/jira/browse/SPARK-28742
>             Project: Spark
>          Issue Type: Bug
>          Components: SQL
>    Affects Versions: 2.4.0, 2.4.3
>            Reporter: Ivan Tsukanov
>            Priority: Major
>
> The following code
> {code:java}
> val rdd = sparkContext.makeRDD(Seq(Row("1")))
> val schema = StructType(Seq(
>   StructField("c1", StringType)
> ))
> val df = sparkSession.createDataFrame(rdd, schema)
> val column = when(col("c1").isin("1"), "1").otherwise(col("c1"))
> (1 to 9).foldLeft(df) { case (acc, _) =>
>   val res = acc.withColumn("c1", column)
>   res.take(1)
>   res
> }
> {code}
> falls with
> {code:java}
> java.lang.StackOverflowError
>    at org.codehaus.janino.CodeContext.flowAnalysis(CodeContext.java:395)
>    ...{code}
> Probably, the problem is spark generates unexplainable big Physical Plan - 
> {code:java}
> val rdd = sparkContext.makeRDD(Seq(Row("1")))
> val schema = StructType(Seq(
>   StructField("c1", StringType)
> ))
> val df = sparkSession.createDataFrame(rdd, schema)
> val column = when(col("c1").isin("1"), "1").otherwise(col("c1"))
> val result = (1 to 9).foldLeft(df) { case (acc, _) =>
>   acc.withColumn("c1", column)
> }
> result.explain()
> {code}
> it shows a plan 18936 symbols length
> {code:java}
> == Physical Plan ==
> *(1) Project [CASE WHEN (CASE WHEN (CASE WHEN (CASE WHEN (CASE WHEN (CASE WHEN (CASE .... 18936 symbols
> +- Scan ExistingRDD[c1#1]  {code}



--
This message was sent by Atlassian JIRA
(v7.6.14#76016)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org