Spark - java.lang.ClassCastException 处理成 udf 类型为 Array[Array[Map[String,String]]] 的列时

Question

I am concatenating two columns in spark of type Array[Map[String,String]] resulting in a new column of type Array[Array[Map[String,String]]] . However I would like to flatten that column to end up having a columns of type Array[Map[String,String]] with the values of both of the original columns

I have read that from Spark 2.4 it would be possible to apply flatten directly on the concatenation of the columns. Something like this:

df.withColumn("concatenation", flatten(array($"colArrayMap1", $"colArrayMap2")))

However I am still with Spark 2.2, so I need to use a udf for that. This is what I wrote:

def flatten_collection(arr: Array[Array[Map[String,String]]]) = {
    if(arr == null)
        null
    else
        arr.flatten
}
  
val flatten_collection_udf = udf(flatten_collection _)

df.withColumn("concatenation", array($"colArrayMap1", $"colArrayMap2")).withColumn("concatenation", flatten_collection_udf($"concatenation")).show(false)

But I am getting the following error:

Caused by: org.apache.spark.SparkException: Failed to execute user defined function($anonfun$1: (array<array<map<string,string>>>) => array<map<string,string>>)
  at org.apache.spark.sql.catalyst.expressions.GeneratedClass$GeneratedIterator.processNext(Unknown Source)
  at org.apache.spark.sql.execution.BufferedRowIterator.hasNext(BufferedRowIterator.java:43)
  at org.apache.spark.sql.execution.WholeStageCodegenExec$$anonfun$8$$anon$1.hasNext(WholeStageCodegenExec.scala:395)
  at org.apache.spark.sql.execution.SparkPlan$$anonfun$2.apply(SparkPlan.scala:234)
  at org.apache.spark.sql.execution.SparkPlan$$anonfun$2.apply(SparkPlan.scala:228)
  at org.apache.spark.rdd.RDD$$anonfun$mapPartitionsInternal$1$$anonfun$apply$25.apply(RDD.scala:835)
  at org.apache.spark.rdd.RDD$$anonfun$mapPartitionsInternal$1$$anonfun$apply$25.apply(RDD.scala:835)
  at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:49)
  at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:323)
  at org.apache.spark.rdd.RDD.iterator(RDD.scala:287)
  at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:87)
  at org.apache.spark.scheduler.Task.run(Task.scala:109)
  at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:380)
  at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
  at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
  at java.lang.Thread.run(Thread.java:748)
Caused by: java.lang.ClassCastException: scala.collection.mutable.WrappedArray$ofRef cannot be cast to [[Lscala.collection.immutable.Map;

I assume the cast error is happening in the udf, but why and how to avoid it?

Besides if someone knows a solution for Spark 2.2 which doesn't require to use UDF even better

Answer 1

Adapted from the answer here . Seq is needed instead of Array .

def concat_arr(
    arr1: Seq[Map[String,String]],
    arr2: Seq[Map[String,String]]
) : Seq[Map[String,String]] =
{
    (arr1 ++ arr2)
}
val concatUDF = udf(concat_arr _)

val df2 = df.withColumn("concatenation", concatUDF($"colArrayMap1", $"colArrayMap2"))

df2.show(false)
+--------------------+--------------------+----------------------------------------+
|colArrayMap1        |colArrayMap2        |concatenation                           |
+--------------------+--------------------+----------------------------------------+
|[[a -> b], [c -> d]]|[[a -> b], [c -> d]]|[[a -> b], [c -> d], [a -> b], [c -> d]]|
+--------------------+--------------------+----------------------------------------+