CODEWITHSUNDEEP
scalaapache-sparkapache-spark-sql
user299791
user299791

Reputation: 2070

How to combine (join) information across an Array[DataFrame]

I have an Array[DataFrame] and I want to check, for each row of each data frame, if there is any change in the values by column. Say I have the first row of three data frames, like:

 (0,1.0,0.4,0.1)
 (0,3.0,0.2,0.1)
 (0,5.0,0.4,0.1)

The first column is the ID, and my ideal output for this ID would be:

 (0, 1, 1, 0)

meaning that the second and third columns changed while the third did not. I attach here a bit of data to replicate my setting

val rdd = sc.parallelize(Array((0,1.0,0.4,0.1),
                               (1,0.9,0.3,0.3),
                               (2,0.2,0.9,0.2),
                               (3,0.9,0.2,0.2),
                               (4,0.3,0.5,0.5)))
val rdd2 = sc.parallelize(Array((0,3.0,0.2,0.1),
                                (1,0.9,0.3,0.3),
                                (2,0.2,0.5,0.2),
                                (3,0.8,0.1,0.1),
                                (4,0.3,0.5,0.5)))
val rdd3 = sc.parallelize(Array((0,5.0,0.4,0.1),
                                (1,0.5,0.3,0.3),
                                (2,0.3,0.3,0.5),
                                (3,0.3,0.3,0.1),
                                (4,0.3,0.5,0.5)))
val df = rdd.toDF("id", "prop1", "prop2", "prop3")
val df2 = rdd2.toDF("id", "prop1", "prop2", "prop3")
val df3 = rdd3.toDF("id", "prop1", "prop2", "prop3")
val result:Array[DataFrame] = new Array[DataFrame](3)
result.update(0, df)
result.update(1,df2)
result.update(2,df3)

How can I map over the array and get my output?

Upvotes: 3

Views: 478

Answers (2)

marios
marios

Reputation: 8996

First we need to join all the DataFrames together.

val combined = result.reduceLeft((a,b) => a.join(b,"id"))

To compare all the columns of the same label (e.g., "prod1"), I found it easier (at least for me) to operate on the RDD level. We fist transform the data into (id, Seq[Double]).

val finalResults = combined.rdd.map{
  x => 
    (x.getInt(0), x.toSeq.tail.map(_.asInstanceOf[Double]))
}.map{ 
  case(i,d) => 
     def checkAllEqual(l: Seq[Double]) = if(l.toSet.size == 1) 0 else 1
     val g = d.grouped(3).toList 
     val g1 = checkAllEqual(g.map(x => x(0)))
     val g2 = checkAllEqual(g.map(x => x(1)))
     val g3 = checkAllEqual(g.map(x => x(2)))
     (i, g1,g2,g3)
}.toDF("id", "prod1", "prod2", "prod3")

finalResults.show()

This will print:

+---+-----+-----+-----+
| id|prod1|prod2|prod3|
+---+-----+-----+-----+
|  0|    1|    1|    0|
|  1|    1|    0|    0|
|  2|    1|    1|    1|
|  3|    1|    1|    1|
|  4|    0|    0|    0|
+---+-----+-----+-----+

Upvotes: 1

zero323
zero323

Reputation: 330433

You can use countDistinct with groupBy:

import org.apache.spark.sql.functions.{countDistinct}

val exprs = Seq("prop1", "prop2", "prop3")
  .map(c => (countDistinct(c) > 1).cast("integer").alias(c))

val combined = result.reduce(_ unionAll _)

val aggregatedViaGroupBy = combined
  .groupBy($"id")
  .agg(exprs.head, exprs.tail: _*)

aggregatedViaGroupBy.show
// +---+-----+-----+-----+
// | id|prop1|prop2|prop3|
// +---+-----+-----+-----+
// |  0|    1|    1|    0|
// |  1|    1|    0|    0|
// |  2|    1|    1|    1|
// |  3|    1|    1|    1|
// |  4|    0|    0|    0|
// +---+-----+-----+-----+

Upvotes: 2

Related Questions