A Scala script that takes a CSV as input on stdin, produces a pivot table on two columns using a count() function and outputs the pivot table to stdout.

CSVPivot.scala

import collection.mutable
import io.Source
import java.io.FileInputStream

def toIntOption(s: String): Option[Int] = try { Some(s.toInt) } catch { case e: NumberFormatException => None }

val intArgs = args.map(toIntOption).flatten

if (args.length != 2 && intArgs.length != 2) {
  System.err.println("usage: <x_column_number> <y_column_number>")
  sys.exit(1)
}

// You can change these if you need different behaviour. 
val containsColumnHeaders = true
val containsRowHeaders = false

val inputStream = System.in
//val inputStream = new FileInputStream("survey.csv")

val lines = Source.fromInputStream(inputStream).getLines().toBuffer

var cells = lines map { _.split(',').toBuffer }

if (containsColumnHeaders)
  cells = cells.drop(1)

if (containsRowHeaders) {
  cells = cells map { _.drop(1) }
}

val argToIndex = if (containsRowHeaders) 2 else 1

val (column1Index, column2Index) = (intArgs(0) - argToIndex, intArgs(1) - argToIndex)

val map = new mutable.HashMap[(String, String), Int]

def pairInRow(row: mutable.Buffer[String]): (String, String) = (row(column1Index), row(column2Index))

// You could change the pivot aggregation fuctnion from count() to soemthing else here 
// by changing 'rows.size' to some other calculation, e.g. (rows.map{_.toInt}).sum
val pairCounts: Map[(String, String), Int] = (for ((pair, rows) <- (cells groupBy pairInRow)) yield (pair, rows.size)).toMap

val outputColumns = (pairCounts.keys map {_._1}).toSet.toSeq.sorted

val outputRows = (pairCounts.keys map {_._2}).toSet.toSeq.sorted

for (column <- outputColumns) {
  print(',')
  print(column)
}
println()

for (row <- outputRows) {
  print(row)
  for (column <- outputColumns) {
    print(',')
    print(pairCounts.getOrElse((column, row), ""))
  }
  println()
}