Goal: Find THE FASTEST GROWING COUNTY IN THE UNITED STATES over the 1969-2011 timeframe.

USPopulation

Goal: Process the 12 million plus records 
from: http://seer.cancer.gov/popdata/download.html
using: a Scala API atop Cascading, aka SCALDING ( Inventors: Avi Bryant, Oscar Boykin, Argyris )
to find: 
THE FASTEST GROWING COUNTY IN THE UNITED STATES over the 1969-2011 timeframe.
-----------------------------------------------------------------------------
RESULTS: Scroll to the very bottom.

First, the scalding source... 
---
import com.twitter.scalding.{Dsl,Source,TextLine,Job,Args,Tsv,RichPipe}
import cascading.pipe.Pipe

case class USPopulationSource(override val p:String) extends TextLine(p) {
  override def transformForRead(pipe : Pipe) = {
    import Dsl._
    RichPipe(pipe).mapTo('line->('year,'state,'fips,'isWhite,'isBlack, 'isHispanic,'isMale,'age,'population)) {
      record:String =>
      val year:Int             = record.slice(0,0+4).toInt
      val state:String         = record.slice(4,4+2)
      val fips:String          = record.slice(6,6+5)
      val isWhite:Boolean      = record.slice(13,13+1).toInt == 1
      val isBlack:Boolean      = record.slice(13,13+1).toInt == 2
      val isHispanic:Boolean   = record.slice(14,14+1).toInt == 1
      val isMale:Boolean       = record.slice(15,15+1).toInt == 1
      val age:Int              = 5*(record.slice(16,16+2).toInt -1)
      val population:Int       = record.slice(18,18+8).toInt

      (year,state,fips,isWhite,isBlack, isHispanic,isMale,age,population)
    }
  }
}

class PopulationStats(args:Args) extends Job(args) {
  val people = USPopulationSource("pop.txt").read
  val fipspipe = TextLine("US_FIPS_Codes.csv").read.mapTo('line->('state,'county,'fips)) {
    line:String =>
    var arr = line.split(",")
    (arr(0),arr(1),(arr(2)+arr(3)))
  }
  people.groupBy('year, 'fips){
      group => group.plus[Int]('population->'population)
  }.groupBy('fips) {
     val init = (0,0.0d)
     type X = (Int,Double)
     type T = (Int,Int)

     // foldLeft[X,T](fieldDef : (Fields,Fields))(init : X)(fn : (X,T) => X)
      group => group.foldLeft[X,T]( ('population,'year) -> ('dummy,'growth))(init:X) {

        (x:X, t:T) =>
          val (population,year) = t
          val (dummy, growth ) = x
          year match {
          case 1969 => (population, 0.0d)
          case 2011 => (population,(population-dummy)/(dummy+0.0d))
          case _ => if (dummy==0) (population,0.0d) else (dummy,0.0d)
      }
    }
  }.project('fips,'growth)
  .joinWithSmaller(('fips-> 'fips), fipspipe)
  .groupAll(_.sortBy('growth))
  .write(Tsv("growth.txt"))
}
------


RESULTS:
13053  -0.5959766162310867	Georgia	Chattahoochee
38083	-0.5825892857142857	North Dakota	Sheridan
38013	-0.575839766325892	North Dakota	Burke
54047	-0.5703608502224419	West Virginia	McDowell
48101	-0.5474033816425121	Texas	Cottle
38047	-0.5436781609195402	North Dakota	Logan
38023	-0.5362287210824968	North Dakota	Divide
38087	-0.5334632878492528	North Dakota	Slope
38037	-0.5310054184226369	North Dakota	Grant
48301	-0.5276381909547738	Texas	Loving

---- some 3000 counties here ----

13117  10.637015231025215	Georgia	Forsyth
12097	10.880533448053345	Florida	Osceola
48397	10.907133440749963	Texas	Rockwall
13135	11.063892016788289	Georgia	Gwinnett
48157	11.168507788849015	Texas	Fort Bend
48491	11.323119312014695	Texas	Williamson
08117	11.39344262295082	Colorado	Summit
48085	11.507137247655564	Texas	Collin
12035	21.135939986360537	Florida	Flagler
08035	36.49576488706366	Colorado	Douglas

CONCLUSION:
Over the 1969-2011 timeframe, Douglas County experienced the highest population growth ( 3600% )
The top-10 counties with the most pop growth are shown above - they are in Texas, Florida & Colorado.
The botom-10 counties are also showh above - North Dakota expectedly hosts 6 of the bottom 10.

Execution Time: 29 seconds, in Scalding local mode, on my powermac
Scalding FTW!