Skip to content

Instantly share code, notes, and snippets.

@krishnanraman
Last active June 13, 2017 03:47
  • Star 6 You must be signed in to star a gist
  • Fork 1 You must be signed in to fork a gist
Star You must be signed in to star a gist
Save krishnanraman/4696053 to your computer and use it in GitHub Desktop.
Goal: Find THE FASTEST GROWING COUNTY IN THE UNITED STATES over the 1969-2011 timeframe.
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!
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment