#!/bin/bash

# $1 - Location of the directory of directories of the data files. ( ../Data/arffData/discrete/ )
# $2 - Location of the Tar3 and Which learners ( ../Src/ )
# $3 - The sample rate of the crossval ( 10 )
# $4 - "i" for equal interval discretization.
#      "f" for equal frequency discretization.
#      "n" for no discretization. ( This uses the learners' built in discretization methods )
# $5 - If we are discretizing, this will be the range of numbers to discretize on.
#      IE "2 4 8 16 32" would mean discretize with 2 bins, 4, bins, etc.
#      If $4 = n, this should just be 1.
# $6 - Location to put just the totals into.

cc compare.c -o com;
cc plotsort.c -o plotsort;
cc sumOfDiff.c -o sumOfDiff;
cc sortSod.c -o sortSod;


## Create the directory to put the sum of difference tables into.
sodDir=./sodData/;
echo "Sum of Diff Directory = $sodDir";
#mkdir $sodDir;

## For each bin number
for d in $5
do
  ## Convert and discretize this data with $d bins.
  dir=../Data/whichData/discrete/$d"binsEq"$4;
  echo "Data Directory = $dir";
  #mkdir $dir;
  #./converter $1 $dir $4 $d

  ## Run Dave Owen's Tar3 Configuration script on the data file.
  ## This will allow us to make sure Tar3 is optimal.
  #./cfgsim.sh $dir/$stem $2/t3 > nowhere;
  #rm -f nowhere;

  ## Cross-validate the data.
  ## Create 10 train-test sets and run each through the learner.
  resDir=results/discrete/$d"binsEq"$4/;
  echo "Results Directory = $resDir";
  mkdir $resDir;
  ./xval $dir $resDir $2 $3;
    
  ## Plot the data generated by the $3 runs with $d bins
  ## Create a directory to put the plots into.
  plotDir=./plotData/discrete/$d"binsEq"$4;
  echo "Plot directory = $plotDir";
  mkdir $plotDir;
  ./genplots $resDir $plotDir/;

  ## Using the files in $resDir, create a win/loss table for each one.
  echo "Tallying Sum Of Differences for $d bins";
  ./sumOfDiff `ls $resDir/*` $sodDir/discrete/$d"binsEq"$4;
  
  ## Sort the sumOfDiff table by which's wins.
  #./sortSod $sodDir/$d"binsEq"$4 "$d Bin Totals:" $6;
done