\#!/bin/bash

# $1 - Location of directory to cross-validate the learners on.
# $2 - Location of the output of the comparison.
# $3 - Location of the Tar3 and Which learners.
# $4 - Sampling rate.

## For each data file that was created in the new directory.
files=`ls $1/*.data`;
for file in $files
do
  ## Do an 10way cross-validation on the data.
  stem=`basename $file .data`;
  echo "Creating $4 Train-Test sets with $1/$stem.data to test the learners on."
  ./makeTrainTest seed=$RANDOM sample=$4 file=$stem $1/$stem.data;

  # For each of the 10 samples run Which and Tar3 on 90% train, then test their rules
  # on 10% test.  Score them and store that score along with other statistics.
  for i in 0 1 2 3 4 5 6 7 8 9
  do
    ## Create some local temp files.
    cp $1/$stem".cfg" train$i".cfg";
    cp $1/$stem".cfg" test$i$".cfg";
    cp $1/$stem".names" train$i".names";
    cp $1/$stem".names" test$i".names";
    mv train$i$stem train$i".data";
    mv test$i$stem test$i".data";

    echo "   Tar3 pass $i on $stem.";
    $3t3 train$i | ./getRules Pass=2 > trules$i;
    echo "   Which pass $i on $stem.";
    $33 100 train$i l 1 | ./getRules Pass=1 > wrules$i;

    ## Compare Which and Tar3 at each of the 10-runs.
    echo "   Comparing using test$i";
    ./compare test$i wrules$i trules$i $2$stem l $i > comLog;

    ## Clean up the temp files.
    rm -f trules$i;
    rm -f wrules$i;
    rm -f train$i*
    rm -f test$i*

  done
  ## Sort the 10 runs on each data set by how great Which's performance over Tar3 was.
  echo "Sorting the compare'd output information by difference( Which - Tar ).";
  ./plotsort $2$stem;
done