#!/bin/bash
Here=$1
Sandbox=$2
Repeats=$3
Use=$4
Pred=$5
SpecialMethods=$6
NormalMethods=$7
ColumnPrunerMethods=$8
Log=$9
let Min=(Use+10)
Drivers="precise proximal"
Trainings="Subset Superset"
Weka="java -Xmx1024M -cp weka.jar"

csvToArff(){
	./csv2arff --ranges rules.config --relation $stem $1;
}

arffWithNums(){
	./cocNums --numbers $1.config $2;
}

generate(){
	for i in *.csv
	do
		mainDataset=${i/_*/}_all.csv;
		stem=${mainDataset/\.*/}
		csvToArff $mainDataset > mainDataset.arff;

		stem=${i/\.*/}
		csvToArff $i > subset.arff
		Size=`./instances subset.arff`;
		[ "$Size" -lt $Min ] && continue
		./blab "\n$stem "

		for ((R = 1; R <= Repeats; R++))
		do
			./blab "$R "
			Seed=$RANDOM

			#generates a test file and two train files
			#one train file is generated from the same subset that the test file was generated (called manual stratification or Subset)
			#the other train file is generated from the superset of the subset that the test file was generated (called Superset)
			gawk -f splitIntoTrainTest.awk Seed=$Seed TestSize=$Use subset.arff mainDataset.arff;

			for Training in $Trainings
			do
				cp train$Training.arff train.arff

				for Num in $Drivers
				do
					arffWithNums $Num test.arff > testN.arff
					arffWithNums $Num train.arff > trainN.arff
					numberOfAttributes=`./attributes testN.arff`

	#---------- No Column Pruner 
					ColumnPruner="None"
					#---------- LC with pre-defined settings
					. $Sandbox/lcSimple testN.arff;
	
					#---------- Only Full Attribute Methods
					for Method in $SpecialMethods
					do
						. $Sandbox/$Method trainN.arff testN.arff;
					done
	
					#---------- Any Attribute Methods
					for Method in $NormalMethods
					do
						. $Sandbox/$Method trainN.arff testN.arff;
					done
	
	#---------- Local Wrapper
					ColumnPruner="LocalWrapper"
					sed 's/ //g' $Num.config > tunings
					./localwrapper tunings train.arff > rankings.dat
					rm -rf subset
					mkdir -p subset/train subset/test
	
					cat rankings.dat train.arff | gawk -f split.awk Stem="subset/train/$stem";
					cat rankings.dat test.arff | gawk -f split.awk Stem="subset/test/$stem";
					
					#---------- Any Attribute Methods For Local Wrapper
					for j in subset/train/*.arff
					do
						numberOfAttributes=`./attributes $j`;
						arffWithNums $Num $j > trainN.arff;
						arffWithNums $Num subset/test/`basename $j` > testN.arff;
						for Method in $NormalMethods;
						do
							. $Sandbox/$Method trainN.arff testN.arff;
						done
					done
	
	#---------- All Other Column Pruners combined with other methods
					for ColumnPruner in $ColumnPrunerMethods
					do
						. $Sandbox/$ColumnPruner
						for Method in $NormalMethods;
						do
							. $Sandbox/$Method $ColumnPruner"_"Train.arff $ColumnPruner"_"Test.arff;
						done
					done
				done
			done
		done
	done
}

generate > $Log