#!/usr/bin/gawk -f

BEGIN  { 
	TrainNormal="train.arff";
	TrainManualStratification="trainManualStratification.arff";
	TestNormal="test.arff"; 
	TestSize = 5;
	Seed = 1;
} 

/^[ \t]*$/ {
	next;
} 

NR==1{
	Seed ? srand(Seed) : srand(1)      
	printf "" > TrainNormal;  
	printf "" > TrainManualStratification;
	printf "" > TestNormal;
}

FILENAME==ARGV[3] && /@relation/,/@data/ { 
	print $0 >> TrainNormal;
	print $0 >> TrainManualStratification;
	print $0 >> TestNormal; 
	next; 
}

FILENAME==ARGV[4] && /@relation/,/@data/ { 
	next; 
}

FILENAME==ARGV[3] {
	InstanceNormal[rand()] = $0;
}

FILENAME==ARGV[4] {
	InstanceManualStratification[rand()] = $0;
}

END { 
	for (i in InstanceNormal)
	{
		if (TestSize > 0)
		{
			#store the ones for test file and print them to the test file as well
			TestArray[TestSize] = InstanceNormal[i];
			print InstanceNormal[i] >> TestNormal;
			TestSize--;
		}

		else
		{
			#print the rest to the train file
			print InstanceNormal[i] >> TrainNormal;
		}
	}

	for (i in InstanceManualStratification)
	{
		UsedBefore = 0;
		#check to see if that instance is in the test file. If not, it can be in the train file used for manual stratification
		for (j in TestArray)
		{
			if (InstanceManualStratification[i] == TestArray[j]) 
			{
				UsedBefore = 1;
			}
		}
		#if not used for test file, it can be used for training	
		if (UsedBefore == 0)
			print InstanceManualStratification[i] >> TrainManualStratification;
	}
}