##################################################################
# find the k-th nearest historial projects near the generated projects

function euclidean(row1,row2,data1,data2,    n,col,d,d1,d2,key,ignorep,i) {
	split("",ignorep,"")
	for (key in Klasses)
		ignorep[Klasses[key]] = 1
	for(col=1;col<=Cols;col++)
	    if (!(col in ignorep)) {
			d1 = normalize(data1,col,data1[row1,col])
			d2 = normalize(data2,col,data2[row2,col])
			d += abs(d1 - d2)^2
			n++
	}
	return  sqrt(d)/sqrt(n)
}
function distance(row1,row2,data1,data2,memo,  d) {
	d = as100(euclidean(row1,row2,data1,data2))
	memo[-1 * d] = row1  # d started at row1
	memo[d]      = row2  # d ended at row2
	return d
}
function normalize(data,col,n,      min,max,d) { 
	min = data["min",col]
	max = data["max",col]
	d = min == max ? 1 : (n - min) /(max - min) 
	return d
}
function neighbors(news,new,olds,old,neighbor,memo,   o,n) {
	for(n=1;n<=news;n++) 
		for(o=1;o<=olds;o++) 
			push2(distance(n,o,new,old,memo),neighbor,n)
}
function knn(k,news,neighbor,memo,ks,    dist,n,most,i,d,sorted) {
	for(n=1;n<=news;n++) {
		most = neighbor[n,0]
		for(i = 1;i <=  most;i++)  
			dist[++d] = neighbor[n,i]
	}
	knnDebug(dist,memo)
	asort(dist,sorted)
	for(i=1;i<=d;i++)  {
		n  = memo[sorted[i]]
		if ( ++ks[n]==1 ) k--
		if (k == 0)  return i
    }	
	return k
}