R/codebook.R

In pdc: Permutation Distribution Clustering

codebook <-
function(x,m=3,t=1, use.fast=T, normalized=T)
{
	if (class(x) != "numeric") {
		stop("X must be of class numeric!")
	}

	if (t*(m-1) >= length(x)) {
		return(NA);
	}
	
	# branch to fast C implementation, if possible
	
 	 if (use.fast) {
	  if (m==2) {
		distribution <- ( .Call("fast_codebook2",x,t) )	 	
	  } else if (m==3) {
		distribution <-( .Call("fast_codebook3",x,t) )
	  } else if (m==4) {
	 	distribution <-( .Call("fast_codebook4", x,t))
	  } else if (m==5) {
	 	distribution <-( .Call("fast_codebook5", x,t))
	  } else if (m==6) {
	 	distribution <-( .Call("fast_codebook6", x,t))
	  } else if (m==7) {
	 	distribution <- ( .Call("fast_codebook7", x,t))
	  }
	 } 

	if (!use.fast | m>7) {
	 

	# generic codeword function
	codeword.func <- codeword;

	distribution <- rep.int(0, factorial(m))
	to <- (length(x)-t*(m-1))
	for(i in 1:to)
	{

		data <- x[seq.int(i,i+t*(m-1),t)]
		
		# skip data containing NA
		if	(any(is.na(data))) { next; }
		
		# calculate permutation index
		number <- codeword.func(data, m);
				
		distribution[number] <- distribution[number] + 1
	}
	
	}
	
	if (normalized) {	
		return(distribution/sum(distribution))
	} else {
		return(distribution)
	}
}