Rcmd::Clustering - Interfaces to clustering algorithms of R language.

    This class is a part of G-language Genome Analysis Environment, 
    collecting interfaces to clustering algorithms of R language.

    &opt_default(output => 'show', label => [], filename => 'hclust.pdf', method => 'ward', distmethod => 'correlation');

    my @args= opt_get(@_);
    my $output= opt_val('output');
    my @label=  @{opt_val('label')};
    my $filename= opt_val('filename');
    my $method= opt_val('method');
    my $distmethod= opt_val('distmethod');

    my @all_methods= ('ward', 'single', 'complate', 'average', 'centroid', 'median', 'mcquitty');
    unless (grep /^$method$/, @all_methods) {
        $method= 'ward';
    }
    $method = 'ward.D2' if ($method = 'word');

    my @all_dist_methods= ('euclidean', 'maximum', 'manhattan', 'canberra', 'binary', 'pearson',
                           'abspearson', 'correlation', 'abscorrelation', 'spearman', 'kendall');
    unless (grep /^$distmethod$/, @all_dist_methods) {
        $distmethod= 'pearson';
    }

    my $rcmd = Rcmd->new();

    my (%data_table, @R_names);
    for my $i (0 .. $#args) {
        my $R_name= 'array'.$i;
        push @R_names, $R_name;

        $data_table{$R_name}= $_[$i];
        $rcmd->array($R_name, @{$args[$i]});
    }

    # 'label' list object (if required)
    $rcmd->sarray('label', @label) if $#label > -1;
    my @R_commands= (
                     'CMP <- complete.cases('.join(', ', @R_names).')',
                     ('d.table <- data.frame('.$R_names[0].'=1:'.($#{$data_table{$R_names[0]}}+1).', row.names=label)')x!! ($#label > -1),
                     ('d.table <- data.frame('.$R_names[0].'=1:'.($#{$data_table{$R_names[0]}}+1).')')x!!                  ($#label == -1)
		     );

    for my $key (@R_names) {
        push @R_commands, $key.' <- '.$key.'[CMP]';
        push @R_commands, 'd.table$'.$key.' <- '.$key;
    }

    $rcmd->exec(
                @R_commands,

                'library("amap")',
                "pdf('./graph/".$filename."')",
                'hc <- hclust(Dist(d.table, method="'.$distmethod.'"), method="'.$method.'")',
                'plot(hc)',
               );

    msg_gimv('graph/'.$filename) if $output eq 'show';

    return '';

    my $rcmd = new Rcmd();
    $rcmd->set_mode('tmp');

    opt_default("iter.max"=>10, filename=>"kmeans.pdf", output=>"show", sampledata=>0, centers=>5);
    my @args = opt_get(@_);
    my $centers = opt_val("centers");
    my $iter = opt_val("iter.max");
    my $output = opt_val("output");
    my $filename = opt_val("filename");
    my $sampledata = opt_val("sampledata");
    my $label = opt_val("label") || '';

    if($sampledata){
	$rcmd->sample_data_for_clustering();
    }else{
	$rcmd->set_clust_data(@args, -label=>$label);
    }

    my @result = $rcmd->exec(
			     'require(stats)',
			     "rclust.kmeans<-kmeans(rclust,$centers,$iter)",
			     'rclust.kmeans$cluster'
			     );

    if($output =~ /g/ || $output =~ /show/){
	$rcmd->exec(
		    "pdf('./graph/".$filename."')",
		    'plot(rclust,col=rclust.kmeans$cluster)',
		    "points(rclust.kmeans\$centers, col=1:$centers,pch=8)"
		    );

	msg_gimv("graph/$filename") if ($output =~ /show/);
    }
    
    $rcmd->set_mode();
    return @result;

    my $rcmd = shift;

    $rcmd->exec(
		'rclust<-rbind(matrix(rnorm(100,sd=0.3),ncol=2),' .
		'matrix(rnorm(100,mean=1,sd=0.3),ncol=2))' 
		);

    my $rcmd = shift;

    my @args = opt_get(@_);
    my $label = opt_val("label") || '';

    my $flag = 0;
    foreach (@args){
	if ($flag == 0){
	    $flag ++;

	    $rcmd->array('rclust', @$_);
	}else{
	    $rcmd->array('tmp', @$_);
	    $rcmd->exec('rclust <- rbind(rclust, tmp)');
	}
    }

    if (length $label){
	$rcmd->sarray('label', @$label);
	$rcmd->exec('dimnames(rclust) <- list(label, NULL)');
    }

    my $rcmd = new Rcmd();
    $rcmd->set_mode('tmp');

    opt_default(filename=>"som.pdf", output=>"show", xdim=>3, ydim=>3, sampledata=>0, topo=>'hexa', neigh=>'gaussian');
    my @args       = opt_get(@_);
    my $xdim       = opt_val("xdim");
    my $ydim       = opt_val("ydim");
    my $filename   = opt_val("filename");
    my $output     = opt_val("output");
    my $sampledata = opt_val("sampledata");
    my $label      = opt_val("label") || '';
    my $topo       = opt_val('topo');
    my $neigh      = opt_val('neigh');

    if($sampledata){
	$rcmd->sample_data_for_clustering();
    }else{
	$rcmd->set_clust_data(@args, -label=>$label);
    }

    $rcmd->exec(
		'require(som)',
		"rclust.som<-som(rclust, $xdim, $ydim, topo='hexa', neigh='gaussian')",
		'rclust.som$visual'
		);

    my @result;
    open(FILE, $rcmd->{log}) || die($!);
    while(<FILE>){
	if(/qerror/){
	    @result = ();
	    while(<FILE>){
		chomp;
		if(/^\d/){
		    my (undef, $x, $y, $qerror) = split(/\s+/, $_, 4);
		    push(@result, [$x, $y, $qerror]);
		}
	    }
	}
    }
    close(FILE);

    if($output =~ /g/ || $output =~ /show/){
	$rcmd->exec(
                    "pdf('./graph/".$filename."')",
		    'plot(rclust.som)'
		    );

	msg_gimv("graph/$filename") if ($output =~ /show/);
    }
    
    $rcmd->set_mode();
    return @result;