#!/usr/bin/perl -w
#
# Author:  Grant Thiltgen
#
# This script is to collect output data from the multiple neural nets
# run to test our training protocol.

use strict;
use English;
use File::Basename;
use Getopt::Long;
use Pod::Usage;

my $numnets=3;  # Number of neural nets to look at.
my $txxalign;   # The alignment, for example t05-thin90
my $dataset;    # The dataset, for example dunbrack-30pc-1763
my $arch;       # The architecture of the neural net.
my $alphaname;  # The name of the alphabet used in neural net training
my $data;       # Data used, for example tr12 or t1c2
GetOptions(
	      "numnets=i" => \$numnets
	    , "txxalign=s" => \$txxalign
	    , "dataset=s" => \$dataset
	    , "arch=s" => \$arch
            , "alphaname=s" => \$alphaname
	    , "data=s" => \$data
	    , "help|?"	=>	sub {pod2usage("verbose"=>1);}
	    , "man"		=> 	sub {pod2usage("verbose"=>2);}
	    )
	   or pod2usage("verbose" => 0);

# Building loops to create the data file.

my $line;       # Information from the files brought into this.
my $value;      # Value of hash for foreach loop.
my %trainbits;  # storing training data in bits
my @temp;       # temporary array for storing data in the line.
my ($a,$b);     # temp values for sorting.
my $i=0;        # temp value for sorting.
my $key;        # value for key in hash.
my @nets;       # Value used to store the nets for checking.
my $average=0;  # Value for storing average of trained networks.

    while ($line = <STDIN>)
    {   
	chomp $line;
	$nets[$i++] = $line;
    }

    print STDOUT "Net\tValue\n";
    print STDOUT "6S\t7N\n";
    for (my $i=0; $i < $numnets; $i++)
    {
	open (TRAIN, "$nets[$i]/quality-reports/$dataset-$txxalign-$arch-$data-$alphaname-mult250-from-empty.train") or print STDERR "Can't open $data training file for $nets[$i]\n";

	while ($line = <TRAIN>)
	{
	    chomp $line;
	    if ($line =~ /^#/)
	    {
		next;
	    }
	    else
	    {
		@temp = split /\s+/, $line;
		$trainbits{$nets[$i]} = $temp[3];
	    }
	}	    
	close TRAIN;
	$average += $trainbits{$nets[$i]};
    }
    $average = $average / $numnets;
    print STDOUT "avg\t$average\n";
    $i = 0;
    open (NETS, ">networks/$dataset-$txxalign-$arch-$data-$alphaname-mult250.nets");
    foreach $key (sort highlow keys %trainbits)
    {
	if ($i++ < 1)
	{
	    print NETS "$key\n";
	}
	print STDOUT "$key\t$trainbits{$key}\n";
    }

sub highlow ($$)
{
    my ($a,$b) = @_;
    $trainbits{$b} <=> $trainbits{$a};
}



__END__