#RDBParse

package RDBParse;
require Exporter;
use FileHandle;
STDERR->autoflush(1);
use English;
use File::Basename;
use lib dirname($PROGRAM_NAME);
use ParseCompare;
@ISA = qw(Exporter);
@EXPORT = qw( RDBParseFiles );

sub RDBParseFiles (@)
{

    $compare_location = "/projects/compbio/lib/2nd-compare/";
    
  @rdb_files = @_;

  foreach $RDBFILE (@rdb_files)
  {
    open RDBFILE or die "ERROR: Couldn't open RDB file $RDBFILE\n";

    # NOTE: This part is ugly - we read in each compare file, even though
    # the main program has to do the same thing for mutual info.

    $RDBFILE =~ /\.(\w*)-?\w*\.rdb$/;   # get alphabet from .alph(-abc).rdb
    $alphabet = $1;
    push @alphabets, $alphabet;

    @compare = ParseCompareFile ($compare_location .
       "t2k." . $alphabet . "-dssp_ehl2.compare");
    $alph_weights{$alphabet} = $compare[1];
    %joint_prob = %{$compare[2]};

    $comments .= "# Comments from $RDBFILE\n";
    $comments .= "# ============================================\n";
    # look for and save comments
    while($line = <RDBFILE>)
    {
      chomp($line);
      last if (!($line =~ /^#/));
      $comments .= $line . "\n";
    }

    # split into an array, all but the first two elements are alphabet letters
    # use these letters as indicies into the hash table

    @letters = split("\t", $line);

    # scan and discard the next line (length information)
    $line = <RDBFILE>;

    # create a new RDB line from each line in the file

    $line_num = 0;
    while($line = <RDBFILE>)
    {
      chomp($line);

      if ($line =~ /^#/)
      {
        $comments .= $line . "\n";
        next;
      }

      @line = split ("\t", $line);
      $line_num++;
      $E_prob = $H_prob = $L_prob = 0;

      for ($i = 2; $i <= $#line; $i++)
      {
        $prob = $line[$i];
        $E_prob += $prob * ${$joint_prob{$letters[$i]}}[0];
        $H_prob += $prob * ${$joint_prob{$letters[$i]}}[1];
        $L_prob += $prob * ${$joint_prob{$letters[$i]}}[2];
      }

      $total_prob = $E_prob + $H_prob + $L_prob;   # Renormalize to remove errors
      $E_prob /= $total_prob;
      $H_prob /= $total_prob;
      $L_prob /= $total_prob;

      @{ $output[$line_num]{$alphabet} } =
        ($line[0], $line[1], $E_prob, $H_prob, $L_prob);
    }
    close (RDBFILE);
    $comments .= "# ============================================\n";
  }

  return ($line_num, $comments, \%alph_weights, \@output);

}