# ACGT String compression and decompression
#
# We compress to fixed length string, 2 bits per symbol -- zero fill in last byte.

my %compress ;	# string of up to 4 characters to compressed item
my @expand ;	  # compressed value to string

BEGIN {
  my @ACGT = qw(A C G T) ;
  my $s = '' ;
  my $v = 0 ;
  foreach my $a (@ACGT) {
    $s = $a ; $compress{$s."\n"} = $compress{$s} = chr($v) ;
    foreach my $b (@ACGT) {
      $s = $a.$b ; $compress{$s."\n"} = $compress{$s} = chr($v) ;
      foreach my $c (@ACGT) {
        $s = $a.$b.$c ; $compress{$s."\n"} = $compress{$s} = chr($v) ;
        foreach my $d (@ACGT) {
          $s = $a.$b.$c.$d ;
          $compress{$s} = chr($v) ;
          $expand[$v++] = $s ;
        } ;
      } ;
    } ;
  } ;

  $compress{"\n"} = '' ;
  $compress{''}   = '' ;
} ;

# Compress ACGT string (all upper case) with or without trailing "\n"
#
#   $compressed = acgt_compress(ACGT_STRING)

sub acgt_compress {
  my $c = '' ;
  $c .= $compress{$_} for unpack('(a4)*', $_[0]) ;
  return $c ;
} ;

# Decompress compressed ACGT, and cut to required length
#
#   $acgt_string = acgt_expand(COMPRESSED, LENGTH)

sub acgt_expand {
  my $e = '' ;
  $e .= $expand[$_] for unpack('C*', $_[0]) ;
  return substr($e, 0, $_[1]) ;
} ;