Re^2: Challenge: "Words" In A String

[gone2015]

A trie ! Beautiful :-)

For repeated use it's obvious to consider preprocessing the dictionary into the trie, and reading that each time.

Of interest is how beautifully this trie will compress, by folding the matching sub-tries. Code to illustrate this given below. On my machine the result is:

  Loading dictionary '2of4brif.txt'... done -- 1.30 Secs
  Writing trie '2of4brif.trie'... 22599/138676 done -- 1.35 Secs
  Reading trie '2of4brif.trie'... done -- 0.25 Secs
  Walking the tries... done -- 1.94 Secs
  Original trie is: 27.1M, new trie is: 5.4M

showing the degree of compression (22599/138676), and that reading the trie takes 0.25 Secs where loading the dictionary takes 1.30 Secs. Result :-)

(The memory footprint is also reduced, but that's not hugely significant.)

---

  use strict;
  use warnings;
  
  use constant DICT  => "2of4brif.txt";
  use constant TRIE  => "2of4brif.trie" ;
  
  sub get_cpu {
    my $t = (times)[0] ;
    if (@_) { $t = sprintf('%4.2f Secs', $t - $_[0]) ; } ;
    return $t ;
  } ;
  
  my $took ;
  
  print STDERR "Loading dictionary '", DICT, "'" ;
  $took = get_cpu() ;
  my $trie  = load_dict(DICT) ;
  print STDERR "... done -- ", get_cpu($took), "\n" ;
  
  print STDERR "Writing trie '", TRIE, "'" ;
  $took = get_cpu() ;
  my ($now, $was) = write_trie(TRIE, $trie) ;
  print STDERR "... done -- ", get_cpu($took), "\n" ;
  
  print STDERR "Reading trie '", TRIE, "'" ;
  $took = get_cpu() ;
  my $check = read_trie(TRIE) ;
  print STDERR "... done -- ", get_cpu($took), "\n" ;
  
  print STDERR "Walking the tries" ;
  $took = get_cpu() ;
  walk('', $trie, $check) ;
  print STDERR "... done -- ", get_cpu($took), "\n" ;

  use Devel::Size qw(total_size) ;

  printf STDERR "Original trie is: %3.1fM, new trie is: %3.1fM\n",
                      total_size( $trie)/(1024 * 1024),
                      total_size($check)/(1024 * 1024) ;

  #===================================================================
+======================
  # Loading the Dictionary into the Trie.
  
  sub load_dict {
    my ($d_name) = @_ ;
  
    # Constructs a trie from the dictionary.
  
    open(my $fh, '<', $d_name)
           or die("Unable to open dictionary '$d_name': $!\n");
  
    my $trie = undef ;
  
    while (<$fh>) {
      s/\s+$// ;  # chomp;
      my $p = \$trie;
      for ( split(//, $_), '!' ) {
        $p = \( $$p->{$_} ) ;
      }
    }
  
    return $trie ;
  }
  
  #===================================================================
+======================
  # Writing the Trie.
  
  my %node_map ;
  my $idx ;
  
  my $node_count ;
  
  sub write_trie {
    my ($t_name, $trie) = @_ ;
  
    open(my $fh, '>', $t_name)
           or die("Unable to create trie '$t_name': $!\n");
  
    $node_count = 0 ;
  
    $idx = 1 ;        # Index 0 reserved for what '!' points at !
    %node_map = ('!' => "$idx") ; # Preset end of word node
  
    write_node($fh, $trie) ;
  
    close $fh ;
  
    return ($idx, $node_count) ;
  } ;
  
  sub write_node {
    my ($fh, $node) = @_ ;
  
    $node_count++ ;
  
    my @chs = sort keys %$node ;  # NB '!' sorts to the front !
  
    my @n = $chs[0] eq '!' ? (shift @chs) : () ;
  
    foreach my $ch (@chs) {
      my $p = write_node($fh, $node->{$ch}) ;
      push @n, $ch.$p ;
    } ;
  
    my $n = join(' ', @n) ;
    my $p ;
    unless (defined($p = $node_map{$n})) {
      $p = $node_map{$n} = sprintf('%X', ++$idx) ;
            # Assign new index and record
      print $fh $n, "\n" ;    # Output new node
    } ;
  
    return $p ;
  } ;
  
  #===================================================================
+======================
  # Reading the Trie.
  
  my @nodes ;
  
  sub read_trie {
    my ($t_name) = @_ ;
  
    open(my $fh, '<', $t_name)
           or die("Unable to open trie '$t_name': $!\n");
  
    @nodes = (undef, {'!' => undef}) ;
  
    while (<$fh>) {
      push @nodes, { map { my ($c, $p) = unpack('AA*', $_) ;
                     ($c, $nodes[hex($p)]) } split } ; # Note hex('') 
+= 0
    } ;
  
    close $fh ;
  
    my $trie = pop @nodes ;
    @nodes = () ;
  
    return $trie ;
  } ;
  
  #===================================================================
+======================
  # Walk two tries to check they are identical

  sub walk {
    my ($w, $ra, $rb) = @_ ;
  
    my @ac = sort keys %$ra ;
    my @bc = sort keys %$rb ;
  
    if (@ac != @bc) { die "node length mismatch \@ '$w': (@ac) vs (@bc
+)" ; } ;
    for my $i (0..$#ac) {
      if ($ac[$i] ne $bc[$i]) { die "node mismatch \@ '$w': (@ac) vs (
+@bc)" ; } ;
    } ;
  
    foreach my $ch (@ac) {
      my $ad = $ra->{$ch} ;
      my $bd = $rb->{$ch} ;
      if ($ch eq '!') {
        if (defined($ad)) { die "'!' with defined down ($ad) in 'a' \@
+ '$w'" ; } ;
        if (defined($bd)) { die "'!' with defined down ($bd) in 'b' \@
+ '$w'" ; } ;
      }
      else {
        walk($w.$ch, $ad, $bd) ;
      } ;
    } ;
  } ;