--- trunk/run.pl	2006/04/17 16:50:53	434
+++ trunk/run.pl	2006/05/15 17:23:38	509
@@ -12,11 +12,13 @@
 use WebPAC::Input 0.03;
 use WebPAC::Store 0.03;
 use WebPAC::Normalize::XML;
+use WebPAC::Normalize::Set;
 use WebPAC::Output::TT;
 use YAML qw/LoadFile/;
 use Getopt::Long;
 use File::Path;
 use Time::HiRes qw/time/;
+use File::Slurp;
 
 =head1 NAME
 
@@ -48,6 +50,16 @@
 
 path to YAML configuration file
 
+=item --force-set
+
+force conversion C<< normalize->path >> in C<config.yml> from
+C<.xml> to C<.pl>
+
+=item --stats
+
+disable indexing and dump statistics about field and subfield
+usage for each input
+
 =back
 
 =cut
@@ -59,6 +71,8 @@
 my $config = 'conf/config.yml';
 my $debug = 0;
 my $only_db_name;
+my $force_set = 0;
+my $stats = 0;
 
 GetOptions(
 	"limit=i" => \$limit,
@@ -68,6 +82,8 @@
 	"only=s" => \$only_db_name,
 	"config" => \$config,
 	"debug" => \$debug,
+	"force-set" => \$force_set,
+	"stats" => \$stats,
 );
 
 $config = LoadFile($config);
@@ -77,9 +93,15 @@
 die "no databases in config file!\n" unless ($config->{databases});
 
 my $log = _new WebPAC::Common()->_get_logger();
+$log->info( "-" x 79 );
 
 my $use_indexer = $config->{use_indexer} || 'hyperestraier';
-$log->info("using $use_indexer indexing engine...");
+if ($stats) {
+	$log->debug("option --stats disables update of indexing engine...");
+	$use_indexer = undef;
+} else {
+	$log->info("using $use_indexer indexing engine...");
+}
 
 my $total_rows = 0;
 my $start_t = time();
@@ -90,32 +112,32 @@
 
 	my $indexer;
 
-	my $indexer_config = $config->{$use_indexer} || $log->logdie("can't find '$use_indexer' part in confguration");
-	$indexer_config->{database} = $database;
-	$indexer_config->{clean} = $clean;
-	$indexer_config->{label} = $db_config->{name};
-
-	# important: clean database just once!
-	$clean = 0;
-
-	if ($use_indexer eq 'hyperestraier') {
-
-		# open Hyper Estraier database
-		use WebPAC::Output::Estraier '0.10';
-		$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
-	
-	} elsif ($use_indexer eq 'kinosearch') {
-
-		# open KinoSearch
-		use WebPAC::Output::KinoSearch;
-		$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path});
-		$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } );
+	if ($use_indexer) {
+		my $indexer_config = $config->{$use_indexer} || $log->logdie("can't find '$use_indexer' part in confguration");
+		$indexer_config->{database} = $database;
+		$indexer_config->{clean} = $clean;
+		$indexer_config->{label} = $db_config->{name};
+
+		if ($use_indexer eq 'hyperestraier') {
+
+			# open Hyper Estraier database
+			use WebPAC::Output::Estraier '0.10';
+			$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
+		
+		} elsif ($use_indexer eq 'kinosearch') {
+
+			# open KinoSearch
+			use WebPAC::Output::KinoSearch;
+			$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path});
+			$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } );
 
-	} else {
-		$log->logdie("unknown use_indexer: $use_indexer");
+		} else {
+			$log->logdie("unknown use_indexer: $use_indexer");
+		}
+
+		$log->logide("can't continue without valid indexer") unless ($indexer);
 	}
 
-	$log->logide("can't continue without valid indexer") unless ($indexer);
 
 	#
 	# now WebPAC::Store
@@ -175,6 +197,7 @@
 			offset => $offset,
 			lookup => $lookup,
 			recode => $input->{recode},
+			stats => $stats,
 		);
 		$log->logdie("can't create input using $input_module") unless ($input);
 
@@ -191,18 +214,34 @@
 			prefix => $input->{name},
 		);
 
+		my $rules;
 		my $normalize_path = $input->{normalize}->{path};
 
+		if ($force_set) {
+			my $new_norm_path = $normalize_path;
+			$new_norm_path =~ s/\.xml$/.pl/;
+			if (-e $new_norm_path) {
+				$log->debug("--force-set replaced $normalize_path with $new_norm_path");
+				$normalize_path = $new_norm_path;
+			} else {
+				$log->debug("--force-set failed on $new_norm_path, fallback to $normalize_path");
+			}
+		}
+
 		if ($normalize_path =~ m/\.xml$/i) {
 			$n->open(
 				tag => $input->{normalize}->{tag},
-				xml_file => $input->{normalize}->{path},
+				xml_file => $normalize_path,
 			);
 		} elsif ($normalize_path =~ m/\.(?:yml|yaml)$/i) {
 			$n->open_yaml(
 				path => $normalize_path,
 				tag => $input->{normalize}->{tag},
 			);
+		} elsif ($normalize_path =~ m/\.(?:pl)$/i) {
+			$n = undef;
+			$log->info("using WebPAC::Normalize::Set to process $normalize_path");
+			$rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!";
 		}
 
 		foreach my $pos ( 0 ... $input_db->size ) {
@@ -217,20 +256,38 @@
 				push @{ $row->{'000'} }, $pos;
 			}
 
-			my $ds = $n->data_structure($row);
+				
+			my $ds;
+			if ($n) {
+				$ds = $n->data_structure($row);
+			} else {
+				$ds = WebPAC::Normalize::Set::data_structure(
+					row => $row,
+					rules => $rules,
+					lookup => $lookup->lookup_hash,
+				);
+
+				$db->save_ds(
+					id => $mfn,
+					ds => $ds,
+					prefix => $input->{name},
+				) if ($ds && !$stats);
+			}
 
 			$indexer->add(
 				id => $input->{name} . "/" . $mfn,
 				ds => $ds,
 				type => $config->{$use_indexer}->{type},
-			);
+			) if ($indexer);
 
 			$total_rows++;
 		}
 
+		$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats);
+
 	};
 
-	eval { $indexer->finish } if ($indexer->can('finish'));
+	eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
 
 	my $dt = time() - $start_t;
 	$log->info("$total_rows records indexed in " .