--- trunk/run.pl	2006/04/17 15:10:04	431
+++ trunk/run.pl	2006/05/15 13:15:01	507
@@ -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 +50,15 @@
 
 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
+
+dump statistics about used fields and subfields in each input
+
 =back
 
 =cut
@@ -59,6 +70,8 @@
 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 +81,8 @@
 	"only=s" => \$only_db_name,
 	"config" => \$config,
 	"debug" => \$debug,
+	"force-set" => \$force_set,
+	"stats" => \$stats,
 );
 
 $config = LoadFile($config);
@@ -105,6 +120,7 @@
 
 		# open KinoSearch
 		use WebPAC::Output::KinoSearch;
+		$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path});
 		$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } );
 
 	} else {
@@ -171,6 +187,7 @@
 			offset => $offset,
 			lookup => $lookup,
 			recode => $input->{recode},
+			stats => $stats,
 		);
 		$log->logdie("can't create input using $input_module") unless ($input);
 
@@ -187,18 +204,34 @@
 			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 ) {
@@ -213,7 +246,12 @@
 				push @{ $row->{'000'} }, $pos;
 			}
 
-			my $ds = $n->data_structure($row);
+			my $ds = $n ? $n->data_structure($row) :
+				WebPAC::Normalize::Set::data_structure(
+					row => $row,
+					rules => $rules,
+					lookup => $lookup->lookup_hash,
+				);
 
 			$indexer->add(
 				id => $input->{name} . "/" . $mfn,
@@ -224,8 +262,12 @@
 			$total_rows++;
 		}
 
+		$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats);
+
 	};
 
+	eval { $indexer->finish } if ($indexer->can('finish'));
+
 	my $dt = time() - $start_t;
 	$log->info("$total_rows records indexed in " .
 		sprintf("%.2f sec [%.2f rec/sec]",