--- trunk/run.pl	2006/05/15 17:49:01	511
+++ trunk/run.pl	2006/06/29 21:18:50	541
@@ -11,14 +11,15 @@
 use WebPAC::Lookup;
 use WebPAC::Input 0.03;
 use WebPAC::Store 0.03;
-use WebPAC::Normalize::XML;
-use WebPAC::Normalize::Set;
+use WebPAC::Normalize;
 use WebPAC::Output::TT;
+use WebPAC::Validate;
 use YAML qw/LoadFile/;
 use Getopt::Long;
 use File::Path;
 use Time::HiRes qw/time/;
 use File::Slurp;
+use MARC::Record 2.0;	# need 2.0 for utf-8 encoding see marcpm.sf.net
 
 =head1 NAME
 
@@ -53,16 +54,15 @@
 
 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
 
 disable indexing and dump statistics about field and subfield
 usage for each input
 
+=item --validate path/to/validation_file
+
+turn on extra validation of imput records, see L<WebPAC::Validation>
+
 =back
 
 =cut
@@ -74,8 +74,8 @@
 my $config = 'conf/config.yml';
 my $debug = 0;
 my $only_filter;
-my $force_set = 0;
 my $stats = 0;
+my $validate_path;
 
 GetOptions(
 	"limit=i" => \$limit,
@@ -85,8 +85,8 @@
 	"only=s" => \$only_filter,
 	"config" => \$config,
 	"debug" => \$debug,
-	"force-set" => \$force_set,
 	"stats" => \$stats,
+	"validate=s" => \$validate_path,
 );
 
 $config = LoadFile($config);
@@ -98,6 +98,11 @@
 my $log = _new WebPAC::Common()->_get_logger();
 $log->info( "-" x 79 );
 
+my $validate;
+$validate = new WebPAC::Validate(
+	path => $validate_path,
+) if ($validate_path);
+
 my $use_indexer = $config->{use_indexer} || 'hyperestraier';
 if ($stats) {
 	$log->debug("option --stats disables update of indexing engine...");
@@ -109,13 +114,14 @@
 my $total_rows = 0;
 my $start_t = time();
 
+my @links;
+my $indexer;
+
 while (my ($database, $db_config) = each %{ $config->{databases} }) {
 
-	my ($only_database,$only_input) = split(m#/#, $only_filter);
+	my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter);
 	next if ($only_database && $database !~ m/$only_database/i);
 
-	my $indexer;
-
 	if ($use_indexer) {
 		my $indexer_config = $config->{$use_indexer} || $log->logdie("can't find '$use_indexer' part in confguration");
 		$indexer_config->{database} = $database;
@@ -182,7 +188,7 @@
 
 	foreach my $input (@inputs) {
 
-		next if ($only_input && $input->{name} =~ m#$only_input#i || $input->{type} =~ m#$only_input#i);
+		next if ($only_input && ($input->{name} !~ m#$only_input#i && $input->{type} !~ m#$only_input#i));
 
 		my $type = lc($input->{type});
 
@@ -190,11 +196,13 @@
 
 		my $lookup = new WebPAC::Lookup(
 			lookup_file => $input->{lookup},
-		);
+		) if ($input->{lookup});
 
 		my $input_module = $config->{webpac}->{inputs}->{$type};
 
-		$log->info("working on input '$input->{name}' in $input->{path} [type: $input->{type}] using $input_module lookup '$input->{lookup}'");
+		$log->info("working on input '$input->{name}' in $input->{path} [type: $input->{type}] using $input_module",
+			$input->{lookup} ? "lookup '$input->{lookup}'" : ""
+		);
 
 		my $input_db = new WebPAC::Input(
 			module => $input_module,
@@ -210,67 +218,58 @@
 		my $maxmfn = $input_db->open(
 			path => $input->{path},
 			code_page => $input->{encoding},	# database encoding
+			%{ $input },
 		);
 
-		my $n = new WebPAC::Normalize::XML(
-		#	filter => { 'foo' => sub { shift } },
-			db => $db,
-			lookup_regex => $lookup->regex,
-			lookup => $lookup,
-			prefix => $input->{name},
-		);
+		my @norm_array = ref($input->{normalize}) eq 'ARRAY' ?
+			@{ $input->{normalize} } : ( $input->{normalize} );
 
-		my $rules;
-		my $normalize_path = $input->{normalize}->{path};
+		foreach my $normalize (@norm_array) {
 
-		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");
-			}
-		}
+			my $rules;
+			my $normalize_path = $normalize->{path} || $log->logdie("can't find normalize path in config");
 
-		if ($normalize_path =~ m/\.xml$/i) {
-			$n->open(
-				tag => $input->{normalize}->{tag},
-				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: $!";
-		}
+			$log->logdie("Found '$normalize_path' as normalization file which isn't supported any more!") unless ( $normalize_path =~ m!\.pl$!i );
 
-		foreach my $pos ( 0 ... $input_db->size ) {
+			my $rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!";
 
-			my $row = $input_db->fetch || next;
+			$log->info("Using $normalize_path for normalization...");
 
-			my $mfn = $row->{'000'}->[0];
+			my $marc_fh;
+			if (my $path = $normalize->{marc21}) {
+				open($marc_fh, '>', $path) ||
+					$log->logdie("can't open MARC output $path: $!");
 
-			if (! $mfn || $mfn !~ m#^\d+$#) {
-				$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
-				$mfn = $pos;
-				push @{ $row->{'000'} }, $pos;
+				$log->info("Creating MARC export file $path\n");
 			}
 
-				
-			my $ds;
-			if ($n) {
-				$ds = $n->data_structure($row);
-			} else {
-				$ds = WebPAC::Normalize::Set::data_structure(
+			# reset position in database
+			$input_db->seek(1);
+
+			foreach my $pos ( 0 ... $input_db->size ) {
+
+				my $row = $input_db->fetch || next;
+
+				my $mfn = $row->{'000'}->[0];
+
+				if (! $mfn || $mfn !~ m#^\d+$#) {
+					$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
+					$mfn = $pos;
+					push @{ $row->{'000'} }, $pos;
+				}
+
+
+				if ($validate) {
+					my @errors = $validate->validate_errors( $row );
+					$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors);
+				}
+
+					
+				my $ds = WebPAC::Normalize::data_structure(
 					row => $row,
 					rules => $rules,
-					lookup => $lookup->lookup_hash,
+					lookup => $lookup ? $lookup->lookup_hash : undef,
+					marc_encoding => 'utf-8',
 				);
 
 				$db->save_ds(
@@ -278,20 +277,31 @@
 					ds => $ds,
 					prefix => $input->{name},
 				) if ($ds && !$stats);
+
+				$indexer->add(
+					id => $input->{name} . "/" . $mfn,
+					ds => $ds,
+					type => $config->{$use_indexer}->{type},
+				) if ($indexer && $ds);
+
+				if ($marc_fh) {
+					my $marc = new MARC::Record;
+					$marc->encoding( 'utf-8' );
+					$marc->add_fields( WebPAC::Normalize::_get_marc21_fields() );
+					print $marc_fh $marc->as_usmarc;
+				}
+
+				$total_rows++;
 			}
 
-			$indexer->add(
-				id => $input->{name} . "/" . $mfn,
-				ds => $ds,
-				type => $config->{$use_indexer}->{type},
-			) if ($indexer);
+			$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats);
 
-			$total_rows++;
-		}
+			# close MARC file
+			close($marc_fh) if ($marc_fh);
 
-		$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats);
+		}
 
-	};
+	}
 
 	eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
 
@@ -308,12 +318,12 @@
 	if (ref($db_config->{links}) eq 'ARRAY') {
 		foreach my $link (@{ $db_config->{links} }) {
 			if ($use_indexer eq 'hyperestraier') {
-				$log->info("adding link $database -> $link->{to} [$link->{credit}]");
-				$indexer->add_link(
+				$log->info("saving link $database -> $link->{to} [$link->{credit}]");
+				push @links, {
 					from => $database,
 					to => $link->{to},
 					credit => $link->{credit},
-				);
+				};
 			} else {
 				$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
 			}
@@ -322,3 +332,7 @@
 
 }
 
+foreach my $link (@links) {
+	$log->info("adding link $link->{from} -> $link->{to} [$link->{credit}]");
+	$indexer->add_link( %{ $link } );
+}