--- trunk/run.pl	2006/09/26 10:23:04	712
+++ trunk/run.pl	2007/05/27 10:50:39	850
@@ -7,12 +7,12 @@
 use lib './lib';
 
 use WebPAC::Common 0.02;
-use WebPAC::Parser 0.04;
-use WebPAC::Input 0.13;
-use WebPAC::Store 0.10;
-use WebPAC::Normalize 0.11;
+use WebPAC::Parser 0.08;
+use WebPAC::Input 0.16;
+use WebPAC::Store 0.14;
+use WebPAC::Normalize 0.22;
 use WebPAC::Output::TT;
-use WebPAC::Validate 0.06;
+use WebPAC::Validate 0.11;
 use WebPAC::Output::MARC;
 use WebPAC::Config;
 use Getopt::Long;
@@ -21,6 +21,7 @@
 use File::Slurp;
 use Data::Dump qw/dump/;
 use Storable qw/dclone/;
+use Pod::Usage qw/pod2usage/;
 
 use Proc::Queue size => 1;
 use POSIX ":sys_wait_h"; # imports WNOHANG
@@ -31,7 +32,7 @@
 
 B<this command will probably go away. Don't get used to it!>
 
-Options:
+=head1 OPTIONS
 
 =over 4
 
@@ -67,17 +68,19 @@
 
 turn on extra validation of imput records, see L<WebPAC::Validation>
 
-=item --marc-normalize conf/normalize/mapping.pl
+=item --validate-delimiters path/to/validate_delimiters_file
 
-This option specifies normalisation file for MARC creation
+this option is used with C<--validate> to turn on extra validation of
+delimiters. If file is non existant, it will be created on first run.
 
-=item --marc-output out/marc/test.marc
+=item --marc-generate
 
-Optional path to output file
+Generate MARC file. This will automatically be on if file contains C<marc*> directives.
+You can use this option as C<--no-marc-generate> to disable MARC generation.
 
 =item --marc-lint
 
-By default turned on if C<--marc-normalize> is used. You can disable lint
+By default turned on if normalisation file has C<marc*> directives. You can disable lint
 messages with C<--no-marc-lint>.
 
 =item --marc-dump
@@ -110,12 +113,14 @@
 my $only_filter;
 my $stats = 0;
 my $validate_path;
-my ($marc_normalize, $marc_output);
+my $validate_delimiters_path;
+my $marc_generate = 1;
 my $marc_lint = 1;
 my $marc_dump = 0;
 my $parallel = 0;
 my $only_links = 0;
 my $merge = 0;
+my $help;
 
 my $log = _new WebPAC::Common()->_get_logger();
 
@@ -129,15 +134,18 @@
 	"debug+" => \$debug,
 	"stats" => \$stats,
 	"validate=s" => \$validate_path,
-	"marc-normalize=s" => \$marc_normalize,
-	"marc-output=s" => \$marc_output,
+	"validate-delimiters=s" => \$validate_delimiters_path,
+	"marc-generate!" => \$marc_generate,
 	"marc-lint!" => \$marc_lint,
 	"marc-dump!" => \$marc_dump,
 	"parallel=i" => \$parallel,
 	"only-links!" => \$only_links,
 	"merge" => \$merge,
+	"help" => \$help,
 );
 
+pod2usage(-verbose => 2) if ($help);
+
 my $config = new WebPAC::Config( path => $config_path );
 
 #print "config = ",dump($config) if ($debug);
@@ -146,6 +154,12 @@
 
 $log->info( "-" x 79 );
 
+my $log_file = 'log';
+
+if (-e $log_file ) {	# && -s $log_file > 5 * 1024 * 1024) {
+	$log->info("moved old log with ", -s $log_file, " bytes to '${log_file}.old'");
+	rename $log_file, "${log_file}.old" || $log->logwarn("can't rename $log_file to ${log_file}.old: $!");
+}
 
 my $estcmd_fh;
 my $estcmd_path = './estcmd-merge.sh';
@@ -156,24 +170,22 @@
 	$log->info("created merge batch file $estcmd_path");
 }
 
-
 my $validate;
 $validate = new WebPAC::Validate(
 	path => $validate_path,
-) if ($validate_path);
-
+	delimiters => $config->webpac('delimiters'),
+	delimiters_path => $validate_delimiters_path,
+) if ($validate_path || $validate_delimiters_path);
 
 my $use_indexer = $config->use_indexer;
+$stats ||= $validate;
 if ($stats) {
-	$log->debug("option --stats disables update of indexing engine...");
+	$log->debug("disabled indexing for stats collection");
 	$use_indexer = undef;
 } else {
 	$log->info("using $use_indexer indexing engine...");
 }
 
-# disable indexing when creating marc
-$use_indexer = undef if ($marc_normalize);
-
 # parse normalize files and create source files for lookup and normalization
 
 my $parser = new WebPAC::Parser( config => $config );
@@ -212,7 +224,7 @@
 	}
 
 	my $indexer;
-	if ($use_indexer) {
+	if ($use_indexer && $parser->have_rules( 'search', $database )) {
 
 		my $cfg_name = $use_indexer;
 		$cfg_name =~ s/\-.*$//;
@@ -285,7 +297,7 @@
 	my $abs_path = abs_path($0);
 	$abs_path =~ s#/[^/]*$#/#;
 
-	my $db_path = $config->get('webpac')->{db_path} . '/' . $database;
+	my $db_path = $config->webpac('db_path');
 
 	if ($clean) {
 		$log->info("creating new database '$database' in $db_path");
@@ -296,7 +308,6 @@
 
 	my $store = new WebPAC::Store(
 		path => $db_path,
-		database => $database,
 		debug => $debug,
 	);
 
@@ -355,7 +366,6 @@
 			delete( $input->{lookup} );
 		}
 
-		my $lookup;
 		my $lookup_coderef;
 
 		if (@lookups) {
@@ -369,11 +379,10 @@
 				WebPAC::Normalize::data_structure(
 					row => $rec,
 					rules => $rules,
-					lookup => $lookup,
 					config => create_ds_config( $db_config, $database, $input, $mfn ),
 				);
 
-				warn "current lookup = ", dump($lookup) if ($lookup);
+				#warn "current lookup: ", dump(WebPAC::Normalize::_get_lookup());
 			};
 
 			WebPAC::Normalize::_set_lookup( undef );
@@ -382,20 +391,46 @@
 
 		};
 
+		my $lookup_jar;
+
 		my $maxmfn = $input_db->open(
 			path => $input->{path},
 			code_page => $input->{encoding},	# database encoding
 			lookup_coderef => $lookup_coderef,
+			lookup => $lookup_jar,
 			%{ $input },
+			load_row => sub {
+				my $a = shift;
+				return $store->load_row(
+					database => $database,
+					input => $input_name,
+					id => $a->{id},
+				);
+			},
+			save_row => sub {
+				my $a = shift;
+				return $store->save_row(
+					database => $database,
+					input => $input_name,
+					id => $a->{id},
+					row => $a->{row},
+				);
+			},
+
 		);
 
 		my $lookup_data = WebPAC::Normalize::_get_lookup();
 
 		if (defined( $lookup_data->{$database}->{$input_name} )) {
-			$log->debug("created following lookups: ", dump( $lookup_data ));
+			$log->debug("created following lookups: ", sub { dump( $lookup_data ) } );
 
 			foreach my $key (keys %{ $lookup_data->{$database}->{$input_name} }) {
-				$store->save_lookup( $database, $input_name, $key, $lookup_data->{$database}->{$input_name}->{$key} );
+				$store->save_lookup(
+					database => $database,
+					input => $input_name,
+					key => $key,
+					data => $lookup_data->{$database}->{$input_name}->{$key},
+				);
 			}
 		}
 
@@ -409,120 +444,147 @@
 			$log->info("Generating report file $path");
 		}
 
-		my @norm_array = ref($input->{normalize}) eq 'ARRAY' ?
-			@{ $input->{normalize} } : ( $input->{normalize} );
-
-		if ($marc_normalize) {
-			@norm_array = ( {
-				path => $marc_normalize,
-				output => $marc_output || "out/marc/${database}-${input_name}.marc",
-			} );
-		}
-
-		foreach my $normalize (@norm_array) {
-
-			my $normalize_path = $normalize->{path} || $log->logdie("can't find normalize path in config");
-
-			$log->logdie("Found '$normalize_path' as normalization file which isn't supported any more!") unless ( $normalize_path =~ m!\.pl$!i );
-
-			my $rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!";
-
-			$log->info("Using $normalize_path for normalization...");
-
-			my $marc = new WebPAC::Output::MARC(
-				path => $normalize->{output},
+		my $marc;
+		if ($marc_generate && $parser->have_rules( 'marc', $database, $input_name )) {
+			$marc = new WebPAC::Output::MARC(
+				path => "out/marc/${database}-${input_name}.marc",
 				lint => $marc_lint,
 				dump => $marc_dump,
-			) if ($normalize->{output});
+			);
+		}
+
+		my $rules = $parser->normalize_rules($database,$input_name) || $log->logdie("no normalize rules found for $database/$input_name");
+		$log->debug("parsed normalize rules:\n$rules");
 
-			# reset position in database
-			$input_db->seek(1);
+		# reset position in database
+		$input_db->seek(1);
 
-			# generate name of config key for indexer (strip everything after -)
-			my $indexer_config = $use_indexer;
-			$indexer_config =~ s/^(\w+)-?.*$/$1/g if ($indexer_config);
+		# generate name of config key for indexer (strip everything after -)
+		my $indexer_config = $use_indexer;
+		$indexer_config =~ s/^(\w+)-?.*$/$1/g if ($indexer_config);
+
+		my $lookup_hash;
+		my $depends = $parser->depends($database,$input_name);
+	
+		if ($depends) {
+			$log->debug("$database/$input_name depends on: ", dump($depends)) if ($depends);
+			$log->logdie("parser->depends didn't return HASH") unless (ref($depends) eq 'HASH');
+
+			foreach my $db (keys %$depends) {
+				foreach my $i (keys %{$depends->{$db}}) {
+					foreach my $k (keys %{$depends->{$db}->{$i}}) {
+						my $t = time();
+						$log->debug("loading lookup $db/$i");
+						$lookup_hash->{$db}->{$i}->{$k} = $store->load_lookup(
+							database => $db,
+							input => $i,
+							key => $k,
+						);
+						$log->debug(sprintf("lookup $db/$i took %.2fs", time() - $t));
+					}
+				}
+			}
 
-			foreach my $pos ( 0 ... $input_db->size ) {
+			$log->debug("lookup_hash = ", sub { dump( $lookup_hash ) });
+		}
 
-				my $row = $input_db->fetch || next;
 
-				my $mfn = $row->{'000'}->[0];
+		foreach my $pos ( 0 ... $input_db->size ) {
 
-				if (! $mfn || $mfn !~ m#^\d+$#) {
-					$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
-					$mfn = $pos;
-					push @{ $row->{'000'} }, $pos;
-				}
+			my $row = $input_db->fetch || next;
 
+			$total_rows++;
 
-				if ($validate) {
-					if ( my $errors = $validate->validate_errors( $row, $input_db->dump ) ) {
-						$log->error( "MFN $mfn validation error:\n",
-							$validate->report_error( $errors )
-						);
-					}
-				}
+			my $mfn = $row->{'000'}->[0];
 
-				my $ds = WebPAC::Normalize::data_structure(
-					row => $row,
-					rules => $rules,
-					lookup => $lookup ? $lookup->lookup_hash : undef,
-					config => create_ds_config( $db_config, $database, $input, $mfn ),
-					marc_encoding => 'utf-8',
-				);
+			if (! $mfn || $mfn !~ m#^\d+$#) {
+				$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
+				$mfn = $pos;
+				push @{ $row->{'000'} }, $pos;
+			}
 
-				$store->save_ds(
-					id => $mfn,
-					ds => $ds,
-					prefix => $input_name,
-				) if ($ds && !$stats);
-
-				$indexer->add(
-					id => "${input_name}/${mfn}",
-					ds => $ds,
-					type => $config->get($indexer_config)->{type},
-				) if ($indexer && $ds);
-
-				if ($marc) {
-					my $i = 0;
-
-					while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) {
-						$marc->add(
-							id => $mfn . ( $i ? "/$i" : '' ),
-							fields => $fields,
-							leader => WebPAC::Normalize::marc_leader(),
-							row => $row,
-						);
-						$i++;
-					}
 
-					$log->info("Created $i instances of MFN $mfn\n") if ($i > 1);
+			if ($validate) {
+				if ( my $errors = $validate->validate_rec( $row, $input_db->dump_ascii ) ) {
+					$log->error( "MFN $mfn validation error:\n",
+						$validate->report_error( $errors )
+					);
 				}
-
-				$total_rows++;
+				next;	# validation doesn't create any output
 			}
 
-			if ($validate) {
-				my $errors = $validate->report;
-				if ($errors) {
-					$log->info("validation errors:\n$errors\n" );
-					print $report_fh "$errors\n" if ($report_fh);
+			my $ds = WebPAC::Normalize::data_structure(
+				row => $row,
+				rules => $rules,
+				lookup => $lookup_hash,
+				config => create_ds_config( $db_config, $database, $input, $mfn ),
+				marc_encoding => 'utf-8',
+				load_row_coderef => sub {
+					my ($database,$input,$mfn) = @_;
+					return $store->load_row(
+						database => $database,
+						input => $input,
+						id => $mfn,
+					);
+				},
+			);
+
+			$log->debug("ds = ", sub { dump($ds) }) if ($ds);
+
+			$store->save_ds(
+				database => $database,
+				input => $input_name,
+				id => $mfn,
+				ds => $ds,
+			) if ($ds && !$stats);
+
+			$indexer->add(
+				id => "${input_name}/${mfn}",
+				ds => $ds,
+				type => $config->get($indexer_config)->{type},
+			) if ($indexer && $ds);
+
+			if ($marc) {
+				my $i = 0;
+
+				while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) {
+					$marc->add(
+						id => $mfn . ( $i ? "/$i" : '' ),
+						fields => $fields,
+						leader => WebPAC::Normalize::_get_marc_leader(),
+						row => $row,
+					);
+					$i++;
 				}
+
+				$log->info("Created $i instances of MFN $mfn\n") if ($i > 1);
 			}
+		}
 
-			if ($stats) {
-				my $s = $input_db->stats;
-				$log->info("statistics of fields usage:\n$s");
-				print $report_fh "Statistics of fields usage:\n$s" if ($report_fh);
+		if ($validate) {
+			my $errors = $validate->report;
+			if ($errors) {
+				$log->info("validation errors:\n$errors\n" );
+				print $report_fh "$errors\n" if ($report_fh);
 			}
 
-			# close MARC file
-			$marc->finish if ($marc);
+			print $report_fh "\nAll possible subfields/delimiter templates:\n", $validate->delimiters_templates( report => 1 ), "\n\n";
+		}
 
-			# close report
-			close($report_fh) if ($report_fh)
+		if ($stats) {
+			my $s = $input_db->stats;
+			$log->info("statistics of fields usage:\n$s");
+			print $report_fh "Statistics of fields usage:\n$s" if ($report_fh);
 		}
 
+		# close MARC file
+		$marc->finish if ($marc);
+
+		# close report
+		close($report_fh) if ($report_fh);
+
+		# reset validate stats
+		$validate->reset if ( $validate );
 	}
 
 	eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
@@ -549,6 +611,9 @@
 	$log->info("all parallel processes finished");
 }
 
+# save new delimiters if needed
+$validate->save_delimiters_templates if ( $validate );
+
 #
 # handle links or merge after indexing
 #