--- trunk/run.pl	2006/08/25 12:31:06	619
+++ trunk/run.pl	2006/09/25 15:26:12	707
@@ -7,14 +7,14 @@
 use lib './lib';
 
 use WebPAC::Common 0.02;
-use WebPAC::Lookup 0.03;
-use WebPAC::Input 0.11;
+use WebPAC::Parser 0.04;
+use WebPAC::Input 0.13;
 use WebPAC::Store 0.03;
 use WebPAC::Normalize 0.11;
 use WebPAC::Output::TT;
-use WebPAC::Validate;
+use WebPAC::Validate 0.06;
 use WebPAC::Output::MARC;
-use YAML qw/LoadFile/;
+use WebPAC::Config;
 use Getopt::Long;
 use File::Path;
 use Time::HiRes qw/time/;
@@ -60,8 +60,8 @@
 
 =item --stats
 
-disable indexing and dump statistics about field and subfield
-usage for each input
+disable indexing, modify_* in configuration and dump statistics about field
+and subfield usage for each input
 
 =item --validate path/to/validation_file
 
@@ -105,7 +105,7 @@
 my $limit;
 
 my $clean = 0;
-my $config = 'conf/config.yml';
+my $config_path;
 my $debug = 0;
 my $only_filter;
 my $stats = 0;
@@ -119,21 +119,13 @@
 
 my $log = _new WebPAC::Common()->_get_logger();
 
-my $hostname = `hostname`;
-chomp($hostname);
-$hostname =~ s/\..+$//;
-if (-e "conf/$hostname.yml") {
-	$config = "conf/$hostname.yml";
-	$log->info("using host configuration file: $config");
-}
-
 GetOptions(
 	"limit=i" => \$limit,
 	"offset=i" => \$offset,
 	"clean" => \$clean,
 	"one=s" => \$only_filter,
 	"only=s" => \$only_filter,
-	"config" => \$config,
+	"config" => \$config_path,
 	"debug+" => \$debug,
 	"stats" => \$stats,
 	"validate=s" => \$validate_path,
@@ -146,11 +138,11 @@
 	"merge" => \$merge,
 );
 
-$config = LoadFile($config);
+my $config = new WebPAC::Config( path => $config_path );
 
 #print "config = ",dump($config) if ($debug);
 
-die "no databases in config file!\n" unless ($config->{databases});
+die "no databases in config file!\n" unless ($config->databases);
 
 $log->info( "-" x 79 );
 
@@ -171,7 +163,7 @@
 ) if ($validate_path);
 
 
-my $use_indexer = $config->{use_indexer} || 'hyperestraier';
+my $use_indexer = $config->use_indexer;
 if ($stats) {
 	$log->debug("option --stats disables update of indexing engine...");
 	$use_indexer = undef;
@@ -182,6 +174,10 @@
 # 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 );
+
 my $total_rows = 0;
 my $start_t = time();
 
@@ -192,7 +188,16 @@
 	Proc::Queue::size($parallel);
 }
 
-while (my ($database, $db_config) = each %{ $config->{databases} }) {
+sub create_ds_config {
+	my ($db_config, $database, $input, $mfn) = @_;
+	my $c = dclone( $db_config );
+	$c->{_} = $database || $log->logconfess("need database");
+	$c->{_mfn} = $mfn || $log->logconfess("need mfn");
+	$c->{input} = $input || $log->logconfess("need input");
+	return $c;
+}
+
+while (my ($database, $db_config) = each %{ $config->databases }) {
 
 	my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter);
 	next if ($only_database && $database !~ m/$only_database/i);
@@ -208,7 +213,11 @@
 
 	my $indexer;
 	if ($use_indexer) {
-		my $indexer_config = $config->{$use_indexer} || $log->logdie("can't find '$use_indexer' part in confguration");
+
+		my $cfg_name = $use_indexer;
+		$cfg_name =~ s/\-.*$//;
+
+		my $indexer_config = $config->get( $cfg_name ) || $log->logdie("can't find '$cfg_name' part in confguration");
 		$indexer_config->{database} = $database;
 		$indexer_config->{clean} = $clean;
 		$indexer_config->{label} = $db_config->{name};
@@ -222,6 +231,12 @@
 			use WebPAC::Output::Estraier '0.10';
 			$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
 		
+		} elsif ($use_indexer eq 'hyperestraier-native') {
+
+			# open Hyper Estraier database
+			use WebPAC::Output::EstraierNative;
+			$indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } );
+
 		} elsif ($use_indexer eq 'kinosearch') {
 
 			# open KinoSearch
@@ -270,7 +285,7 @@
 	my $abs_path = abs_path($0);
 	$abs_path =~ s#/[^/]*$#/#;
 
-	my $db_path = $config->{webpac}->{db_path} . '/' . $database;
+	my $db_path = $config->get('webpac')->{db_path} . '/' . $database;
 
 	if ($clean) {
 		$log->info("creating new database '$database' in $db_path");
@@ -299,51 +314,91 @@
 		$log->info("database $database doesn't have inputs defined");
 	}
 
-	my @supported_inputs = keys %{ $config->{webpac}->{inputs} };
-
 	foreach my $input (@inputs) {
 
 		next if ($only_input && ($input->{name} !~ m#$only_input#i && $input->{type} !~ m#$only_input#i));
 
 		my $type = lc($input->{type});
 
-		die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs));
+		die "I know only how to handle input types ", join(",", $config->webpac('inputs') ), " not '$type'!\n" unless (grep(/$type/, $config->webpac('inputs')));
 
-		my $lookup;
-		if ($input->{lookup}) {
-			$lookup = new WebPAC::Lookup(
-				lookup_file => $input->{lookup},
-			);
-			delete( $input->{lookup} );
-		}
+		my $input_module = $config->webpac('inputs')->{$type};
 
-		my $input_module = $config->{webpac}->{inputs}->{$type};
+		my @lookups = $parser->have_lookup_create($database, $input);
 
 		$log->info("working on input '$input->{name}' in $input->{path} [type: $input->{type}] using $input_module",
-			$input->{lookup} ? "lookup '$input->{lookup}'" : ""
+			@lookups ? " creating lookups: ".join(", ", @lookups) : ""
 		);
 
+		if ($stats) {
+			# disable modification of records if --stats is in use
+			delete($input->{modify_records});
+			delete($input->{modify_file});
+		}
+
 		my $input_db = new WebPAC::Input(
 			module => $input_module,
-			encoding => $config->{webpac}->{webpac_encoding},
+			encoding => $config->webpac('webpac_encoding'),
 			limit => $limit || $input->{limit},
 			offset => $offset,
-			lookup_coderef => sub {
-				my $rec = shift || return;
-				$lookup->add( $rec );
-			},
 			recode => $input->{recode},
 			stats => $stats,
 			modify_records => $input->{modify_records},
+			modify_file => $input->{modify_file},
 		);
 		$log->logdie("can't create input using $input_module") unless ($input);
 
+		if (defined( $input->{lookup} )) {
+			$log->warn("$database/", $input->{name}, " has depriciated lookup definition, removing it...");
+			delete( $input->{lookup} );
+		}
+
+		my $lookup;
+		my $lookup_coderef;
+
+		if (@lookups) {
+
+			my $rules = $parser->lookup_create_rules($database, $input) || $log->logdie("no rules found for $database/$input");
+
+			$lookup_coderef = sub {
+				my $rec = shift || die "need rec!";
+				my $mfn = $rec->{'000'}->[0] || die "need mfn in 000";
+
+				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);
+			};
+
+			WebPAC::Normalize::_set_lookup( undef );
+
+			$log->debug("created lookup_coderef using:\n$rules");
+
+		};
+
 		my $maxmfn = $input_db->open(
 			path => $input->{path},
 			code_page => $input->{encoding},	# database encoding
+			lookup_coderef => $lookup_coderef,
 			%{ $input },
 		);
 
+		$log->debug("created following lookups: ", dump( WebPAC::Normalize::_get_lookup() ) );
+
+		my $report_fh;
+		if ($stats || $validate) {
+			my $path = "out/report/" . $database . '-' . $input->{name} . '.txt';
+			open($report_fh, '>', $path) || $log->logdie("can't open $path: $!");
+
+			print $report_fh "Report for database '$database' input '$input->{name}' records ",
+				$offset || 1, "-", $limit || $input->{limit} || $maxmfn, "\n\n";
+			$log->info("Generating report file $path");
+		}
+
 		my @norm_array = ref($input->{normalize}) eq 'ARRAY' ?
 			@{ $input->{normalize} } : ( $input->{normalize} );
 
@@ -373,6 +428,10 @@
 			# 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);
+
 			foreach my $pos ( 0 ... $input_db->size ) {
 
 				my $row = $input_db->fetch || next;
@@ -387,26 +446,18 @@
 
 
 				if ($validate) {
-					my @errors = $validate->validate_errors( $row );
-					$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors);
+					if ( my $errors = $validate->validate_errors( $row, $input_db->dump ) ) {
+						$log->error( "MFN $mfn validation error:\n",
+							$validate->report_error( $errors )
+						);
+					}
 				}
 
-				my $ds_config = dclone($db_config);
-
-				# default values -> database key
-				$ds_config->{_} = $database;
-
-				# current mfn
-				$ds_config->{_mfn} = $mfn;
-
-				# attach current input
-				$ds_config->{input} = $input;
-
 				my $ds = WebPAC::Normalize::data_structure(
 					row => $row,
 					rules => $rules,
 					lookup => $lookup ? $lookup->lookup_hash : undef,
-					config => $ds_config,
+					config => create_ds_config( $db_config, $database, $input, $mfn ),
 					marc_encoding => 'utf-8',
 				);
 
@@ -419,7 +470,7 @@
 				$indexer->add(
 					id => $input->{name} . "/" . $mfn,
 					ds => $ds,
-					type => $config->{$use_indexer}->{type},
+					type => $config->get($indexer_config)->{type},
 				) if ($indexer && $ds);
 
 				if ($marc) {
@@ -441,11 +492,25 @@
 				$total_rows++;
 			}
 
-			$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats);
+			if ($validate) {
+				my $errors = $validate->report;
+				if ($errors) {
+					$log->info("validation errors:\n$errors\n" );
+					print $report_fh "$errors\n" 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)
 		}
 
 	}