--- trunk/run.pl 2006/05/16 15:23:12 516 +++ trunk/run.pl 2006/07/02 21:22:26 567 @@ -4,15 +4,13 @@ use Cwd qw/abs_path/; use File::Temp qw/tempdir/; -use Data::Dumper; use lib './lib'; use WebPAC::Common 0.02; 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/; @@ -20,6 +18,9 @@ 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 +use MARC::Lint; +use Data::Dump qw/dump/; =head1 NAME @@ -54,11 +55,6 @@ path to YAML configuration file -=item --force-set - -force conversion C<< normalize->path >> in C from -C<.xml> to C<.pl> - =item --stats disable indexing and dump statistics about field and subfield @@ -68,6 +64,23 @@ turn on extra validation of imput records, see L +=item --marc-normalize conf/normalize/mapping.pl + +This option specifies normalisation file for MARC creation + +=item --marc-output out/marc/test.marc + +Optional path to output file + +=item --marc-lint + +By default turned on if C<--marc-normalize> is used. You can disable lint +messages with C<--no-marc-lint>. + +=item --marc-dump + +Force dump or input and marc record for debugging. + =back =cut @@ -79,9 +92,11 @@ my $config = 'conf/config.yml'; my $debug = 0; my $only_filter; -my $force_set = 0; my $stats = 0; my $validate_path; +my ($marc_normalize, $marc_output); +my $marc_lint = 1; +my $marc_dump = 0; GetOptions( "limit=i" => \$limit, @@ -90,15 +105,18 @@ "one=s" => \$only_filter, "only=s" => \$only_filter, "config" => \$config, - "debug" => \$debug, - "force-set" => \$force_set, + "debug+" => \$debug, "stats" => \$stats, "validate=s" => \$validate_path, + "marc-normalize=s" => \$marc_normalize, + "marc-output=s" => \$marc_output, + "marc-lint!" => \$marc_lint, + "marc-dump!" => \$marc_dump, ); $config = LoadFile($config); -print "config = ",Dumper($config) if ($debug); +print "config = ",dump($config) if ($debug); die "no databases in config file!\n" unless ($config->{databases}); @@ -118,16 +136,22 @@ $log->info("using $use_indexer indexing engine..."); } +# disable indexing when creating marc +$use_indexer = undef if ($marc_normalize); + my $total_rows = 0; my $start_t = time(); +my @links; +my $indexer; + +my $lint = new MARC::Lint if ($marc_lint); + 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); - 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; @@ -202,11 +226,13 @@ 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, @@ -222,73 +248,63 @@ 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 = $input->{normalize}->{path}; - - if ($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 $normalize_path"); - } + if ($marc_normalize) { + @norm_array = ( { + path => $marc_normalize, + output => $marc_output || 'out/marc/' . $database . '-' . $input->{name} . '.marc', + } ); } - if ($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: $!"; - } + foreach my $normalize (@norm_array) { - foreach my $pos ( 0 ... $input_db->size ) { + my $normalize_path = $normalize->{path} || $log->logdie("can't find normalize path in config"); - my $row = $input_db->fetch || next; + $log->logdie("Found '$normalize_path' as normalization file which isn't supported any more!") unless ( $normalize_path =~ m!\.pl$!i ); - my $mfn = $row->{'000'}->[0]; + my $rules = read_file( $normalize_path ) or die "can't open $normalize_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("Using $normalize_path for normalization..."); + my $marc_fh; + if (my $path = $normalize->{output}) { + open($marc_fh, '>', $path) || + $log->logdie("can't open MARC output $path: $!"); - if ($validate) { - my @errors = $validate->validate_errors( $row ); - $log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors); + $log->info("Creating MARC export file $path", $marc_lint ? ' (with lint)' : '', "\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( @@ -296,20 +312,67 @@ 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' ); + my @marc_fields = WebPAC::Normalize::_get_marc_fields(); + if (! @marc_fields) { + $log->warn("MARC record $mfn is empty, skipping"); + } else { + $marc->add_fields( @marc_fields ); + + # tweak leader + if (my $new_l = WebPAC::Normalize::marc_leader()) { + + my $leader = $marc->leader; + + foreach my $o ( keys %$new_l ) { + my $insert = $new_l->{$o}; + $leader = substr($leader, 0, $o) . + $insert . substr($leader, $o+length($insert)); + } + $marc->leader( $leader ); + } + + if ($marc_lint) { + $lint->check_record( $marc ); + my $err = join( "\n", $lint->warnings ); + $log->error("MARC lint detected warning on MFN $mfn\tleader:", $marc->leader(), "\n", + ">>>>> Original imput row:\n",dump($row), "\n", + ">>>>> Normalized MARC row:\n",dump(@marc_fields), "\n", + ">>>>> MARC lint warnings:\n",$err,"\n" + ) if ($err); + } + + if ($marc_dump) { + $log->info("MARC record on MFN $mfn\tleader:", $marc->leader(), "\n", + ">>>>> Original imput row:\n",dump($row), "\n", + ">>>>> Normalized MARC row:\n",dump(@marc_fields), "\n", + ); + } + + 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 && $indexer->can('finish')); @@ -323,15 +386,15 @@ # # add Hyper Estraier links to other databases # - if (ref($db_config->{links}) eq 'ARRAY') { + if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) { 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}]"); } @@ -340,3 +403,7 @@ } +foreach my $link (@links) { + $log->info("adding link $link->{from} -> $link->{to} [$link->{credit}]"); + $indexer->add_link( %{ $link } ); +}