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;
use WebPAC::Output::TT;
-use WebPAC::Output::Estraier '0.10';
+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
+use MARC::Lint;
+use Data::Dump qw/dump/;
=head1 NAME
remove database and Hyper Estraier index before indexing
-=item --one=database_name
+=item --only=database_name/input_filter
-reindex just single database
+reindex just single database (legacy name is --one)
+
+C</input_filter> is optional part which can be C<name>
+or C<type> from input
=item --config conf/config.yml
path to YAML configuration file
+=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>
+
+=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
my $clean = 0;
my $config = 'conf/config.yml';
my $debug = 0;
-my $one_db_name;
+my $only_filter;
+my $stats = 0;
+my $validate_path;
+my ($marc_normalize, $marc_output);
+my $marc_lint = 1;
+my $marc_dump = 0;
GetOptions(
"limit=i" => \$limit,
"offset=i" => \$offset,
"clean" => \$clean,
- "one=s" => \$one_db_name,
+ "one=s" => \$only_filter,
+ "only=s" => \$only_filter,
"config" => \$config,
- "debug" => \$debug,
+ "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});
+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...");
+ $use_indexer = undef;
+} else {
+ $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 %{ $config->{databases} }) {
- next if ($one_db_name && $database !~ m/$one_db_name/i);
+ my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter);
+ next if ($only_database && $database !~ m/$only_database/i);
- my $log = _new WebPAC::Common()->_get_logger();
+ if ($use_indexer) {
+ my $indexer_config = $config->{$use_indexer} || $log->logdie("can't find '$use_indexer' part in confguration");
+ $indexer_config->{database} = $database;
+ $indexer_config->{clean} = $clean;
+ $indexer_config->{label} = $db_config->{name};
- #
- # open Hyper Estraier database
- #
+ if ($use_indexer eq 'hyperestraier') {
+
+ # open Hyper Estraier database
+ use WebPAC::Output::Estraier '0.10';
+ $indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
+
+ } elsif ($use_indexer eq 'kinosearch') {
+
+ # open KinoSearch
+ use WebPAC::Output::KinoSearch;
+ $indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path});
+ $indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } );
+
+ } else {
+ $log->logdie("unknown use_indexer: $use_indexer");
+ }
- my $est_config = $config->{hyperestraier} || $log->logdie("can't find 'hyperestraier' part in confguration");
- $est_config->{database} = $database;
- $est_config->{clean} = $clean;
- $est_config->{label} = $db_config->{name};
+ $log->logide("can't continue without valid indexer") unless ($indexer);
+ }
- my $est = new WebPAC::Output::Estraier( %{ $est_config } );
#
# now WebPAC::Store
my $db_path = $config->{webpac}->{db_path} . '/' . $database;
if ($clean) {
- $log->info("creating new database $database in $db_path");
+ $log->info("creating new database '$database' in $db_path");
rmtree( $db_path ) || $log->warn("can't remove $db_path: $!");
} else {
- $log->debug("working on $database in $db_path");
+ $log->info("working on database '$database' in $db_path");
}
my $db = new WebPAC::Store(
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));
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->{path}' [$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,
limit => $limit || $input->{limit},
offset => $offset,
lookup => $lookup,
+ recode => $input->{recode},
+ stats => $stats,
);
$log->logdie("can't create input using $input_module") unless ($input);
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 $normalize_path = $input->{normalize}->{path};
-
- if ($normalize_path =~ m/\.xml$/i) {
- $n->open(
- tag => $input->{normalize}->{tag},
- xml_file => $input->{normalize}->{path},
- );
- } elsif ($normalize_path =~ m/\.(?:yml|yaml)$/i) {
- $n->open_yaml(
- path => $normalize_path,
- tag => $input->{normalize}->{tag},
- );
+ if ($marc_normalize) {
+ @norm_array = ( {
+ path => $marc_normalize,
+ output => $marc_output || 'out/marc/' . $database . '-' . $input->{name} . '.marc',
+ } );
}
- foreach my $pos ( 0 ... $input_db->size ) {
+ 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 $row = $input_db->fetch || next;
+ my $marc_fh;
+ if (my $path = $normalize->{output}) {
+ open($marc_fh, '>', $path) ||
+ $log->logdie("can't open MARC output $path: $!");
- my $mfn = $row->{'000'}->[0];
+ $log->info("Creating MARC export file $path", $marc_lint ? ' (with lint)' : '', "\n");
+ }
- if (! $mfn || $mfn !~ m#^\d+$#) {
- $log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
- $mfn = $pos;
- push @{ $row->{'000'} }, $pos;
+ # 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->lookup_hash : undef,
+ marc_encoding => 'utf-8',
+ );
+
+ $db->save_ds(
+ id => $mfn,
+ 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 $mfn\n",
+ "Original imput row: ",dump($row), "\n",
+ "Normalized MARC row: ",dump(@marc_fields), "\n",
+ "MARC lint warnings: ",$err,"\n"
+ ) if ($err);
+ }
+
+ if ($marc_dump) {
+ $log->info("MARC record on MFN $mfn\tleader:", $marc->leader(), "\n",
+ "Original imput row: ",dump($row), "\n",
+ "Normalized MARC row: ",dump(@marc_fields), "\n",
+ );
+ }
+
+ print $marc_fh $marc->as_usmarc;
+ }
+ }
+
+ $total_rows++;
}
- my $ds = $n->data_structure($row);
+ $log->info("statistics of fields usage:\n", $input_db->stats) if ($stats);
- $est->add(
- id => $input->{name} . "/" . $mfn,
- ds => $ds,
- type => $config->{hyperestraier}->{type},
- );
+ # close MARC file
+ close($marc_fh) if ($marc_fh);
- $total_rows++;
}
- };
+ }
+
+ eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
my $dt = time() - $start_t;
- $log->info("$total_rows records indexed in " .
- sprintf("%.2f sec [%.2f rec/sec]",
+ $log->info("$total_rows records ", $indexer ? "indexed " : "",
+ sprintf("in %.2f sec [%.2f rec/sec]",
$dt, ($total_rows / $dt)
)
);
#
# add Hyper Estraier links to other databases
#
- if (ref($db_config->{links}) eq 'ARRAY') {
+ if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) {
foreach my $link (@{ $db_config->{links} }) {
- $log->info("adding link $database -> $link->{to} [$link->{credit}]");
- $est->add_link(
- from => $database,
- to => $link->{to},
- credit => $link->{credit},
- );
+ if ($use_indexer eq 'hyperestraier') {
+ $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}]");
+ }
}
}
}
+foreach my $link (@links) {
+ $log->info("adding link $link->{from} -> $link->{to} [$link->{credit}]");
+ $indexer->add_link( %{ $link } );
+}
projects / webpac2 / blobdiff