webpac2/trunk/run.pl

#!/usr/bin/perl -w

use strict;

use Cwd qw/abs_path/;
use File::Temp qw/tempdir/;
use lib './lib';

use WebPAC::Common 0.02;
use WebPAC::Lookup;
use WebPAC::Input 0.03;
use WebPAC::Store 0.03;
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
use MARC::Lint;
use Data::Dump qw/dump/;

=head1 NAME

run.pl - start WebPAC indexing

B<this command will probably go away. Don't get used to it!>

Options:

=over 4

=item --offset 42

start loading (all) databases at offset 42

=item --limit 100

limit loading to 100 records

=item --clean

remove database and Hyper Estraier index before indexing

=item --only=database_name/input_filter

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 $offset;
my $limit;

my $clean = 0;
my $config = 'conf/config.yml';
my $debug = 0;
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" => \$only_filter,
        "only=s" => \$only_filter,
        "config" => \$config,
        "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 = ",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} }) {

        my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter);
        next if ($only_database && $database !~ m/$only_database/i);

        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};

                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");
                }

                $log->logide("can't continue without valid indexer") unless ($indexer);
        }


        #
        # now WebPAC::Store
        #
        my $abs_path = abs_path($0);
        $abs_path =~ s#/[^/]*$#/#;

        my $db_path = $config->{webpac}->{db_path} . '/' . $database;

        if ($clean) {
                $log->info("creating new database '$database' in $db_path");
                rmtree( $db_path ) || $log->warn("can't remove $db_path: $!");
        } else {
                $log->info("working on database '$database' in $db_path");
        }

        my $db = new WebPAC::Store(
                path => $db_path,
                database => $database,
                debug => $debug,
        );


        #
        # now, iterate through input formats
        #

        my @inputs;
        if (ref($db_config->{input}) eq 'ARRAY') {
                @inputs = @{ $db_config->{input} };
        } elsif ($db_config->{input}) {
                push @inputs, $db_config->{input};
        } else {
                $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));

                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",
                        $input->{lookup} ? "lookup '$input->{lookup}'" : ""
                );

                my $input_db = new WebPAC::Input(
                        module => $input_module,
                        code_page => $config->{webpac}->{webpac_encoding},
                        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 @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_fh;
                        if (my $path = $normalize->{output}) {
                                open($marc_fh, '>', $path) ||
                                        $log->logdie("can't open MARC output $path: $!");

                                $log->info("Creating MARC export file $path", $marc_lint ? ' (with lint)' : '', "\n");
                        }

                        # 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\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++;
                        }

                        $log->info("statistics of fields usage:\n", $input_db->stats) if ($stats);

                        # close MARC file
                        close($marc_fh) if ($marc_fh);

                }

        }

        eval { $indexer->finish } if ($indexer && $indexer->can('finish'));

        my $dt = time() - $start_t;
        $log->info("$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' && $use_indexer) {
                foreach my $link (@{ $db_config->{links} }) {
                        if ($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 } );
}
1	dpavlin	74	#!/usr/bin/perl -w
2
3			use strict;
4
5			use Cwd qw/abs_path/;
6			use File::Temp qw/tempdir/;
7			use lib './lib';
8
9	dpavlin	255	use WebPAC::Common 0.02;
10	dpavlin	74	use WebPAC::Lookup;
11	dpavlin	287	use WebPAC::Input 0.03;
12	dpavlin	209	use WebPAC::Store 0.03;
13	dpavlin	536	use WebPAC::Normalize;
14	dpavlin	74	use WebPAC::Output::TT;
15	dpavlin	516	use WebPAC::Validate;
16	dpavlin	141	use YAML qw/LoadFile/;
17	dpavlin	301	use Getopt::Long;
18			use File::Path;
19	dpavlin	389	use Time::HiRes qw/time/;
20	dpavlin	492	use File::Slurp;
21	dpavlin	541	use MARC::Record 2.0; # need 2.0 for utf-8 encoding see marcpm.sf.net
22	dpavlin	556	use MARC::Lint;
23			use Data::Dump qw/dump/;
24	dpavlin	74
25	dpavlin	301	=head1 NAME
26	dpavlin	76
27	dpavlin	301	run.pl - start WebPAC indexing
28	dpavlin	141
29	dpavlin	301	B<this command will probably go away. Don't get used to it!>
30	dpavlin	141
31	dpavlin	301	Options:
32
33			=over 4
34
35			=item --offset 42
36
37			start loading (all) databases at offset 42
38
39			=item --limit 100
40
41			limit loading to 100 records
42
43			=item --clean
44
45			remove database and Hyper Estraier index before indexing
46
47	dpavlin	510	=item --only=database_name/input_filter
48	dpavlin	335
49	dpavlin	423	reindex just single database (legacy name is --one)
50	dpavlin	335
51	dpavlin	510	C</input_filter> is optional part which can be C<name>
52			or C<type> from input
53
54	dpavlin	301	=item --config conf/config.yml
55
56			path to YAML configuration file
57
58	dpavlin	507	=item --stats
59
60	dpavlin	509	disable indexing and dump statistics about field and subfield
61			usage for each input
62	dpavlin	507
63	dpavlin	516	=item --validate path/to/validation_file
64
65			turn on extra validation of imput records, see L<WebPAC::Validation>
66
67	dpavlin	552	=item --marc-normalize conf/normalize/mapping.pl
68
69			This option specifies normalisation file for MARC creation
70
71			=item --marc-output out/marc/test.marc
72
73			Optional path to output file
74
75	dpavlin	556	=item --marc-lint
76
77			By default turned on if C<--marc-normalize> is used. You can disable lint
78	dpavlin	558	messages with C<--no-marc-lint>.
79	dpavlin	556
80	dpavlin	559	=item --marc-dump
81
82			Force dump or input and marc record for debugging.
83
84	dpavlin	301	=back
85
86			=cut
87
88			my $offset;
89			my $limit;
90
91			my $clean = 0;
92			my $config = 'conf/config.yml';
93			my $debug = 0;
94	dpavlin	510	my $only_filter;
95	dpavlin	507	my $stats = 0;
96	dpavlin	516	my $validate_path;
97	dpavlin	552	my ($marc_normalize, $marc_output);
98	dpavlin	556	my $marc_lint = 1;
99	dpavlin	559	my $marc_dump = 0;
100	dpavlin	301
101			GetOptions(
102			"limit=i" => \$limit,
103			"offset=i" => \$offset,
104			"clean" => \$clean,
105	dpavlin	510	"one=s" => \$only_filter,
106			"only=s" => \$only_filter,
107	dpavlin	301	"config" => \$config,
108	dpavlin	560	"debug+" => \$debug,
109	dpavlin	507	"stats" => \$stats,
110	dpavlin	516	"validate=s" => \$validate_path,
111	dpavlin	552	"marc-normalize=s" => \$marc_normalize,
112			"marc-output=s" => \$marc_output,
113	dpavlin	556	"marc-lint!" => \$marc_lint,
114	dpavlin	559	"marc-dump!" => \$marc_dump,
115	dpavlin	301	);
116
117			$config = LoadFile($config);
118
119	dpavlin	556	print "config = ",dump($config) if ($debug);
120	dpavlin	301
121	dpavlin	210	die "no databases in config file!\n" unless ($config->{databases});
122	dpavlin	431
123			my $log = _new WebPAC::Common()->_get_logger();
124	dpavlin	509	$log->info( "-" x 79 );
125	dpavlin	431
126	dpavlin	516	my $validate;
127			$validate = new WebPAC::Validate(
128			path => $validate_path,
129			) if ($validate_path);
130
131	dpavlin	430	my $use_indexer = $config->{use_indexer} \|\| 'hyperestraier';
132	dpavlin	509	if ($stats) {
133			$log->debug("option --stats disables update of indexing engine...");
134			$use_indexer = undef;
135			} else {
136			$log->info("using $use_indexer indexing engine...");
137			}
138	dpavlin	141
139	dpavlin	552	# disable indexing when creating marc
140			$use_indexer = undef if ($marc_normalize);
141
142	dpavlin	213	my $total_rows = 0;
143	dpavlin	389	my $start_t = time();
144	dpavlin	213
145	dpavlin	529	my @links;
146			my $indexer;
147
148	dpavlin	556	my $lint = new MARC::Lint if ($marc_lint);
149
150	dpavlin	210	while (my ($database, $db_config) = each %{ $config->{databases} }) {
151	dpavlin	141
152	dpavlin	512	my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter);
153	dpavlin	510	next if ($only_database && $database !~ m/$only_database/i);
154	dpavlin	335
155	dpavlin	509	if ($use_indexer) {
156			my $indexer_config = $config->{$use_indexer} \|\| $log->logdie("can't find '$use_indexer' part in confguration");
157			$indexer_config->{database} = $database;
158			$indexer_config->{clean} = $clean;
159			$indexer_config->{label} = $db_config->{name};
160	dpavlin	431
161	dpavlin	509	if ($use_indexer eq 'hyperestraier') {
162	dpavlin	255
163	dpavlin	509	# open Hyper Estraier database
164			use WebPAC::Output::Estraier '0.10';
165			$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
166
167			} elsif ($use_indexer eq 'kinosearch') {
168	dpavlin	430
169	dpavlin	509	# open KinoSearch
170			use WebPAC::Output::KinoSearch;
171			$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path});
172			$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } );
173	dpavlin	431
174	dpavlin	509	} else {
175			$log->logdie("unknown use_indexer: $use_indexer");
176			}
177
178			$log->logide("can't continue without valid indexer") unless ($indexer);
179	dpavlin	430	}
180
181
182	dpavlin	255	#
183			# now WebPAC::Store
184			#
185	dpavlin	210	my $abs_path = abs_path($0);
186			$abs_path =~ s#/[^/]*$#/#;
187	dpavlin	141
188	dpavlin	210	my $db_path = $config->{webpac}->{db_path} . '/' . $database;
189	dpavlin	74
190	dpavlin	301	if ($clean) {
191	dpavlin	511	$log->info("creating new database '$database' in $db_path");
192	dpavlin	301	rmtree( $db_path ) \|\| $log->warn("can't remove $db_path: $!");
193			} else {
194	dpavlin	511	$log->info("working on database '$database' in $db_path");
195	dpavlin	301	}
196	dpavlin	255
197	dpavlin	210	my $db = new WebPAC::Store(
198			path => $db_path,
199	dpavlin	217	database => $database,
200	dpavlin	301	debug => $debug,
201	dpavlin	210	);
202	dpavlin	74
203	dpavlin	233
204	dpavlin	213	#
205			# now, iterate through input formats
206			#
207	dpavlin	74
208	dpavlin	213	my @inputs;
209			if (ref($db_config->{input}) eq 'ARRAY') {
210			@inputs = @{ $db_config->{input} };
211	dpavlin	255	} elsif ($db_config->{input}) {
212			push @inputs, $db_config->{input};
213	dpavlin	213	} else {
214	dpavlin	255	$log->info("database $database doesn't have inputs defined");
215	dpavlin	213	}
216	dpavlin	74
217	dpavlin	286	my @supported_inputs = keys %{ $config->{webpac}->{inputs} };
218
219	dpavlin	213	foreach my $input (@inputs) {
220	dpavlin	233
221	dpavlin	512	next if ($only_input && ($input->{name} !~ m#$only_input#i && $input->{type} !~ m#$only_input#i));
222	dpavlin	510
223	dpavlin	233	my $type = lc($input->{type});
224
225	dpavlin	286	die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs));
226	dpavlin	233
227			my $lookup = new WebPAC::Lookup(
228			lookup_file => $input->{lookup},
229	dpavlin	520	) if ($input->{lookup});
230	dpavlin	233
231	dpavlin	286	my $input_module = $config->{webpac}->{inputs}->{$type};
232	dpavlin	74
233	dpavlin	523	$log->info("working on input '$input->{name}' in $input->{path} [type: $input->{type}] using $input_module",
234			$input->{lookup} ? "lookup '$input->{lookup}'" : ""
235			);
236	dpavlin	286
237	dpavlin	287	my $input_db = new WebPAC::Input(
238			module => $input_module,
239	dpavlin	213	code_page => $config->{webpac}->{webpac_encoding},
240	dpavlin	301	limit => $limit \|\| $input->{limit},
241			offset => $offset,
242	dpavlin	251	lookup => $lookup,
243	dpavlin	416	recode => $input->{recode},
244	dpavlin	507	stats => $stats,
245	dpavlin	287	);
246	dpavlin	286	$log->logdie("can't create input using $input_module") unless ($input);
247	dpavlin	113
248	dpavlin	287	my $maxmfn = $input_db->open(
249	dpavlin	285	path => $input->{path},
250	dpavlin	213	code_page => $input->{encoding}, # database encoding
251	dpavlin	523	%{ $input },
252	dpavlin	213	);
253	dpavlin	113
254	dpavlin	539	my @norm_array = ref($input->{normalize}) eq 'ARRAY' ?
255			@{ $input->{normalize} } : ( $input->{normalize} );
256	dpavlin	210
257	dpavlin	552	if ($marc_normalize) {
258			@norm_array = ( {
259			path => $marc_normalize,
260			output => $marc_output \|\| 'out/marc/' . $database . '-' . $input->{name} . '.marc',
261			} );
262			}
263
264	dpavlin	539	foreach my $normalize (@norm_array) {
265	dpavlin	492
266	dpavlin	539	my $normalize_path = $normalize->{path} \|\| $log->logdie("can't find normalize path in config");
267	dpavlin	269
268	dpavlin	539	$log->logdie("Found '$normalize_path' as normalization file which isn't supported any more!") unless ( $normalize_path =~ m!\.pl$!i );
269	dpavlin	210
270	dpavlin	539	my $rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!";
271	dpavlin	213
272	dpavlin	539	$log->info("Using $normalize_path for normalization...");
273	dpavlin	213
274	dpavlin	540	my $marc_fh;
275	dpavlin	552	if (my $path = $normalize->{output}) {
276	dpavlin	540	open($marc_fh, '>', $path) \|\|
277			$log->logdie("can't open MARC output $path: $!");
278
279	dpavlin	556	$log->info("Creating MARC export file $path", $marc_lint ? ' (with lint)' : '', "\n");
280	dpavlin	540	}
281
282	dpavlin	539	# reset position in database
283			$input_db->seek(1);
284	dpavlin	290
285	dpavlin	539	foreach my $pos ( 0 ... $input_db->size ) {
286	dpavlin	516
287	dpavlin	539	my $row = $input_db->fetch \|\| next;
288	dpavlin	516
289	dpavlin	539	my $mfn = $row->{'000'}->[0];
290	dpavlin	213
291	dpavlin	539	if (! $mfn \|\| $mfn !~ m#^\d+$#) {
292			$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
293			$mfn = $pos;
294			push @{ $row->{'000'} }, $pos;
295			}
296	dpavlin	508
297	dpavlin	213
298	dpavlin	539	if ($validate) {
299			my @errors = $validate->validate_errors( $row );
300			$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors);
301			}
302	dpavlin	213
303	dpavlin	539	my $ds = WebPAC::Normalize::data_structure(
304			row => $row,
305			rules => $rules,
306			lookup => $lookup ? $lookup->lookup_hash : undef,
307	dpavlin	541	marc_encoding => 'utf-8',
308	dpavlin	539	);
309	dpavlin	507
310	dpavlin	539	$db->save_ds(
311			id => $mfn,
312			ds => $ds,
313			prefix => $input->{name},
314			) if ($ds && !$stats);
315	dpavlin	210
316	dpavlin	539	$indexer->add(
317			id => $input->{name} . "/" . $mfn,
318			ds => $ds,
319			type => $config->{$use_indexer}->{type},
320	dpavlin	540	) if ($indexer && $ds);
321	dpavlin	539
322	dpavlin	540	if ($marc_fh) {
323			my $marc = new MARC::Record;
324	dpavlin	541	$marc->encoding( 'utf-8' );
325	dpavlin	556	my @marc_fields = WebPAC::Normalize::_get_marc_fields();
326			if (! @marc_fields) {
327			$log->warn("MARC record $mfn is empty, skipping");
328			} else {
329			$marc->add_fields( @marc_fields );
330	dpavlin	559
331	dpavlin	564	# tweak leader
332			if (my $new_l = WebPAC::Normalize::marc_leader()) {
333
334			my $leader = $marc->leader;
335
336			foreach my $o ( keys %$new_l ) {
337			my $insert = $new_l->{$o};
338			$leader = substr($leader, 0, $o) .
339			$insert . substr($leader, $o+length($insert));
340			}
341			$marc->leader( $leader );
342			}
343
344	dpavlin	556	if ($marc_lint) {
345			$lint->check_record( $marc );
346			my $err = join( "\n", $lint->warnings );
347	dpavlin	567	$log->error("MARC lint detected warning on MFN $mfn\tleader:", $marc->leader(), "\n",
348			">>>>> Original imput row:\n",dump($row), "\n",
349			">>>>> Normalized MARC row:\n",dump(@marc_fields), "\n",
350			">>>>> MARC lint warnings:\n",$err,"\n"
351	dpavlin	556	) if ($err);
352			}
353	dpavlin	559
354			if ($marc_dump) {
355	dpavlin	564	$log->info("MARC record on MFN $mfn\tleader:", $marc->leader(), "\n",
356	dpavlin	567	">>>>> Original imput row:\n",dump($row), "\n",
357			">>>>> Normalized MARC row:\n",dump(@marc_fields), "\n",
358	dpavlin	559	);
359			}
360
361	dpavlin	556	print $marc_fh $marc->as_usmarc;
362			}
363	dpavlin	540	}
364
365	dpavlin	539	$total_rows++;
366			}
367
368			$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats);
369
370	dpavlin	540	# close MARC file
371			close($marc_fh) if ($marc_fh);
372	dpavlin	539
373	dpavlin	540	}
374
375	dpavlin	539	}
376
377	dpavlin	509	eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
378	dpavlin	434
379	dpavlin	389	my $dt = time() - $start_t;
380	dpavlin	511	$log->info("$total_rows records ", $indexer ? "indexed " : "",
381			sprintf("in %.2f sec [%.2f rec/sec]",
382	dpavlin	389	$dt, ($total_rows / $dt)
383			)
384			);
385	dpavlin	255
386	dpavlin	431	#
387			# add Hyper Estraier links to other databases
388			#
389	dpavlin	552	if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) {
390	dpavlin	431	foreach my $link (@{ $db_config->{links} }) {
391			if ($use_indexer eq 'hyperestraier') {
392	dpavlin	529	$log->info("saving link $database -> $link->{to} [$link->{credit}]");
393			push @links, {
394	dpavlin	430	from => $database,
395			to => $link->{to},
396			credit => $link->{credit},
397	dpavlin	529	};
398	dpavlin	431	} else {
399			$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
400	dpavlin	430	}
401	dpavlin	255	}
402			}
403
404	dpavlin	210	}
405
406	dpavlin	529	foreach my $link (@links) {
407			$log->info("adding link $link->{from} -> $link->{to} [$link->{credit}]");
408			$indexer->add_link( %{ $link } );
409			}