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 0.03;
use WebPAC::Input 0.07;
use WebPAC::Store 0.03;
use WebPAC::Normalize 0.11;
use WebPAC::Output::TT;
use WebPAC::Validate;
use WebPAC::Output::MARC;
use YAML qw/LoadFile/;
use Getopt::Long;
use File::Path;
use Time::HiRes qw/time/;
use File::Slurp;
use Data::Dump qw/dump/;
use Storable qw/dclone/;

use Proc::Queue size => 1;
use POSIX ":sys_wait_h"; # imports WNOHANG

=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.

=item --parallel 4

Run databases in parallel (aproximatly same as number of processors in
machine if you want to use full load)

=item --only-links

Create just links

=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;
my $parallel = 0;
my $only_links = 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,
        "parallel=i" => \$parallel,
        "only-links!" => \$only_links,
);

$config = LoadFile($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;

if ($parallel) {
        $log->info("Using $parallel processes for speedup");
        Proc::Queue::size($parallel);
}

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 ($parallel) {
                my $f=fork;
                if(defined ($f) and $f==0) {
                        $log->info("Created processes $$ for speedup");
                } else {
                        next;
                }
        }

        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;
                $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);
        }


        #
        # store 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, sub {
                                        $log->info("adding link $database -> $link->{to} [$link->{credit}]");
                                        $indexer->add_link(
                                                from => $database,
                                                to => $link->{to},
                                                credit => $link->{credit},
                                        );
                                };
                        } else {
                                $log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
                        }
                }
        }
        next if ($only_links);


        #
        # 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 $lookup;
                if ($input->{lookup}) {
                        $lookup = new WebPAC::Lookup(
                                lookup_file => $input->{lookup},
                        );
                        delete( $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,
                        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},
                );
                $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 = new WebPAC::Output::MARC(
                                path => $normalize->{output},
                                lint => $marc_lint,
                                dump => $marc_dump,
                        ) if ($normalize->{output});

                        # 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 $ds_config = dclone($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,
                                        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) {
                                        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);
                                }

                                $total_rows++;
                        }

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

                        # close MARC file
                        $marc->finish if ($marc);

                }

        }

        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)
                )
        );


        # end forked process
        if ($parallel) {
                $log->info("parallel process $$ finished");
                exit(0);
        }

}

if ($parallel) {
        # wait all children to finish
        sleep(1) while wait != -1;
        $log->info("all parallel processes finished");
}


foreach my $link (@links) {
        $log->logdie("coderef in link ", Dumper($link), " is ", ref($link), " and not CODE") unless (ref($link) eq 'CODE');
        $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	588	use WebPAC::Lookup 0.03;
11	dpavlin	585	use WebPAC::Input 0.07;
12	dpavlin	209	use WebPAC::Store 0.03;
13	dpavlin	581	use WebPAC::Normalize 0.11;
14	dpavlin	74	use WebPAC::Output::TT;
15	dpavlin	516	use WebPAC::Validate;
16	dpavlin	578	use WebPAC::Output::MARC;
17	dpavlin	141	use YAML qw/LoadFile/;
18	dpavlin	301	use Getopt::Long;
19			use File::Path;
20	dpavlin	389	use Time::HiRes qw/time/;
21	dpavlin	492	use File::Slurp;
22	dpavlin	556	use Data::Dump qw/dump/;
23	dpavlin	595	use Storable qw/dclone/;
24	dpavlin	74
25	dpavlin	606	use Proc::Queue size => 1;
26			use POSIX ":sys_wait_h"; # imports WNOHANG
27
28	dpavlin	301	=head1 NAME
29	dpavlin	76
30	dpavlin	301	run.pl - start WebPAC indexing
31	dpavlin	141
32	dpavlin	301	B<this command will probably go away. Don't get used to it!>
33	dpavlin	141
34	dpavlin	301	Options:
35
36			=over 4
37
38			=item --offset 42
39
40			start loading (all) databases at offset 42
41
42			=item --limit 100
43
44			limit loading to 100 records
45
46			=item --clean
47
48			remove database and Hyper Estraier index before indexing
49
50	dpavlin	510	=item --only=database_name/input_filter
51	dpavlin	335
52	dpavlin	423	reindex just single database (legacy name is --one)
53	dpavlin	335
54	dpavlin	510	C</input_filter> is optional part which can be C<name>
55			or C<type> from input
56
57	dpavlin	301	=item --config conf/config.yml
58
59			path to YAML configuration file
60
61	dpavlin	507	=item --stats
62
63	dpavlin	509	disable indexing and dump statistics about field and subfield
64			usage for each input
65	dpavlin	507
66	dpavlin	516	=item --validate path/to/validation_file
67
68			turn on extra validation of imput records, see L<WebPAC::Validation>
69
70	dpavlin	552	=item --marc-normalize conf/normalize/mapping.pl
71
72			This option specifies normalisation file for MARC creation
73
74			=item --marc-output out/marc/test.marc
75
76			Optional path to output file
77
78	dpavlin	556	=item --marc-lint
79
80			By default turned on if C<--marc-normalize> is used. You can disable lint
81	dpavlin	558	messages with C<--no-marc-lint>.
82	dpavlin	556
83	dpavlin	559	=item --marc-dump
84
85			Force dump or input and marc record for debugging.
86
87	dpavlin	606	=item --parallel 4
88
89			Run databases in parallel (aproximatly same as number of processors in
90			machine if you want to use full load)
91
92	dpavlin	607	=item --only-links
93
94			Create just links
95
96	dpavlin	301	=back
97
98			=cut
99
100			my $offset;
101			my $limit;
102
103			my $clean = 0;
104			my $config = 'conf/config.yml';
105			my $debug = 0;
106	dpavlin	510	my $only_filter;
107	dpavlin	507	my $stats = 0;
108	dpavlin	516	my $validate_path;
109	dpavlin	552	my ($marc_normalize, $marc_output);
110	dpavlin	556	my $marc_lint = 1;
111	dpavlin	559	my $marc_dump = 0;
112	dpavlin	606	my $parallel = 0;
113	dpavlin	607	my $only_links = 0;
114	dpavlin	606
115	dpavlin	301	GetOptions(
116			"limit=i" => \$limit,
117			"offset=i" => \$offset,
118			"clean" => \$clean,
119	dpavlin	510	"one=s" => \$only_filter,
120			"only=s" => \$only_filter,
121	dpavlin	301	"config" => \$config,
122	dpavlin	560	"debug+" => \$debug,
123	dpavlin	507	"stats" => \$stats,
124	dpavlin	516	"validate=s" => \$validate_path,
125	dpavlin	552	"marc-normalize=s" => \$marc_normalize,
126			"marc-output=s" => \$marc_output,
127	dpavlin	556	"marc-lint!" => \$marc_lint,
128	dpavlin	559	"marc-dump!" => \$marc_dump,
129	dpavlin	606	"parallel=i" => \$parallel,
130	dpavlin	607	"only-links!" => \$only_links,
131	dpavlin	301	);
132
133			$config = LoadFile($config);
134
135	dpavlin	556	print "config = ",dump($config) if ($debug);
136	dpavlin	301
137	dpavlin	210	die "no databases in config file!\n" unless ($config->{databases});
138	dpavlin	431
139			my $log = _new WebPAC::Common()->_get_logger();
140	dpavlin	509	$log->info( "-" x 79 );
141	dpavlin	431
142	dpavlin	516	my $validate;
143			$validate = new WebPAC::Validate(
144			path => $validate_path,
145			) if ($validate_path);
146
147	dpavlin	430	my $use_indexer = $config->{use_indexer} \|\| 'hyperestraier';
148	dpavlin	509	if ($stats) {
149			$log->debug("option --stats disables update of indexing engine...");
150			$use_indexer = undef;
151			} else {
152			$log->info("using $use_indexer indexing engine...");
153			}
154	dpavlin	141
155	dpavlin	552	# disable indexing when creating marc
156			$use_indexer = undef if ($marc_normalize);
157
158	dpavlin	213	my $total_rows = 0;
159	dpavlin	389	my $start_t = time();
160	dpavlin	213
161	dpavlin	529	my @links;
162
163	dpavlin	606	if ($parallel) {
164			$log->info("Using $parallel processes for speedup");
165			Proc::Queue::size($parallel);
166			}
167
168	dpavlin	210	while (my ($database, $db_config) = each %{ $config->{databases} }) {
169	dpavlin	141
170	dpavlin	512	my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter);
171	dpavlin	510	next if ($only_database && $database !~ m/$only_database/i);
172	dpavlin	335
173	dpavlin	606	if ($parallel) {
174			my $f=fork;
175			if(defined ($f) and $f==0) {
176			$log->info("Created processes $$ for speedup");
177			} else {
178			next;
179			}
180			}
181
182	dpavlin	607	my $indexer;
183	dpavlin	509	if ($use_indexer) {
184			my $indexer_config = $config->{$use_indexer} \|\| $log->logdie("can't find '$use_indexer' part in confguration");
185			$indexer_config->{database} = $database;
186			$indexer_config->{clean} = $clean;
187			$indexer_config->{label} = $db_config->{name};
188	dpavlin	431
189	dpavlin	509	if ($use_indexer eq 'hyperestraier') {
190	dpavlin	255
191	dpavlin	509	# open Hyper Estraier database
192			use WebPAC::Output::Estraier '0.10';
193			$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
194
195			} elsif ($use_indexer eq 'kinosearch') {
196	dpavlin	430
197	dpavlin	509	# open KinoSearch
198			use WebPAC::Output::KinoSearch;
199			$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path});
200			$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } );
201	dpavlin	431
202	dpavlin	509	} else {
203			$log->logdie("unknown use_indexer: $use_indexer");
204			}
205
206			$log->logide("can't continue without valid indexer") unless ($indexer);
207	dpavlin	430	}
208
209
210	dpavlin	255	#
211	dpavlin	607	# store Hyper Estraier links to other databases
212			#
213			if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) {
214			foreach my $link (@{ $db_config->{links} }) {
215			if ($use_indexer eq 'hyperestraier') {
216			$log->info("saving link $database -> $link->{to} [$link->{credit}]");
217			push @links, sub {
218			$log->info("adding link $database -> $link->{to} [$link->{credit}]");
219			$indexer->add_link(
220			from => $database,
221			to => $link->{to},
222			credit => $link->{credit},
223			);
224			};
225			} else {
226			$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
227			}
228			}
229			}
230			next if ($only_links);
231
232
233			#
234	dpavlin	255	# now WebPAC::Store
235			#
236	dpavlin	210	my $abs_path = abs_path($0);
237			$abs_path =~ s#/[^/]*$#/#;
238	dpavlin	141
239	dpavlin	210	my $db_path = $config->{webpac}->{db_path} . '/' . $database;
240	dpavlin	74
241	dpavlin	301	if ($clean) {
242	dpavlin	511	$log->info("creating new database '$database' in $db_path");
243	dpavlin	301	rmtree( $db_path ) \|\| $log->warn("can't remove $db_path: $!");
244			} else {
245	dpavlin	511	$log->info("working on database '$database' in $db_path");
246	dpavlin	301	}
247	dpavlin	255
248	dpavlin	210	my $db = new WebPAC::Store(
249			path => $db_path,
250	dpavlin	217	database => $database,
251	dpavlin	301	debug => $debug,
252	dpavlin	210	);
253	dpavlin	74
254	dpavlin	233
255	dpavlin	213	#
256			# now, iterate through input formats
257			#
258	dpavlin	74
259	dpavlin	213	my @inputs;
260			if (ref($db_config->{input}) eq 'ARRAY') {
261			@inputs = @{ $db_config->{input} };
262	dpavlin	255	} elsif ($db_config->{input}) {
263			push @inputs, $db_config->{input};
264	dpavlin	213	} else {
265	dpavlin	255	$log->info("database $database doesn't have inputs defined");
266	dpavlin	213	}
267	dpavlin	74
268	dpavlin	286	my @supported_inputs = keys %{ $config->{webpac}->{inputs} };
269
270	dpavlin	213	foreach my $input (@inputs) {
271	dpavlin	233
272	dpavlin	512	next if ($only_input && ($input->{name} !~ m#$only_input#i && $input->{type} !~ m#$only_input#i));
273	dpavlin	510
274	dpavlin	233	my $type = lc($input->{type});
275
276	dpavlin	286	die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs));
277	dpavlin	233
278	dpavlin	588	my $lookup;
279			if ($input->{lookup}) {
280			$lookup = new WebPAC::Lookup(
281			lookup_file => $input->{lookup},
282			);
283			delete( $input->{lookup} );
284			}
285	dpavlin	233
286	dpavlin	286	my $input_module = $config->{webpac}->{inputs}->{$type};
287	dpavlin	74
288	dpavlin	523	$log->info("working on input '$input->{name}' in $input->{path} [type: $input->{type}] using $input_module",
289			$input->{lookup} ? "lookup '$input->{lookup}'" : ""
290			);
291	dpavlin	286
292	dpavlin	287	my $input_db = new WebPAC::Input(
293			module => $input_module,
294	dpavlin	585	encoding => $config->{webpac}->{webpac_encoding},
295	dpavlin	301	limit => $limit \|\| $input->{limit},
296			offset => $offset,
297	dpavlin	585	lookup_coderef => sub {
298			my $rec = shift \|\| return;
299			$lookup->add( $rec );
300			},
301	dpavlin	416	recode => $input->{recode},
302	dpavlin	507	stats => $stats,
303	dpavlin	600	modify_records => $input->{modify_records},
304	dpavlin	287	);
305	dpavlin	286	$log->logdie("can't create input using $input_module") unless ($input);
306	dpavlin	113
307	dpavlin	287	my $maxmfn = $input_db->open(
308	dpavlin	285	path => $input->{path},
309	dpavlin	213	code_page => $input->{encoding}, # database encoding
310	dpavlin	523	%{ $input },
311	dpavlin	213	);
312	dpavlin	113
313	dpavlin	539	my @norm_array = ref($input->{normalize}) eq 'ARRAY' ?
314			@{ $input->{normalize} } : ( $input->{normalize} );
315	dpavlin	210
316	dpavlin	552	if ($marc_normalize) {
317			@norm_array = ( {
318			path => $marc_normalize,
319			output => $marc_output \|\| 'out/marc/' . $database . '-' . $input->{name} . '.marc',
320			} );
321			}
322
323	dpavlin	539	foreach my $normalize (@norm_array) {
324	dpavlin	492
325	dpavlin	539	my $normalize_path = $normalize->{path} \|\| $log->logdie("can't find normalize path in config");
326	dpavlin	269
327	dpavlin	539	$log->logdie("Found '$normalize_path' as normalization file which isn't supported any more!") unless ( $normalize_path =~ m!\.pl$!i );
328	dpavlin	210
329	dpavlin	539	my $rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!";
330	dpavlin	213
331	dpavlin	539	$log->info("Using $normalize_path for normalization...");
332	dpavlin	213
333	dpavlin	578	my $marc = new WebPAC::Output::MARC(
334			path => $normalize->{output},
335			lint => $marc_lint,
336			dump => $marc_dump,
337			) if ($normalize->{output});
338	dpavlin	540
339	dpavlin	539	# reset position in database
340			$input_db->seek(1);
341	dpavlin	290
342	dpavlin	539	foreach my $pos ( 0 ... $input_db->size ) {
343	dpavlin	516
344	dpavlin	539	my $row = $input_db->fetch \|\| next;
345	dpavlin	516
346	dpavlin	539	my $mfn = $row->{'000'}->[0];
347	dpavlin	213
348	dpavlin	539	if (! $mfn \|\| $mfn !~ m#^\d+$#) {
349			$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
350			$mfn = $pos;
351			push @{ $row->{'000'} }, $pos;
352			}
353	dpavlin	508
354	dpavlin	213
355	dpavlin	539	if ($validate) {
356			my @errors = $validate->validate_errors( $row );
357			$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors);
358			}
359	dpavlin	213
360	dpavlin	595	my $ds_config = dclone($db_config);
361
362			# default values -> database key
363			$ds_config->{_} = $database;
364
365			# current mfn
366			$ds_config->{_mfn} = $mfn;
367
368			# attach current input
369			$ds_config->{input} = $input;
370
371	dpavlin	539	my $ds = WebPAC::Normalize::data_structure(
372			row => $row,
373			rules => $rules,
374			lookup => $lookup ? $lookup->lookup_hash : undef,
375	dpavlin	595	config => $ds_config,
376	dpavlin	541	marc_encoding => 'utf-8',
377	dpavlin	539	);
378	dpavlin	507
379	dpavlin	539	$db->save_ds(
380			id => $mfn,
381			ds => $ds,
382			prefix => $input->{name},
383			) if ($ds && !$stats);
384	dpavlin	210
385	dpavlin	539	$indexer->add(
386			id => $input->{name} . "/" . $mfn,
387			ds => $ds,
388			type => $config->{$use_indexer}->{type},
389	dpavlin	540	) if ($indexer && $ds);
390	dpavlin	539
391	dpavlin	580	if ($marc) {
392			my $i = 0;
393	dpavlin	559
394	dpavlin	580	while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) {
395			$marc->add(
396			id => $mfn . ( $i ? "/$i" : '' ),
397			fields => $fields,
398			leader => WebPAC::Normalize::marc_leader(),
399			row => $row,
400			);
401			$i++;
402			}
403
404			$log->info("Created $i instances of MFN $mfn\n") if ($i > 1);
405			}
406
407	dpavlin	539	$total_rows++;
408			}
409
410			$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats);
411
412	dpavlin	540	# close MARC file
413	dpavlin	578	$marc->finish if ($marc);
414	dpavlin	539
415	dpavlin	540	}
416
417	dpavlin	539	}
418
419	dpavlin	509	eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
420	dpavlin	434
421	dpavlin	389	my $dt = time() - $start_t;
422	dpavlin	511	$log->info("$total_rows records ", $indexer ? "indexed " : "",
423			sprintf("in %.2f sec [%.2f rec/sec]",
424	dpavlin	389	$dt, ($total_rows / $dt)
425			)
426			);
427	dpavlin	255
428
429	dpavlin	606	# end forked process
430			if ($parallel) {
431			$log->info("parallel process $$ finished");
432			exit(0);
433			}
434
435	dpavlin	210	}
436
437	dpavlin	606	if ($parallel) {
438			# wait all children to finish
439			sleep(1) while wait != -1;
440			$log->info("all parallel processes finished");
441			}
442
443	dpavlin	607
444	dpavlin	529	foreach my $link (@links) {
445	dpavlin	607	$log->logdie("coderef in link ", Dumper($link), " is ", ref($link), " and not CODE") unless (ref($link) eq 'CODE');
446			$link->();
447	dpavlin	529	}
448	dpavlin	606