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/;

=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;

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 $lookup;
                if ($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,
                );
                $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 = 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) {
                                        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)
                )
        );

        #
        # 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	#!/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	use WebPAC::Common 0.02;
10	use WebPAC::Lookup 0.03;
11	use WebPAC::Input 0.07;
12	use WebPAC::Store 0.03;
13	use WebPAC::Normalize 0.11;
14	use WebPAC::Output::TT;
15	use WebPAC::Validate;
16	use WebPAC::Output::MARC;
17	use YAML qw/LoadFile/;
18	use Getopt::Long;
19	use File::Path;
20	use Time::HiRes qw/time/;
21	use File::Slurp;
22	use Data::Dump qw/dump/;
23
24	=head1 NAME
25
26	run.pl - start WebPAC indexing
27
28	B<this command will probably go away. Don't get used to it!>
29
30	Options:
31
32	=over 4
33
34	=item --offset 42
35
36	start loading (all) databases at offset 42
37
38	=item --limit 100
39
40	limit loading to 100 records
41
42	=item --clean
43
44	remove database and Hyper Estraier index before indexing
45
46	=item --only=database_name/input_filter
47
48	reindex just single database (legacy name is --one)
49
50	C</input_filter> is optional part which can be C<name>
51	or C<type> from input
52
53	=item --config conf/config.yml
54
55	path to YAML configuration file
56
57	=item --stats
58
59	disable indexing and dump statistics about field and subfield
60	usage for each input
61
62	=item --validate path/to/validation_file
63
64	turn on extra validation of imput records, see L<WebPAC::Validation>
65
66	=item --marc-normalize conf/normalize/mapping.pl
67
68	This option specifies normalisation file for MARC creation
69
70	=item --marc-output out/marc/test.marc
71
72	Optional path to output file
73
74	=item --marc-lint
75
76	By default turned on if C<--marc-normalize> is used. You can disable lint
77	messages with C<--no-marc-lint>.
78
79	=item --marc-dump
80
81	Force dump or input and marc record for debugging.
82
83	=back
84
85	=cut
86
87	my $offset;
88	my $limit;
89
90	my $clean = 0;
91	my $config = 'conf/config.yml';
92	my $debug = 0;
93	my $only_filter;
94	my $stats = 0;
95	my $validate_path;
96	my ($marc_normalize, $marc_output);
97	my $marc_lint = 1;
98	my $marc_dump = 0;
99
100	GetOptions(
101	"limit=i" => \$limit,
102	"offset=i" => \$offset,
103	"clean" => \$clean,
104	"one=s" => \$only_filter,
105	"only=s" => \$only_filter,
106	"config" => \$config,
107	"debug+" => \$debug,
108	"stats" => \$stats,
109	"validate=s" => \$validate_path,
110	"marc-normalize=s" => \$marc_normalize,
111	"marc-output=s" => \$marc_output,
112	"marc-lint!" => \$marc_lint,
113	"marc-dump!" => \$marc_dump,
114	);
115
116	$config = LoadFile($config);
117
118	print "config = ",dump($config) if ($debug);
119
120	die "no databases in config file!\n" unless ($config->{databases});
121
122	my $log = _new WebPAC::Common()->_get_logger();
123	$log->info( "-" x 79 );
124
125	my $validate;
126	$validate = new WebPAC::Validate(
127	path => $validate_path,
128	) if ($validate_path);
129
130	my $use_indexer = $config->{use_indexer} \|\| 'hyperestraier';
131	if ($stats) {
132	$log->debug("option --stats disables update of indexing engine...");
133	$use_indexer = undef;
134	} else {
135	$log->info("using $use_indexer indexing engine...");
136	}
137
138	# disable indexing when creating marc
139	$use_indexer = undef if ($marc_normalize);
140
141	my $total_rows = 0;
142	my $start_t = time();
143
144	my @links;
145	my $indexer;
146
147	while (my ($database, $db_config) = each %{ $config->{databases} }) {
148
149	my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter);
150	next if ($only_database && $database !~ m/$only_database/i);
151
152	if ($use_indexer) {
153	my $indexer_config = $config->{$use_indexer} \|\| $log->logdie("can't find '$use_indexer' part in confguration");
154	$indexer_config->{database} = $database;
155	$indexer_config->{clean} = $clean;
156	$indexer_config->{label} = $db_config->{name};
157
158	if ($use_indexer eq 'hyperestraier') {
159
160	# open Hyper Estraier database
161	use WebPAC::Output::Estraier '0.10';
162	$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
163
164	} elsif ($use_indexer eq 'kinosearch') {
165
166	# open KinoSearch
167	use WebPAC::Output::KinoSearch;
168	$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path});
169	$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } );
170
171	} else {
172	$log->logdie("unknown use_indexer: $use_indexer");
173	}
174
175	$log->logide("can't continue without valid indexer") unless ($indexer);
176	}
177
178
179	#
180	# now WebPAC::Store
181	#
182	my $abs_path = abs_path($0);
183	$abs_path =~ s#/[^/]*$#/#;
184
185	my $db_path = $config->{webpac}->{db_path} . '/' . $database;
186
187	if ($clean) {
188	$log->info("creating new database '$database' in $db_path");
189	rmtree( $db_path ) \|\| $log->warn("can't remove $db_path: $!");
190	} else {
191	$log->info("working on database '$database' in $db_path");
192	}
193
194	my $db = new WebPAC::Store(
195	path => $db_path,
196	database => $database,
197	debug => $debug,
198	);
199
200
201	#
202	# now, iterate through input formats
203	#
204
205	my @inputs;
206	if (ref($db_config->{input}) eq 'ARRAY') {
207	@inputs = @{ $db_config->{input} };
208	} elsif ($db_config->{input}) {
209	push @inputs, $db_config->{input};
210	} else {
211	$log->info("database $database doesn't have inputs defined");
212	}
213
214	my @supported_inputs = keys %{ $config->{webpac}->{inputs} };
215
216	foreach my $input (@inputs) {
217
218	next if ($only_input && ($input->{name} !~ m#$only_input#i && $input->{type} !~ m#$only_input#i));
219
220	my $type = lc($input->{type});
221
222	die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs));
223
224	my $lookup;
225	if ($input->{lookup}) {
226	$lookup = new WebPAC::Lookup(
227	lookup_file => $input->{lookup},
228	);
229	delete( $input->{lookup} );
230	}
231
232	my $input_module = $config->{webpac}->{inputs}->{$type};
233
234	$log->info("working on input '$input->{name}' in $input->{path} [type: $input->{type}] using $input_module",
235	$input->{lookup} ? "lookup '$input->{lookup}'" : ""
236	);
237
238	my $input_db = new WebPAC::Input(
239	module => $input_module,
240	encoding => $config->{webpac}->{webpac_encoding},
241	limit => $limit \|\| $input->{limit},
242	offset => $offset,
243	lookup_coderef => sub {
244	my $rec = shift \|\| return;
245	$lookup->add( $rec );
246	},
247	recode => $input->{recode},
248	stats => $stats,
249	);
250	$log->logdie("can't create input using $input_module") unless ($input);
251
252	my $maxmfn = $input_db->open(
253	path => $input->{path},
254	code_page => $input->{encoding}, # database encoding
255	%{ $input },
256	);
257
258	my @norm_array = ref($input->{normalize}) eq 'ARRAY' ?
259	@{ $input->{normalize} } : ( $input->{normalize} );
260
261	if ($marc_normalize) {
262	@norm_array = ( {
263	path => $marc_normalize,
264	output => $marc_output \|\| 'out/marc/' . $database . '-' . $input->{name} . '.marc',
265	} );
266	}
267
268	foreach my $normalize (@norm_array) {
269
270	my $normalize_path = $normalize->{path} \|\| $log->logdie("can't find normalize path in config");
271
272	$log->logdie("Found '$normalize_path' as normalization file which isn't supported any more!") unless ( $normalize_path =~ m!\.pl$!i );
273
274	my $rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!";
275
276	$log->info("Using $normalize_path for normalization...");
277
278	my $marc = new WebPAC::Output::MARC(
279	path => $normalize->{output},
280	lint => $marc_lint,
281	dump => $marc_dump,
282	) if ($normalize->{output});
283
284	# reset position in database
285	$input_db->seek(1);
286
287	foreach my $pos ( 0 ... $input_db->size ) {
288
289	my $row = $input_db->fetch \|\| next;
290
291	my $mfn = $row->{'000'}->[0];
292
293	if (! $mfn \|\| $mfn !~ m#^\d+$#) {
294	$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
295	$mfn = $pos;
296	push @{ $row->{'000'} }, $pos;
297	}
298
299
300	if ($validate) {
301	my @errors = $validate->validate_errors( $row );
302	$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors);
303	}
304
305	my $ds = WebPAC::Normalize::data_structure(
306	row => $row,
307	rules => $rules,
308	lookup => $lookup ? $lookup->lookup_hash : undef,
309	marc_encoding => 'utf-8',
310	);
311
312	$db->save_ds(
313	id => $mfn,
314	ds => $ds,
315	prefix => $input->{name},
316	) if ($ds && !$stats);
317
318	$indexer->add(
319	id => $input->{name} . "/" . $mfn,
320	ds => $ds,
321	type => $config->{$use_indexer}->{type},
322	) if ($indexer && $ds);
323
324	if ($marc) {
325	my $i = 0;
326
327	while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) {
328	$marc->add(
329	id => $mfn . ( $i ? "/$i" : '' ),
330	fields => $fields,
331	leader => WebPAC::Normalize::marc_leader(),
332	row => $row,
333	);
334	$i++;
335	}
336
337	$log->info("Created $i instances of MFN $mfn\n") if ($i > 1);
338	}
339
340	$total_rows++;
341	}
342
343	$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats);
344
345	# close MARC file
346	$marc->finish if ($marc);
347
348	}
349
350	}
351
352	eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
353
354	my $dt = time() - $start_t;
355	$log->info("$total_rows records ", $indexer ? "indexed " : "",
356	sprintf("in %.2f sec [%.2f rec/sec]",
357	$dt, ($total_rows / $dt)
358	)
359	);
360
361	#
362	# add Hyper Estraier links to other databases
363	#
364	if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) {
365	foreach my $link (@{ $db_config->{links} }) {
366	if ($use_indexer eq 'hyperestraier') {
367	$log->info("saving link $database -> $link->{to} [$link->{credit}]");
368	push @links, {
369	from => $database,
370	to => $link->{to},
371	credit => $link->{credit},
372	};
373	} else {
374	$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
375	}
376	}
377	}
378
379	}
380
381	foreach my $link (@links) {
382	$log->info("adding link $link->{from} -> $link->{to} [$link->{credit}]");
383	$indexer->add_link( %{ $link } );
384	}