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/;

=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,
                        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)
                )
        );

        #
        # 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	use Storable qw/dclone/;
24
25	=head1 NAME
26
27	run.pl - start WebPAC indexing
28
29	B<this command will probably go away. Don't get used to it!>
30
31	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	=item --only=database_name/input_filter
48
49	reindex just single database (legacy name is --one)
50
51	C</input_filter> is optional part which can be C<name>
52	or C<type> from input
53
54	=item --config conf/config.yml
55
56	path to YAML configuration file
57
58	=item --stats
59
60	disable indexing and dump statistics about field and subfield
61	usage for each input
62
63	=item --validate path/to/validation_file
64
65	turn on extra validation of imput records, see L<WebPAC::Validation>
66
67	=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	=item --marc-lint
76
77	By default turned on if C<--marc-normalize> is used. You can disable lint
78	messages with C<--no-marc-lint>.
79
80	=item --marc-dump
81
82	Force dump or input and marc record for debugging.
83
84	=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	my $only_filter;
95	my $stats = 0;
96	my $validate_path;
97	my ($marc_normalize, $marc_output);
98	my $marc_lint = 1;
99	my $marc_dump = 0;
100
101	GetOptions(
102	"limit=i" => \$limit,
103	"offset=i" => \$offset,
104	"clean" => \$clean,
105	"one=s" => \$only_filter,
106	"only=s" => \$only_filter,
107	"config" => \$config,
108	"debug+" => \$debug,
109	"stats" => \$stats,
110	"validate=s" => \$validate_path,
111	"marc-normalize=s" => \$marc_normalize,
112	"marc-output=s" => \$marc_output,
113	"marc-lint!" => \$marc_lint,
114	"marc-dump!" => \$marc_dump,
115	);
116
117	$config = LoadFile($config);
118
119	print "config = ",dump($config) if ($debug);
120
121	die "no databases in config file!\n" unless ($config->{databases});
122
123	my $log = _new WebPAC::Common()->_get_logger();
124	$log->info( "-" x 79 );
125
126	my $validate;
127	$validate = new WebPAC::Validate(
128	path => $validate_path,
129	) if ($validate_path);
130
131	my $use_indexer = $config->{use_indexer} \|\| 'hyperestraier';
132	if ($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
139	# disable indexing when creating marc
140	$use_indexer = undef if ($marc_normalize);
141
142	my $total_rows = 0;
143	my $start_t = time();
144
145	my @links;
146	my $indexer;
147
148	while (my ($database, $db_config) = each %{ $config->{databases} }) {
149
150	my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter);
151	next if ($only_database && $database !~ m/$only_database/i);
152
153	if ($use_indexer) {
154	my $indexer_config = $config->{$use_indexer} \|\| $log->logdie("can't find '$use_indexer' part in confguration");
155	$indexer_config->{database} = $database;
156	$indexer_config->{clean} = $clean;
157	$indexer_config->{label} = $db_config->{name};
158
159	if ($use_indexer eq 'hyperestraier') {
160
161	# open Hyper Estraier database
162	use WebPAC::Output::Estraier '0.10';
163	$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
164
165	} elsif ($use_indexer eq 'kinosearch') {
166
167	# open KinoSearch
168	use WebPAC::Output::KinoSearch;
169	$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path});
170	$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } );
171
172	} else {
173	$log->logdie("unknown use_indexer: $use_indexer");
174	}
175
176	$log->logide("can't continue without valid indexer") unless ($indexer);
177	}
178
179
180	#
181	# now WebPAC::Store
182	#
183	my $abs_path = abs_path($0);
184	$abs_path =~ s#/[^/]*$#/#;
185
186	my $db_path = $config->{webpac}->{db_path} . '/' . $database;
187
188	if ($clean) {
189	$log->info("creating new database '$database' in $db_path");
190	rmtree( $db_path ) \|\| $log->warn("can't remove $db_path: $!");
191	} else {
192	$log->info("working on database '$database' in $db_path");
193	}
194
195	my $db = new WebPAC::Store(
196	path => $db_path,
197	database => $database,
198	debug => $debug,
199	);
200
201
202	#
203	# now, iterate through input formats
204	#
205
206	my @inputs;
207	if (ref($db_config->{input}) eq 'ARRAY') {
208	@inputs = @{ $db_config->{input} };
209	} elsif ($db_config->{input}) {
210	push @inputs, $db_config->{input};
211	} else {
212	$log->info("database $database doesn't have inputs defined");
213	}
214
215	my @supported_inputs = keys %{ $config->{webpac}->{inputs} };
216
217	foreach my $input (@inputs) {
218
219	next if ($only_input && ($input->{name} !~ m#$only_input#i && $input->{type} !~ m#$only_input#i));
220
221	my $type = lc($input->{type});
222
223	die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs));
224
225	my $lookup;
226	if ($input->{lookup}) {
227	$lookup = new WebPAC::Lookup(
228	lookup_file => $input->{lookup},
229	);
230	delete( $input->{lookup} );
231	}
232
233	my $input_module = $config->{webpac}->{inputs}->{$type};
234
235	$log->info("working on input '$input->{name}' in $input->{path} [type: $input->{type}] using $input_module",
236	$input->{lookup} ? "lookup '$input->{lookup}'" : ""
237	);
238
239	my $input_db = new WebPAC::Input(
240	module => $input_module,
241	encoding => $config->{webpac}->{webpac_encoding},
242	limit => $limit \|\| $input->{limit},
243	offset => $offset,
244	lookup_coderef => sub {
245	my $rec = shift \|\| return;
246	$lookup->add( $rec );
247	},
248	recode => $input->{recode},
249	stats => $stats,
250	modify_records => $input->{modify_records},
251	);
252	$log->logdie("can't create input using $input_module") unless ($input);
253
254	my $maxmfn = $input_db->open(
255	path => $input->{path},
256	code_page => $input->{encoding}, # database encoding
257	%{ $input },
258	);
259
260	my @norm_array = ref($input->{normalize}) eq 'ARRAY' ?
261	@{ $input->{normalize} } : ( $input->{normalize} );
262
263	if ($marc_normalize) {
264	@norm_array = ( {
265	path => $marc_normalize,
266	output => $marc_output \|\| 'out/marc/' . $database . '-' . $input->{name} . '.marc',
267	} );
268	}
269
270	foreach my $normalize (@norm_array) {
271
272	my $normalize_path = $normalize->{path} \|\| $log->logdie("can't find normalize path in config");
273
274	$log->logdie("Found '$normalize_path' as normalization file which isn't supported any more!") unless ( $normalize_path =~ m!\.pl$!i );
275
276	my $rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!";
277
278	$log->info("Using $normalize_path for normalization...");
279
280	my $marc = new WebPAC::Output::MARC(
281	path => $normalize->{output},
282	lint => $marc_lint,
283	dump => $marc_dump,
284	) if ($normalize->{output});
285
286	# reset position in database
287	$input_db->seek(1);
288
289	foreach my $pos ( 0 ... $input_db->size ) {
290
291	my $row = $input_db->fetch \|\| next;
292
293	my $mfn = $row->{'000'}->[0];
294
295	if (! $mfn \|\| $mfn !~ m#^\d+$#) {
296	$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
297	$mfn = $pos;
298	push @{ $row->{'000'} }, $pos;
299	}
300
301
302	if ($validate) {
303	my @errors = $validate->validate_errors( $row );
304	$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors);
305	}
306
307	my $ds_config = dclone($db_config);
308
309	# default values -> database key
310	$ds_config->{_} = $database;
311
312	# current mfn
313	$ds_config->{_mfn} = $mfn;
314
315	# attach current input
316	$ds_config->{input} = $input;
317
318	my $ds = WebPAC::Normalize::data_structure(
319	row => $row,
320	rules => $rules,
321	lookup => $lookup ? $lookup->lookup_hash : undef,
322	config => $ds_config,
323	marc_encoding => 'utf-8',
324	);
325
326	$db->save_ds(
327	id => $mfn,
328	ds => $ds,
329	prefix => $input->{name},
330	) if ($ds && !$stats);
331
332	$indexer->add(
333	id => $input->{name} . "/" . $mfn,
334	ds => $ds,
335	type => $config->{$use_indexer}->{type},
336	) if ($indexer && $ds);
337
338	if ($marc) {
339	my $i = 0;
340
341	while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) {
342	$marc->add(
343	id => $mfn . ( $i ? "/$i" : '' ),
344	fields => $fields,
345	leader => WebPAC::Normalize::marc_leader(),
346	row => $row,
347	);
348	$i++;
349	}
350
351	$log->info("Created $i instances of MFN $mfn\n") if ($i > 1);
352	}
353
354	$total_rows++;
355	}
356
357	$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats);
358
359	# close MARC file
360	$marc->finish if ($marc);
361
362	}
363
364	}
365
366	eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
367
368	my $dt = time() - $start_t;
369	$log->info("$total_rows records ", $indexer ? "indexed " : "",
370	sprintf("in %.2f sec [%.2f rec/sec]",
371	$dt, ($total_rows / $dt)
372	)
373	);
374
375	#
376	# add Hyper Estraier links to other databases
377	#
378	if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) {
379	foreach my $link (@{ $db_config->{links} }) {
380	if ($use_indexer eq 'hyperestraier') {
381	$log->info("saving link $database -> $link->{to} [$link->{credit}]");
382	push @links, {
383	from => $database,
384	to => $link->{to},
385	credit => $link->{credit},
386	};
387	} else {
388	$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
389	}
390	}
391	}
392
393	}
394
395	foreach my $link (@links) {
396	$log->info("adding link $link->{from} -> $link->{to} [$link->{credit}]");
397	$indexer->add_link( %{ $link } );
398	}