webpac2/trunk/run.pl

#!/usr/bin/perl -w

use strict;

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

use WebPAC::Common 0.02;
use WebPAC::Lookup;
use WebPAC::Input 0.03;
use WebPAC::Store 0.03;
use WebPAC::Normalize;
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

=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>

=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;

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,
);

$config = LoadFile($config);

print "config = ",Dumper($config) if ($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...");
}

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 = 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} );

                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 = $normalize->{marc21}) {
                                open($marc_fh, '>', $path) ||
                                        $log->logdie("can't open MARC output $path: $!");

                                $log->info("Creating MARC export file $path\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' );
                                        $marc->add_fields( WebPAC::Normalize::_get_marc21_fields() );
                                        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') {
                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 Data::Dumper;
8	use lib './lib';
9
10	use WebPAC::Common 0.02;
11	use WebPAC::Lookup;
12	use WebPAC::Input 0.03;
13	use WebPAC::Store 0.03;
14	use WebPAC::Normalize;
15	use WebPAC::Output::TT;
16	use WebPAC::Validate;
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 MARC::Record 2.0; # need 2.0 for utf-8 encoding see marcpm.sf.net
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	=back
67
68	=cut
69
70	my $offset;
71	my $limit;
72
73	my $clean = 0;
74	my $config = 'conf/config.yml';
75	my $debug = 0;
76	my $only_filter;
77	my $stats = 0;
78	my $validate_path;
79
80	GetOptions(
81	"limit=i" => \$limit,
82	"offset=i" => \$offset,
83	"clean" => \$clean,
84	"one=s" => \$only_filter,
85	"only=s" => \$only_filter,
86	"config" => \$config,
87	"debug" => \$debug,
88	"stats" => \$stats,
89	"validate=s" => \$validate_path,
90	);
91
92	$config = LoadFile($config);
93
94	print "config = ",Dumper($config) if ($debug);
95
96	die "no databases in config file!\n" unless ($config->{databases});
97
98	my $log = _new WebPAC::Common()->_get_logger();
99	$log->info( "-" x 79 );
100
101	my $validate;
102	$validate = new WebPAC::Validate(
103	path => $validate_path,
104	) if ($validate_path);
105
106	my $use_indexer = $config->{use_indexer} \|\| 'hyperestraier';
107	if ($stats) {
108	$log->debug("option --stats disables update of indexing engine...");
109	$use_indexer = undef;
110	} else {
111	$log->info("using $use_indexer indexing engine...");
112	}
113
114	my $total_rows = 0;
115	my $start_t = time();
116
117	my @links;
118	my $indexer;
119
120	while (my ($database, $db_config) = each %{ $config->{databases} }) {
121
122	my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter);
123	next if ($only_database && $database !~ m/$only_database/i);
124
125	if ($use_indexer) {
126	my $indexer_config = $config->{$use_indexer} \|\| $log->logdie("can't find '$use_indexer' part in confguration");
127	$indexer_config->{database} = $database;
128	$indexer_config->{clean} = $clean;
129	$indexer_config->{label} = $db_config->{name};
130
131	if ($use_indexer eq 'hyperestraier') {
132
133	# open Hyper Estraier database
134	use WebPAC::Output::Estraier '0.10';
135	$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
136
137	} elsif ($use_indexer eq 'kinosearch') {
138
139	# open KinoSearch
140	use WebPAC::Output::KinoSearch;
141	$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path});
142	$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } );
143
144	} else {
145	$log->logdie("unknown use_indexer: $use_indexer");
146	}
147
148	$log->logide("can't continue without valid indexer") unless ($indexer);
149	}
150
151
152	#
153	# now WebPAC::Store
154	#
155	my $abs_path = abs_path($0);
156	$abs_path =~ s#/[^/]*$#/#;
157
158	my $db_path = $config->{webpac}->{db_path} . '/' . $database;
159
160	if ($clean) {
161	$log->info("creating new database '$database' in $db_path");
162	rmtree( $db_path ) \|\| $log->warn("can't remove $db_path: $!");
163	} else {
164	$log->info("working on database '$database' in $db_path");
165	}
166
167	my $db = new WebPAC::Store(
168	path => $db_path,
169	database => $database,
170	debug => $debug,
171	);
172
173
174	#
175	# now, iterate through input formats
176	#
177
178	my @inputs;
179	if (ref($db_config->{input}) eq 'ARRAY') {
180	@inputs = @{ $db_config->{input} };
181	} elsif ($db_config->{input}) {
182	push @inputs, $db_config->{input};
183	} else {
184	$log->info("database $database doesn't have inputs defined");
185	}
186
187	my @supported_inputs = keys %{ $config->{webpac}->{inputs} };
188
189	foreach my $input (@inputs) {
190
191	next if ($only_input && ($input->{name} !~ m#$only_input#i && $input->{type} !~ m#$only_input#i));
192
193	my $type = lc($input->{type});
194
195	die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs));
196
197	my $lookup = new WebPAC::Lookup(
198	lookup_file => $input->{lookup},
199	) if ($input->{lookup});
200
201	my $input_module = $config->{webpac}->{inputs}->{$type};
202
203	$log->info("working on input '$input->{name}' in $input->{path} [type: $input->{type}] using $input_module",
204	$input->{lookup} ? "lookup '$input->{lookup}'" : ""
205	);
206
207	my $input_db = new WebPAC::Input(
208	module => $input_module,
209	code_page => $config->{webpac}->{webpac_encoding},
210	limit => $limit \|\| $input->{limit},
211	offset => $offset,
212	lookup => $lookup,
213	recode => $input->{recode},
214	stats => $stats,
215	);
216	$log->logdie("can't create input using $input_module") unless ($input);
217
218	my $maxmfn = $input_db->open(
219	path => $input->{path},
220	code_page => $input->{encoding}, # database encoding
221	%{ $input },
222	);
223
224	my @norm_array = ref($input->{normalize}) eq 'ARRAY' ?
225	@{ $input->{normalize} } : ( $input->{normalize} );
226
227	foreach my $normalize (@norm_array) {
228
229	my $normalize_path = $normalize->{path} \|\| $log->logdie("can't find normalize path in config");
230
231	$log->logdie("Found '$normalize_path' as normalization file which isn't supported any more!") unless ( $normalize_path =~ m!\.pl$!i );
232
233	my $rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!";
234
235	$log->info("Using $normalize_path for normalization...");
236
237	my $marc_fh;
238	if (my $path = $normalize->{marc21}) {
239	open($marc_fh, '>', $path) \|\|
240	$log->logdie("can't open MARC output $path: $!");
241
242	$log->info("Creating MARC export file $path\n");
243	}
244
245	# reset position in database
246	$input_db->seek(1);
247
248	foreach my $pos ( 0 ... $input_db->size ) {
249
250	my $row = $input_db->fetch \|\| next;
251
252	my $mfn = $row->{'000'}->[0];
253
254	if (! $mfn \|\| $mfn !~ m#^\d+$#) {
255	$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
256	$mfn = $pos;
257	push @{ $row->{'000'} }, $pos;
258	}
259
260
261	if ($validate) {
262	my @errors = $validate->validate_errors( $row );
263	$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors);
264	}
265
266
267	my $ds = WebPAC::Normalize::data_structure(
268	row => $row,
269	rules => $rules,
270	lookup => $lookup ? $lookup->lookup_hash : undef,
271	marc_encoding => 'utf-8',
272	);
273
274	$db->save_ds(
275	id => $mfn,
276	ds => $ds,
277	prefix => $input->{name},
278	) if ($ds && !$stats);
279
280	$indexer->add(
281	id => $input->{name} . "/" . $mfn,
282	ds => $ds,
283	type => $config->{$use_indexer}->{type},
284	) if ($indexer && $ds);
285
286	if ($marc_fh) {
287	my $marc = new MARC::Record;
288	$marc->encoding( 'utf-8' );
289	$marc->add_fields( WebPAC::Normalize::_get_marc21_fields() );
290	print $marc_fh $marc->as_usmarc;
291	}
292
293	$total_rows++;
294	}
295
296	$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats);
297
298	# close MARC file
299	close($marc_fh) if ($marc_fh);
300
301	}
302
303	}
304
305	eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
306
307	my $dt = time() - $start_t;
308	$log->info("$total_rows records ", $indexer ? "indexed " : "",
309	sprintf("in %.2f sec [%.2f rec/sec]",
310	$dt, ($total_rows / $dt)
311	)
312	);
313
314	#
315	# add Hyper Estraier links to other databases
316	#
317	if (ref($db_config->{links}) eq 'ARRAY') {
318	foreach my $link (@{ $db_config->{links} }) {
319	if ($use_indexer eq 'hyperestraier') {
320	$log->info("saving link $database -> $link->{to} [$link->{credit}]");
321	push @links, {
322	from => $database,
323	to => $link->{to},
324	credit => $link->{credit},
325	};
326	} else {
327	$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
328	}
329	}
330	}
331
332	}
333
334	foreach my $link (@links) {
335	$log->info("adding link $link->{from} -> $link->{to} [$link->{credit}]");
336	$indexer->add_link( %{ $link } );
337	}