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>

=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

=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);

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,
);

$config = LoadFile($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...");
}

# 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 = new WebPAC::Lookup(
                        lookup_file => $input->{lookup},
                ) if ($input->{lookup});

                my $input_module = $config->{webpac}->{inputs}->{$type};

                $log->info("working on input '$input->{name}' in $input->{path} [type: $input->{type}] using $input_module",
                        $input->{lookup} ? "lookup '$input->{lookup}'" : ""
                );

                my $input_db = new WebPAC::Input(
                        module => $input_module,
                        code_page => $config->{webpac}->{webpac_encoding},
                        limit => $limit || $input->{limit},
                        offset => $offset,
                        lookup => $lookup,
                        recode => $input->{recode},
                        stats => $stats,
                );
                $log->logdie("can't create input using $input_module") unless ($input);

                my $maxmfn = $input_db->open(
                        path => $input->{path},
                        code_page => $input->{encoding},        # database encoding
                        %{ $input },
                );

                my @norm_array = ref($input->{normalize}) eq 'ARRAY' ?
                        @{ $input->{normalize} } : ( $input->{normalize} );

                if ($marc_normalize) {
                        @norm_array = ( {
                                path => $marc_normalize,
                                output => $marc_output || 'out/marc/' . $database . '-' . $input->{name} . '.marc',
                        } );
                }

                foreach my $normalize (@norm_array) {

                        my $normalize_path = $normalize->{path} || $log->logdie("can't find normalize path in config");

                        $log->logdie("Found '$normalize_path' as normalization file which isn't supported any more!") unless ( $normalize_path =~ m!\.pl$!i );

                        my $rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!";

                        $log->info("Using $normalize_path for normalization...");

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

                                $log->info("Creating MARC export file $path\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_marc_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' && $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 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	=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	=back
75
76	=cut
77
78	my $offset;
79	my $limit;
80
81	my $clean = 0;
82	my $config = 'conf/config.yml';
83	my $debug = 0;
84	my $only_filter;
85	my $stats = 0;
86	my $validate_path;
87	my ($marc_normalize, $marc_output);
88
89	GetOptions(
90	"limit=i" => \$limit,
91	"offset=i" => \$offset,
92	"clean" => \$clean,
93	"one=s" => \$only_filter,
94	"only=s" => \$only_filter,
95	"config" => \$config,
96	"debug" => \$debug,
97	"stats" => \$stats,
98	"validate=s" => \$validate_path,
99	"marc-normalize=s" => \$marc_normalize,
100	"marc-output=s" => \$marc_output,
101	);
102
103	$config = LoadFile($config);
104
105	print "config = ",Dumper($config) if ($debug);
106
107	die "no databases in config file!\n" unless ($config->{databases});
108
109	my $log = _new WebPAC::Common()->_get_logger();
110	$log->info( "-" x 79 );
111
112	my $validate;
113	$validate = new WebPAC::Validate(
114	path => $validate_path,
115	) if ($validate_path);
116
117	my $use_indexer = $config->{use_indexer} \|\| 'hyperestraier';
118	if ($stats) {
119	$log->debug("option --stats disables update of indexing engine...");
120	$use_indexer = undef;
121	} else {
122	$log->info("using $use_indexer indexing engine...");
123	}
124
125	# disable indexing when creating marc
126	$use_indexer = undef if ($marc_normalize);
127
128	my $total_rows = 0;
129	my $start_t = time();
130
131	my @links;
132	my $indexer;
133
134	while (my ($database, $db_config) = each %{ $config->{databases} }) {
135
136	my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter);
137	next if ($only_database && $database !~ m/$only_database/i);
138
139	if ($use_indexer) {
140	my $indexer_config = $config->{$use_indexer} \|\| $log->logdie("can't find '$use_indexer' part in confguration");
141	$indexer_config->{database} = $database;
142	$indexer_config->{clean} = $clean;
143	$indexer_config->{label} = $db_config->{name};
144
145	if ($use_indexer eq 'hyperestraier') {
146
147	# open Hyper Estraier database
148	use WebPAC::Output::Estraier '0.10';
149	$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
150
151	} elsif ($use_indexer eq 'kinosearch') {
152
153	# open KinoSearch
154	use WebPAC::Output::KinoSearch;
155	$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path});
156	$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } );
157
158	} else {
159	$log->logdie("unknown use_indexer: $use_indexer");
160	}
161
162	$log->logide("can't continue without valid indexer") unless ($indexer);
163	}
164
165
166	#
167	# now WebPAC::Store
168	#
169	my $abs_path = abs_path($0);
170	$abs_path =~ s#/[^/]*$#/#;
171
172	my $db_path = $config->{webpac}->{db_path} . '/' . $database;
173
174	if ($clean) {
175	$log->info("creating new database '$database' in $db_path");
176	rmtree( $db_path ) \|\| $log->warn("can't remove $db_path: $!");
177	} else {
178	$log->info("working on database '$database' in $db_path");
179	}
180
181	my $db = new WebPAC::Store(
182	path => $db_path,
183	database => $database,
184	debug => $debug,
185	);
186
187
188	#
189	# now, iterate through input formats
190	#
191
192	my @inputs;
193	if (ref($db_config->{input}) eq 'ARRAY') {
194	@inputs = @{ $db_config->{input} };
195	} elsif ($db_config->{input}) {
196	push @inputs, $db_config->{input};
197	} else {
198	$log->info("database $database doesn't have inputs defined");
199	}
200
201	my @supported_inputs = keys %{ $config->{webpac}->{inputs} };
202
203	foreach my $input (@inputs) {
204
205	next if ($only_input && ($input->{name} !~ m#$only_input#i && $input->{type} !~ m#$only_input#i));
206
207	my $type = lc($input->{type});
208
209	die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs));
210
211	my $lookup = new WebPAC::Lookup(
212	lookup_file => $input->{lookup},
213	) if ($input->{lookup});
214
215	my $input_module = $config->{webpac}->{inputs}->{$type};
216
217	$log->info("working on input '$input->{name}' in $input->{path} [type: $input->{type}] using $input_module",
218	$input->{lookup} ? "lookup '$input->{lookup}'" : ""
219	);
220
221	my $input_db = new WebPAC::Input(
222	module => $input_module,
223	code_page => $config->{webpac}->{webpac_encoding},
224	limit => $limit \|\| $input->{limit},
225	offset => $offset,
226	lookup => $lookup,
227	recode => $input->{recode},
228	stats => $stats,
229	);
230	$log->logdie("can't create input using $input_module") unless ($input);
231
232	my $maxmfn = $input_db->open(
233	path => $input->{path},
234	code_page => $input->{encoding}, # database encoding
235	%{ $input },
236	);
237
238	my @norm_array = ref($input->{normalize}) eq 'ARRAY' ?
239	@{ $input->{normalize} } : ( $input->{normalize} );
240
241	if ($marc_normalize) {
242	@norm_array = ( {
243	path => $marc_normalize,
244	output => $marc_output \|\| 'out/marc/' . $database . '-' . $input->{name} . '.marc',
245	} );
246	}
247
248	foreach my $normalize (@norm_array) {
249
250	my $normalize_path = $normalize->{path} \|\| $log->logdie("can't find normalize path in config");
251
252	$log->logdie("Found '$normalize_path' as normalization file which isn't supported any more!") unless ( $normalize_path =~ m!\.pl$!i );
253
254	my $rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!";
255
256	$log->info("Using $normalize_path for normalization...");
257
258	my $marc_fh;
259	if (my $path = $normalize->{output}) {
260	open($marc_fh, '>', $path) \|\|
261	$log->logdie("can't open MARC output $path: $!");
262
263	$log->info("Creating MARC export file $path\n");
264	}
265
266	# reset position in database
267	$input_db->seek(1);
268
269	foreach my $pos ( 0 ... $input_db->size ) {
270
271	my $row = $input_db->fetch \|\| next;
272
273	my $mfn = $row->{'000'}->[0];
274
275	if (! $mfn \|\| $mfn !~ m#^\d+$#) {
276	$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
277	$mfn = $pos;
278	push @{ $row->{'000'} }, $pos;
279	}
280
281
282	if ($validate) {
283	my @errors = $validate->validate_errors( $row );
284	$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors);
285	}
286
287
288	my $ds = WebPAC::Normalize::data_structure(
289	row => $row,
290	rules => $rules,
291	lookup => $lookup ? $lookup->lookup_hash : undef,
292	marc_encoding => 'utf-8',
293	);
294
295	$db->save_ds(
296	id => $mfn,
297	ds => $ds,
298	prefix => $input->{name},
299	) if ($ds && !$stats);
300
301	$indexer->add(
302	id => $input->{name} . "/" . $mfn,
303	ds => $ds,
304	type => $config->{$use_indexer}->{type},
305	) if ($indexer && $ds);
306
307	if ($marc_fh) {
308	my $marc = new MARC::Record;
309	$marc->encoding( 'utf-8' );
310	$marc->add_fields( WebPAC::Normalize::_get_marc_fields() );
311	print $marc_fh $marc->as_usmarc;
312	}
313
314	$total_rows++;
315	}
316
317	$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats);
318
319	# close MARC file
320	close($marc_fh) if ($marc_fh);
321
322	}
323
324	}
325
326	eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
327
328	my $dt = time() - $start_t;
329	$log->info("$total_rows records ", $indexer ? "indexed " : "",
330	sprintf("in %.2f sec [%.2f rec/sec]",
331	$dt, ($total_rows / $dt)
332	)
333	);
334
335	#
336	# add Hyper Estraier links to other databases
337	#
338	if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) {
339	foreach my $link (@{ $db_config->{links} }) {
340	if ($use_indexer eq 'hyperestraier') {
341	$log->info("saving link $database -> $link->{to} [$link->{credit}]");
342	push @links, {
343	from => $database,
344	to => $link->{to},
345	credit => $link->{credit},
346	};
347	} else {
348	$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
349	}
350	}
351	}
352
353	}
354
355	foreach my $link (@links) {
356	$log->info("adding link $link->{from} -> $link->{to} [$link->{credit}]");
357	$indexer->add_link( %{ $link } );
358	}