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;

=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 $rules;
                        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,
                                );

                                $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->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	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 Data::Dumper;
8			use lib './lib';
9
10	dpavlin	255	use WebPAC::Common 0.02;
11	dpavlin	74	use WebPAC::Lookup;
12	dpavlin	287	use WebPAC::Input 0.03;
13	dpavlin	209	use WebPAC::Store 0.03;
14	dpavlin	536	use WebPAC::Normalize;
15	dpavlin	74	use WebPAC::Output::TT;
16	dpavlin	516	use WebPAC::Validate;
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	540	use MARC::Record;
23	dpavlin	74
24	dpavlin	301	=head1 NAME
25	dpavlin	76
26	dpavlin	301	run.pl - start WebPAC indexing
27	dpavlin	141
28	dpavlin	301	B<this command will probably go away. Don't get used to it!>
29	dpavlin	141
30	dpavlin	301	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	dpavlin	510	=item --only=database_name/input_filter
47	dpavlin	335
48	dpavlin	423	reindex just single database (legacy name is --one)
49	dpavlin	335
50	dpavlin	510	C</input_filter> is optional part which can be C<name>
51			or C<type> from input
52
53	dpavlin	301	=item --config conf/config.yml
54
55			path to YAML configuration file
56
57	dpavlin	507	=item --stats
58
59	dpavlin	509	disable indexing and dump statistics about field and subfield
60			usage for each input
61	dpavlin	507
62	dpavlin	516	=item --validate path/to/validation_file
63
64			turn on extra validation of imput records, see L<WebPAC::Validation>
65
66	dpavlin	301	=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	dpavlin	510	my $only_filter;
77	dpavlin	507	my $stats = 0;
78	dpavlin	516	my $validate_path;
79	dpavlin	301
80			GetOptions(
81			"limit=i" => \$limit,
82			"offset=i" => \$offset,
83			"clean" => \$clean,
84	dpavlin	510	"one=s" => \$only_filter,
85			"only=s" => \$only_filter,
86	dpavlin	301	"config" => \$config,
87			"debug" => \$debug,
88	dpavlin	507	"stats" => \$stats,
89	dpavlin	516	"validate=s" => \$validate_path,
90	dpavlin	301	);
91
92			$config = LoadFile($config);
93
94			print "config = ",Dumper($config) if ($debug);
95
96	dpavlin	210	die "no databases in config file!\n" unless ($config->{databases});
97	dpavlin	431
98			my $log = _new WebPAC::Common()->_get_logger();
99	dpavlin	509	$log->info( "-" x 79 );
100	dpavlin	431
101	dpavlin	516	my $validate;
102			$validate = new WebPAC::Validate(
103			path => $validate_path,
104			) if ($validate_path);
105
106	dpavlin	430	my $use_indexer = $config->{use_indexer} \|\| 'hyperestraier';
107	dpavlin	509	if ($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	dpavlin	141
114	dpavlin	213	my $total_rows = 0;
115	dpavlin	389	my $start_t = time();
116	dpavlin	213
117	dpavlin	529	my @links;
118			my $indexer;
119
120	dpavlin	210	while (my ($database, $db_config) = each %{ $config->{databases} }) {
121	dpavlin	141
122	dpavlin	512	my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter);
123	dpavlin	510	next if ($only_database && $database !~ m/$only_database/i);
124	dpavlin	335
125	dpavlin	509	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} = $db_config->{name};
130	dpavlin	431
131	dpavlin	509	if ($use_indexer eq 'hyperestraier') {
132	dpavlin	255
133	dpavlin	509	# 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	dpavlin	430
139	dpavlin	509	# 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	dpavlin	431
144	dpavlin	509	} else {
145			$log->logdie("unknown use_indexer: $use_indexer");
146			}
147
148			$log->logide("can't continue without valid indexer") unless ($indexer);
149	dpavlin	430	}
150
151
152	dpavlin	255	#
153			# now WebPAC::Store
154			#
155	dpavlin	210	my $abs_path = abs_path($0);
156			$abs_path =~ s#/[^/]*$#/#;
157	dpavlin	141
158	dpavlin	210	my $db_path = $config->{webpac}->{db_path} . '/' . $database;
159	dpavlin	74
160	dpavlin	301	if ($clean) {
161	dpavlin	511	$log->info("creating new database '$database' in $db_path");
162	dpavlin	301	rmtree( $db_path ) \|\| $log->warn("can't remove $db_path: $!");
163			} else {
164	dpavlin	511	$log->info("working on database '$database' in $db_path");
165	dpavlin	301	}
166	dpavlin	255
167	dpavlin	210	my $db = new WebPAC::Store(
168			path => $db_path,
169	dpavlin	217	database => $database,
170	dpavlin	301	debug => $debug,
171	dpavlin	210	);
172	dpavlin	74
173	dpavlin	233
174	dpavlin	213	#
175			# now, iterate through input formats
176			#
177	dpavlin	74
178	dpavlin	213	my @inputs;
179			if (ref($db_config->{input}) eq 'ARRAY') {
180			@inputs = @{ $db_config->{input} };
181	dpavlin	255	} elsif ($db_config->{input}) {
182			push @inputs, $db_config->{input};
183	dpavlin	213	} else {
184	dpavlin	255	$log->info("database $database doesn't have inputs defined");
185	dpavlin	213	}
186	dpavlin	74
187	dpavlin	286	my @supported_inputs = keys %{ $config->{webpac}->{inputs} };
188
189	dpavlin	213	foreach my $input (@inputs) {
190	dpavlin	233
191	dpavlin	512	next if ($only_input && ($input->{name} !~ m#$only_input#i && $input->{type} !~ m#$only_input#i));
192	dpavlin	510
193	dpavlin	233	my $type = lc($input->{type});
194
195	dpavlin	286	die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs));
196	dpavlin	233
197			my $lookup = new WebPAC::Lookup(
198			lookup_file => $input->{lookup},
199	dpavlin	520	) if ($input->{lookup});
200	dpavlin	233
201	dpavlin	286	my $input_module = $config->{webpac}->{inputs}->{$type};
202	dpavlin	74
203	dpavlin	523	$log->info("working on input '$input->{name}' in $input->{path} [type: $input->{type}] using $input_module",
204			$input->{lookup} ? "lookup '$input->{lookup}'" : ""
205			);
206	dpavlin	286
207	dpavlin	287	my $input_db = new WebPAC::Input(
208			module => $input_module,
209	dpavlin	213	code_page => $config->{webpac}->{webpac_encoding},
210	dpavlin	301	limit => $limit \|\| $input->{limit},
211			offset => $offset,
212	dpavlin	251	lookup => $lookup,
213	dpavlin	416	recode => $input->{recode},
214	dpavlin	507	stats => $stats,
215	dpavlin	287	);
216	dpavlin	286	$log->logdie("can't create input using $input_module") unless ($input);
217	dpavlin	113
218	dpavlin	287	my $maxmfn = $input_db->open(
219	dpavlin	285	path => $input->{path},
220	dpavlin	213	code_page => $input->{encoding}, # database encoding
221	dpavlin	523	%{ $input },
222	dpavlin	213	);
223	dpavlin	113
224	dpavlin	539	my @norm_array = ref($input->{normalize}) eq 'ARRAY' ?
225			@{ $input->{normalize} } : ( $input->{normalize} );
226	dpavlin	210
227	dpavlin	539	foreach my $normalize (@norm_array) {
228	dpavlin	492
229	dpavlin	539	my $rules;
230			my $normalize_path = $normalize->{path} \|\| $log->logdie("can't find normalize path in config");
231	dpavlin	269
232	dpavlin	539	$log->logdie("Found '$normalize_path' as normalization file which isn't supported any more!") unless ( $normalize_path =~ m!\.pl$!i );
233	dpavlin	210
234	dpavlin	539	my $rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!";
235	dpavlin	213
236	dpavlin	539	$log->info("Using $normalize_path for normalization...");
237	dpavlin	213
238	dpavlin	540	my $marc_fh;
239			if (my $path = $normalize->{marc21}) {
240			open($marc_fh, '>', $path) \|\|
241			$log->logdie("can't open MARC output $path: $!");
242
243			$log->info("Creating MARC export file $path\n");
244			}
245
246	dpavlin	539	# reset position in database
247			$input_db->seek(1);
248	dpavlin	290
249	dpavlin	539	foreach my $pos ( 0 ... $input_db->size ) {
250	dpavlin	516
251	dpavlin	539	my $row = $input_db->fetch \|\| next;
252	dpavlin	516
253	dpavlin	539	my $mfn = $row->{'000'}->[0];
254	dpavlin	213
255	dpavlin	539	if (! $mfn \|\| $mfn !~ m#^\d+$#) {
256			$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
257			$mfn = $pos;
258			push @{ $row->{'000'} }, $pos;
259			}
260	dpavlin	508
261	dpavlin	213
262	dpavlin	539	if ($validate) {
263			my @errors = $validate->validate_errors( $row );
264			$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors);
265			}
266	dpavlin	213
267	dpavlin	539
268			my $ds = WebPAC::Normalize::data_structure(
269			row => $row,
270			rules => $rules,
271			lookup => $lookup ? $lookup->lookup_hash : undef,
272			);
273	dpavlin	507
274	dpavlin	539	$db->save_ds(
275			id => $mfn,
276			ds => $ds,
277			prefix => $input->{name},
278			) if ($ds && !$stats);
279	dpavlin	210
280	dpavlin	539	$indexer->add(
281			id => $input->{name} . "/" . $mfn,
282			ds => $ds,
283			type => $config->{$use_indexer}->{type},
284	dpavlin	540	) if ($indexer && $ds);
285	dpavlin	539
286	dpavlin	540	if ($marc_fh) {
287			my $marc = new MARC::Record;
288			$marc->add_fields( WebPAC::Normalize::_get_marc21_fields() );
289			print $marc_fh $marc->as_usmarc;
290			}
291
292	dpavlin	539	$total_rows++;
293			}
294
295			$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats);
296
297	dpavlin	540	# close MARC file
298			close($marc_fh) if ($marc_fh);
299	dpavlin	539
300	dpavlin	540	}
301
302	dpavlin	539	}
303
304	dpavlin	509	eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
305	dpavlin	434
306	dpavlin	389	my $dt = time() - $start_t;
307	dpavlin	511	$log->info("$total_rows records ", $indexer ? "indexed " : "",
308			sprintf("in %.2f sec [%.2f rec/sec]",
309	dpavlin	389	$dt, ($total_rows / $dt)
310			)
311			);
312	dpavlin	255
313	dpavlin	431	#
314			# add Hyper Estraier links to other databases
315			#
316			if (ref($db_config->{links}) eq 'ARRAY') {
317			foreach my $link (@{ $db_config->{links} }) {
318			if ($use_indexer eq 'hyperestraier') {
319	dpavlin	529	$log->info("saving link $database -> $link->{to} [$link->{credit}]");
320			push @links, {
321	dpavlin	430	from => $database,
322			to => $link->{to},
323			credit => $link->{credit},
324	dpavlin	529	};
325	dpavlin	431	} else {
326			$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
327	dpavlin	430	}
328	dpavlin	255	}
329			}
330
331	dpavlin	210	}
332
333	dpavlin	529	foreach my $link (@links) {
334			$log->info("adding link $link->{from} -> $link->{to} [$link->{credit}]");
335			$indexer->add_link( %{ $link } );
336			}