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::XML;
use WebPAC::Normalize::Set;
use WebPAC::Output::TT;
use YAML qw/LoadFile/;
use Getopt::Long;
use File::Path;
use Time::HiRes qw/time/;
use File::Slurp;

=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 --force-set

force conversion C<< normalize->path >> in C<config.yml> from
C<.xml> to C<.pl>

=item --stats

disable indexing and dump statistics about field and subfield
usage for each input

=back

=cut

my $offset;
my $limit;

my $clean = 0;
my $config = 'conf/config.yml';
my $debug = 0;
my $only_filter;
my $force_set = 0;
my $stats = 0;

GetOptions(
        "limit=i" => \$limit,
        "offset=i" => \$offset,
        "clean" => \$clean,
        "one=s" => \$only_filter,
        "only=s" => \$only_filter,
        "config" => \$config,
        "debug" => \$debug,
        "force-set" => \$force_set,
        "stats" => \$stats,
);

$config = LoadFile($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 $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();

while (my ($database, $db_config) = each %{ $config->{databases} }) {

        my ($only_database,$only_input) = split(m#/#, $only_filter);
        next if ($only_database && $database !~ m/$only_database/i);

        my $indexer;

        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->debug("working on $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},
                );

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

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

                my $n = new WebPAC::Normalize::XML(
                #       filter => { 'foo' => sub { shift } },
                        db => $db,
                        lookup_regex => $lookup->regex,
                        lookup => $lookup,
                        prefix => $input->{name},
                );

                my $rules;
                my $normalize_path = $input->{normalize}->{path};

                if ($force_set) {
                        my $new_norm_path = $normalize_path;
                        $new_norm_path =~ s/\.xml$/.pl/;
                        if (-e $new_norm_path) {
                                $log->debug("--force-set replaced $normalize_path with $new_norm_path");
                                $normalize_path = $new_norm_path;
                        } else {
                                $log->debug("--force-set failed on $new_norm_path, fallback to $normalize_path");
                        }
                }

                if ($normalize_path =~ m/\.xml$/i) {
                        $n->open(
                                tag => $input->{normalize}->{tag},
                                xml_file => $normalize_path,
                        );
                } elsif ($normalize_path =~ m/\.(?:yml|yaml)$/i) {
                        $n->open_yaml(
                                path => $normalize_path,
                                tag => $input->{normalize}->{tag},
                        );
                } elsif ($normalize_path =~ m/\.(?:pl)$/i) {
                        $n = undef;
                        $log->info("using WebPAC::Normalize::Set to process $normalize_path");
                        $rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!";
                }

                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;
                        }

                                
                        my $ds;
                        if ($n) {
                                $ds = $n->data_structure($row);
                        } else {
                                $ds = WebPAC::Normalize::Set::data_structure(
                                        row => $row,
                                        rules => $rules,
                                        lookup => $lookup->lookup_hash,
                                );

                                $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);

                        $total_rows++;
                }

                $log->info("statistics of fields usage:\n", $input_db->stats) if ($stats);

        };

        eval { $indexer->finish } if ($indexer && $indexer->can('finish'));

        my $dt = time() - $start_t;
        $log->info("$total_rows records indexed in " .
                sprintf("%.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("adding link $database -> $link->{to} [$link->{credit}]");
                                $indexer->add_link(
                                        from => $database,
                                        to => $link->{to},
                                        credit => $link->{credit},
                                );
                        } else {
                                $log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
                        }
                }
        }

}

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	74	use WebPAC::Normalize::XML;
15	dpavlin	492	use WebPAC::Normalize::Set;
16	dpavlin	74	use WebPAC::Output::TT;
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	74
23	dpavlin	301	=head1 NAME
24	dpavlin	76
25	dpavlin	301	run.pl - start WebPAC indexing
26	dpavlin	141
27	dpavlin	301	B<this command will probably go away. Don't get used to it!>
28	dpavlin	141
29	dpavlin	301	Options:
30
31			=over 4
32
33			=item --offset 42
34
35			start loading (all) databases at offset 42
36
37			=item --limit 100
38
39			limit loading to 100 records
40
41			=item --clean
42
43			remove database and Hyper Estraier index before indexing
44
45	dpavlin	510	=item --only=database_name/input_filter
46	dpavlin	335
47	dpavlin	423	reindex just single database (legacy name is --one)
48	dpavlin	335
49	dpavlin	510	C</input_filter> is optional part which can be C<name>
50			or C<type> from input
51
52	dpavlin	301	=item --config conf/config.yml
53
54			path to YAML configuration file
55
56	dpavlin	492	=item --force-set
57
58	dpavlin	508	force conversion C<< normalize->path >> in C<config.yml> from
59	dpavlin	492	C<.xml> to C<.pl>
60
61	dpavlin	507	=item --stats
62
63	dpavlin	509	disable indexing and dump statistics about field and subfield
64			usage for each input
65	dpavlin	507
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	492	my $force_set = 0;
78	dpavlin	507	my $stats = 0;
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	492	"force-set" => \$force_set,
89	dpavlin	507	"stats" => \$stats,
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	430	my $use_indexer = $config->{use_indexer} \|\| 'hyperestraier';
102	dpavlin	509	if ($stats) {
103			$log->debug("option --stats disables update of indexing engine...");
104			$use_indexer = undef;
105			} else {
106			$log->info("using $use_indexer indexing engine...");
107			}
108	dpavlin	141
109	dpavlin	213	my $total_rows = 0;
110	dpavlin	389	my $start_t = time();
111	dpavlin	213
112	dpavlin	210	while (my ($database, $db_config) = each %{ $config->{databases} }) {
113	dpavlin	141
114	dpavlin	510	my ($only_database,$only_input) = split(m#/#, $only_filter);
115			next if ($only_database && $database !~ m/$only_database/i);
116	dpavlin	335
117	dpavlin	430	my $indexer;
118	dpavlin	255
119	dpavlin	509	if ($use_indexer) {
120			my $indexer_config = $config->{$use_indexer} \|\| $log->logdie("can't find '$use_indexer' part in confguration");
121			$indexer_config->{database} = $database;
122			$indexer_config->{clean} = $clean;
123			$indexer_config->{label} = $db_config->{name};
124	dpavlin	431
125	dpavlin	509	if ($use_indexer eq 'hyperestraier') {
126	dpavlin	255
127	dpavlin	509	# open Hyper Estraier database
128			use WebPAC::Output::Estraier '0.10';
129			$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
130
131			} elsif ($use_indexer eq 'kinosearch') {
132	dpavlin	430
133	dpavlin	509	# open KinoSearch
134			use WebPAC::Output::KinoSearch;
135			$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path});
136			$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } );
137	dpavlin	431
138	dpavlin	509	} else {
139			$log->logdie("unknown use_indexer: $use_indexer");
140			}
141
142			$log->logide("can't continue without valid indexer") unless ($indexer);
143	dpavlin	430	}
144
145
146	dpavlin	255	#
147			# now WebPAC::Store
148			#
149	dpavlin	210	my $abs_path = abs_path($0);
150			$abs_path =~ s#/[^/]*$#/#;
151	dpavlin	141
152	dpavlin	210	my $db_path = $config->{webpac}->{db_path} . '/' . $database;
153	dpavlin	74
154	dpavlin	301	if ($clean) {
155			$log->info("creating new database $database in $db_path");
156			rmtree( $db_path ) \|\| $log->warn("can't remove $db_path: $!");
157			} else {
158	dpavlin	391	$log->debug("working on $database in $db_path");
159	dpavlin	301	}
160	dpavlin	255
161	dpavlin	210	my $db = new WebPAC::Store(
162			path => $db_path,
163	dpavlin	217	database => $database,
164	dpavlin	301	debug => $debug,
165	dpavlin	210	);
166	dpavlin	74
167	dpavlin	233
168	dpavlin	213	#
169			# now, iterate through input formats
170			#
171	dpavlin	74
172	dpavlin	213	my @inputs;
173			if (ref($db_config->{input}) eq 'ARRAY') {
174			@inputs = @{ $db_config->{input} };
175	dpavlin	255	} elsif ($db_config->{input}) {
176			push @inputs, $db_config->{input};
177	dpavlin	213	} else {
178	dpavlin	255	$log->info("database $database doesn't have inputs defined");
179	dpavlin	213	}
180	dpavlin	74
181	dpavlin	286	my @supported_inputs = keys %{ $config->{webpac}->{inputs} };
182
183	dpavlin	213	foreach my $input (@inputs) {
184	dpavlin	233
185	dpavlin	510	next if ($only_input && $input->{name} =~ m#$only_input#i \|\| $input->{type} =~ m#$only_input#i);
186
187	dpavlin	233	my $type = lc($input->{type});
188
189	dpavlin	286	die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs));
190	dpavlin	233
191			my $lookup = new WebPAC::Lookup(
192			lookup_file => $input->{lookup},
193			);
194
195	dpavlin	286	my $input_module = $config->{webpac}->{inputs}->{$type};
196	dpavlin	74
197	dpavlin	391	$log->info("working on input '$input->{path}' [$input->{type}] using $input_module lookup '$input->{lookup}'");
198	dpavlin	286
199	dpavlin	287	my $input_db = new WebPAC::Input(
200			module => $input_module,
201	dpavlin	213	code_page => $config->{webpac}->{webpac_encoding},
202	dpavlin	301	limit => $limit \|\| $input->{limit},
203			offset => $offset,
204	dpavlin	251	lookup => $lookup,
205	dpavlin	416	recode => $input->{recode},
206	dpavlin	507	stats => $stats,
207	dpavlin	287	);
208	dpavlin	286	$log->logdie("can't create input using $input_module") unless ($input);
209	dpavlin	113
210	dpavlin	287	my $maxmfn = $input_db->open(
211	dpavlin	285	path => $input->{path},
212	dpavlin	213	code_page => $input->{encoding}, # database encoding
213			);
214	dpavlin	113
215	dpavlin	213	my $n = new WebPAC::Normalize::XML(
216			# filter => { 'foo' => sub { shift } },
217			db => $db,
218			lookup_regex => $lookup->regex,
219			lookup => $lookup,
220	dpavlin	221	prefix => $input->{name},
221	dpavlin	213	);
222	dpavlin	113
223	dpavlin	492	my $rules;
224	dpavlin	269	my $normalize_path = $input->{normalize}->{path};
225	dpavlin	210
226	dpavlin	492	if ($force_set) {
227			my $new_norm_path = $normalize_path;
228			$new_norm_path =~ s/\.xml$/.pl/;
229			if (-e $new_norm_path) {
230	dpavlin	493	$log->debug("--force-set replaced $normalize_path with $new_norm_path");
231	dpavlin	492	$normalize_path = $new_norm_path;
232			} else {
233	dpavlin	493	$log->debug("--force-set failed on $new_norm_path, fallback to $normalize_path");
234	dpavlin	492	}
235			}
236
237	dpavlin	269	if ($normalize_path =~ m/\.xml$/i) {
238			$n->open(
239			tag => $input->{normalize}->{tag},
240	dpavlin	492	xml_file => $normalize_path,
241	dpavlin	269	);
242			} elsif ($normalize_path =~ m/\.(?:yml\|yaml)$/i) {
243			$n->open_yaml(
244			path => $normalize_path,
245			tag => $input->{normalize}->{tag},
246			);
247	dpavlin	492	} elsif ($normalize_path =~ m/\.(?:pl)$/i) {
248			$n = undef;
249			$log->info("using WebPAC::Normalize::Set to process $normalize_path");
250			$rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!";
251	dpavlin	269	}
252
253	dpavlin	290	foreach my $pos ( 0 ... $input_db->size ) {
254	dpavlin	210
255	dpavlin	287	my $row = $input_db->fetch \|\| next;
256	dpavlin	213
257	dpavlin	291	my $mfn = $row->{'000'}->[0];
258	dpavlin	213
259	dpavlin	291	if (! $mfn \|\| $mfn !~ m#^\d+$#) {
260	dpavlin	290	$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
261			$mfn = $pos;
262	dpavlin	291	push @{ $row->{'000'} }, $pos;
263	dpavlin	290	}
264
265	dpavlin	508
266			my $ds;
267			if ($n) {
268			$ds = $n->data_structure($row);
269			} else {
270			$ds = WebPAC::Normalize::Set::data_structure(
271	dpavlin	492	row => $row,
272			rules => $rules,
273			lookup => $lookup->lookup_hash,
274			);
275	dpavlin	213
276	dpavlin	508	$db->save_ds(
277			id => $mfn,
278			ds => $ds,
279			prefix => $input->{name},
280	dpavlin	509	) if ($ds && !$stats);
281	dpavlin	508	}
282
283	dpavlin	430	$indexer->add(
284	dpavlin	291	id => $input->{name} . "/" . $mfn,
285	dpavlin	213	ds => $ds,
286	dpavlin	430	type => $config->{$use_indexer}->{type},
287	dpavlin	509	) if ($indexer);
288	dpavlin	213
289			$total_rows++;
290			}
291
292	dpavlin	507	$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats);
293
294	dpavlin	210	};
295
296	dpavlin	509	eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
297	dpavlin	434
298	dpavlin	389	my $dt = time() - $start_t;
299			$log->info("$total_rows records indexed in " .
300			sprintf("%.2f sec [%.2f rec/sec]",
301			$dt, ($total_rows / $dt)
302			)
303			);
304	dpavlin	255
305	dpavlin	431	#
306			# add Hyper Estraier links to other databases
307			#
308			if (ref($db_config->{links}) eq 'ARRAY') {
309			foreach my $link (@{ $db_config->{links} }) {
310			if ($use_indexer eq 'hyperestraier') {
311	dpavlin	430	$log->info("adding link $database -> $link->{to} [$link->{credit}]");
312			$indexer->add_link(
313			from => $database,
314			to => $link->{to},
315			credit => $link->{credit},
316			);
317	dpavlin	431	} else {
318			$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
319	dpavlin	430	}
320	dpavlin	255	}
321			}
322
323	dpavlin	210	}
324