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

reindex just single database (legacy name is --one)

=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

dump statistics about used fields and subfields in each input

=back

=cut

my $offset;
my $limit;

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

GetOptions(
        "limit=i" => \$limit,
        "offset=i" => \$offset,
        "clean" => \$clean,
        "one=s" => \$only_db_name,
        "only=s" => \$only_db_name,
        "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();

my $use_indexer = $config->{use_indexer} || 'hyperestraier';
$log->info("using $use_indexer indexing engine...");

my $total_rows = 0;
my $start_t = time();

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

        next if ($only_db_name && $database !~ m/$only_db_name/i);

        my $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) {

                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 = $n ? $n->data_structure($row) :
                                WebPAC::Normalize::Set::data_structure(
                                        row => $row,
                                        rules => $rules,
                                        lookup => $lookup->lookup_hash,
                                );

                        $indexer->add(
                                id => $input->{name} . "/" . $mfn,
                                ds => $ds,
                                type => $config->{$use_indexer}->{type},
                        );

                        $total_rows++;
                }

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

        };

        eval { $indexer->finish } if ($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	423	=item --only=database_name
46	dpavlin	335
47	dpavlin	423	reindex just single database (legacy name is --one)
48	dpavlin	335
49	dpavlin	301	=item --config conf/config.yml
50
51			path to YAML configuration file
52
53	dpavlin	492	=item --force-set
54
55			force conversion C<normalize->path> in C<config.yml> from
56			C<.xml> to C<.pl>
57
58	dpavlin	507	=item --stats
59
60			dump statistics about used fields and subfields in each input
61
62	dpavlin	301	=back
63
64			=cut
65
66			my $offset;
67			my $limit;
68
69			my $clean = 0;
70			my $config = 'conf/config.yml';
71			my $debug = 0;
72	dpavlin	423	my $only_db_name;
73	dpavlin	492	my $force_set = 0;
74	dpavlin	507	my $stats = 0;
75	dpavlin	301
76			GetOptions(
77			"limit=i" => \$limit,
78			"offset=i" => \$offset,
79			"clean" => \$clean,
80	dpavlin	423	"one=s" => \$only_db_name,
81			"only=s" => \$only_db_name,
82	dpavlin	301	"config" => \$config,
83			"debug" => \$debug,
84	dpavlin	492	"force-set" => \$force_set,
85	dpavlin	507	"stats" => \$stats,
86	dpavlin	301	);
87
88			$config = LoadFile($config);
89
90			print "config = ",Dumper($config) if ($debug);
91
92	dpavlin	210	die "no databases in config file!\n" unless ($config->{databases});
93	dpavlin	431
94			my $log = _new WebPAC::Common()->_get_logger();
95
96	dpavlin	430	my $use_indexer = $config->{use_indexer} \|\| 'hyperestraier';
97	dpavlin	431	$log->info("using $use_indexer indexing engine...");
98	dpavlin	141
99	dpavlin	213	my $total_rows = 0;
100	dpavlin	389	my $start_t = time();
101	dpavlin	213
102	dpavlin	210	while (my ($database, $db_config) = each %{ $config->{databases} }) {
103	dpavlin	141
104	dpavlin	423	next if ($only_db_name && $database !~ m/$only_db_name/i);
105	dpavlin	335
106	dpavlin	430	my $indexer;
107	dpavlin	255
108	dpavlin	431	my $indexer_config = $config->{$use_indexer} \|\| $log->logdie("can't find '$use_indexer' part in confguration");
109			$indexer_config->{database} = $database;
110			$indexer_config->{clean} = $clean;
111			$indexer_config->{label} = $db_config->{name};
112
113	dpavlin	430	if ($use_indexer eq 'hyperestraier') {
114	dpavlin	255
115	dpavlin	430	# open Hyper Estraier database
116			use WebPAC::Output::Estraier '0.10';
117	dpavlin	431	$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
118
119			} elsif ($use_indexer eq 'kinosearch') {
120	dpavlin	430
121	dpavlin	431	# open KinoSearch
122			use WebPAC::Output::KinoSearch;
123	dpavlin	434	$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path});
124	dpavlin	431	$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } );
125
126	dpavlin	430	} else {
127			$log->logdie("unknown use_indexer: $use_indexer");
128			}
129
130			$log->logide("can't continue without valid indexer") unless ($indexer);
131
132	dpavlin	255	#
133			# now WebPAC::Store
134			#
135	dpavlin	210	my $abs_path = abs_path($0);
136			$abs_path =~ s#/[^/]*$#/#;
137	dpavlin	141
138	dpavlin	210	my $db_path = $config->{webpac}->{db_path} . '/' . $database;
139	dpavlin	74
140	dpavlin	301	if ($clean) {
141			$log->info("creating new database $database in $db_path");
142			rmtree( $db_path ) \|\| $log->warn("can't remove $db_path: $!");
143			} else {
144	dpavlin	391	$log->debug("working on $database in $db_path");
145	dpavlin	301	}
146	dpavlin	255
147	dpavlin	210	my $db = new WebPAC::Store(
148			path => $db_path,
149	dpavlin	217	database => $database,
150	dpavlin	301	debug => $debug,
151	dpavlin	210	);
152	dpavlin	74
153	dpavlin	233
154	dpavlin	213	#
155			# now, iterate through input formats
156			#
157	dpavlin	74
158	dpavlin	213	my @inputs;
159			if (ref($db_config->{input}) eq 'ARRAY') {
160			@inputs = @{ $db_config->{input} };
161	dpavlin	255	} elsif ($db_config->{input}) {
162			push @inputs, $db_config->{input};
163	dpavlin	213	} else {
164	dpavlin	255	$log->info("database $database doesn't have inputs defined");
165	dpavlin	213	}
166	dpavlin	74
167	dpavlin	286	my @supported_inputs = keys %{ $config->{webpac}->{inputs} };
168
169	dpavlin	213	foreach my $input (@inputs) {
170	dpavlin	233
171			my $type = lc($input->{type});
172
173	dpavlin	286	die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs));
174	dpavlin	233
175			my $lookup = new WebPAC::Lookup(
176			lookup_file => $input->{lookup},
177			);
178
179	dpavlin	286	my $input_module = $config->{webpac}->{inputs}->{$type};
180	dpavlin	74
181	dpavlin	391	$log->info("working on input '$input->{path}' [$input->{type}] using $input_module lookup '$input->{lookup}'");
182	dpavlin	286
183	dpavlin	287	my $input_db = new WebPAC::Input(
184			module => $input_module,
185	dpavlin	213	code_page => $config->{webpac}->{webpac_encoding},
186	dpavlin	301	limit => $limit \|\| $input->{limit},
187			offset => $offset,
188	dpavlin	251	lookup => $lookup,
189	dpavlin	416	recode => $input->{recode},
190	dpavlin	507	stats => $stats,
191	dpavlin	287	);
192	dpavlin	286	$log->logdie("can't create input using $input_module") unless ($input);
193	dpavlin	113
194	dpavlin	287	my $maxmfn = $input_db->open(
195	dpavlin	285	path => $input->{path},
196	dpavlin	213	code_page => $input->{encoding}, # database encoding
197			);
198	dpavlin	113
199	dpavlin	213	my $n = new WebPAC::Normalize::XML(
200			# filter => { 'foo' => sub { shift } },
201			db => $db,
202			lookup_regex => $lookup->regex,
203			lookup => $lookup,
204	dpavlin	221	prefix => $input->{name},
205	dpavlin	213	);
206	dpavlin	113
207	dpavlin	492	my $rules;
208	dpavlin	269	my $normalize_path = $input->{normalize}->{path};
209	dpavlin	210
210	dpavlin	492	if ($force_set) {
211			my $new_norm_path = $normalize_path;
212			$new_norm_path =~ s/\.xml$/.pl/;
213			if (-e $new_norm_path) {
214	dpavlin	493	$log->debug("--force-set replaced $normalize_path with $new_norm_path");
215	dpavlin	492	$normalize_path = $new_norm_path;
216			} else {
217	dpavlin	493	$log->debug("--force-set failed on $new_norm_path, fallback to $normalize_path");
218	dpavlin	492	}
219			}
220
221	dpavlin	269	if ($normalize_path =~ m/\.xml$/i) {
222			$n->open(
223			tag => $input->{normalize}->{tag},
224	dpavlin	492	xml_file => $normalize_path,
225	dpavlin	269	);
226			} elsif ($normalize_path =~ m/\.(?:yml\|yaml)$/i) {
227			$n->open_yaml(
228			path => $normalize_path,
229			tag => $input->{normalize}->{tag},
230			);
231	dpavlin	492	} elsif ($normalize_path =~ m/\.(?:pl)$/i) {
232			$n = undef;
233			$log->info("using WebPAC::Normalize::Set to process $normalize_path");
234			$rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!";
235	dpavlin	269	}
236
237	dpavlin	290	foreach my $pos ( 0 ... $input_db->size ) {
238	dpavlin	210
239	dpavlin	287	my $row = $input_db->fetch \|\| next;
240	dpavlin	213
241	dpavlin	291	my $mfn = $row->{'000'}->[0];
242	dpavlin	213
243	dpavlin	291	if (! $mfn \|\| $mfn !~ m#^\d+$#) {
244	dpavlin	290	$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
245			$mfn = $pos;
246	dpavlin	291	push @{ $row->{'000'} }, $pos;
247	dpavlin	290	}
248
249	dpavlin	492	my $ds = $n ? $n->data_structure($row) :
250			WebPAC::Normalize::Set::data_structure(
251			row => $row,
252			rules => $rules,
253			lookup => $lookup->lookup_hash,
254			);
255	dpavlin	213
256	dpavlin	430	$indexer->add(
257	dpavlin	291	id => $input->{name} . "/" . $mfn,
258	dpavlin	213	ds => $ds,
259	dpavlin	430	type => $config->{$use_indexer}->{type},
260	dpavlin	213	);
261
262			$total_rows++;
263			}
264
265	dpavlin	507	$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats);
266
267	dpavlin	210	};
268
269	dpavlin	434	eval { $indexer->finish } if ($indexer->can('finish'));
270
271	dpavlin	389	my $dt = time() - $start_t;
272			$log->info("$total_rows records indexed in " .
273			sprintf("%.2f sec [%.2f rec/sec]",
274			$dt, ($total_rows / $dt)
275			)
276			);
277	dpavlin	255
278	dpavlin	431	#
279			# add Hyper Estraier links to other databases
280			#
281			if (ref($db_config->{links}) eq 'ARRAY') {
282			foreach my $link (@{ $db_config->{links} }) {
283			if ($use_indexer eq 'hyperestraier') {
284	dpavlin	430	$log->info("adding link $database -> $link->{to} [$link->{credit}]");
285			$indexer->add_link(
286			from => $database,
287			to => $link->{to},
288			credit => $link->{credit},
289			);
290	dpavlin	431	} else {
291			$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
292	dpavlin	430	}
293	dpavlin	255	}
294			}
295
296	dpavlin	210	}
297