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 WebPAC::Validate;
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

=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 $force_set = 0;
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,
        "force-set" => \$force_set,
        "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 $n = new WebPAC::Normalize::XML(
                #       filter => { 'foo' => sub { shift } },
                        db => $db,
                        lookup_regex => $lookup ? $lookup->regex : undef,
                        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;
                        }


                        if ($validate) {
                                my @errors = $validate->validate_errors( $row );
                                $log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors);
                        }

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

                        $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 ", $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	74	use WebPAC::Normalize::XML;
15	dpavlin	492	use WebPAC::Normalize::Set;
16	dpavlin	74	use WebPAC::Output::TT;
17	dpavlin	516	use WebPAC::Validate;
18	dpavlin	141	use YAML qw/LoadFile/;
19	dpavlin	301	use Getopt::Long;
20			use File::Path;
21	dpavlin	389	use Time::HiRes qw/time/;
22	dpavlin	492	use File::Slurp;
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	492	=item --force-set
58
59	dpavlin	508	force conversion C<< normalize->path >> in C<config.yml> from
60	dpavlin	492	C<.xml> to C<.pl>
61
62	dpavlin	507	=item --stats
63
64	dpavlin	509	disable indexing and dump statistics about field and subfield
65			usage for each input
66	dpavlin	507
67	dpavlin	516	=item --validate path/to/validation_file
68
69			turn on extra validation of imput records, see L<WebPAC::Validation>
70
71	dpavlin	301	=back
72
73			=cut
74
75			my $offset;
76			my $limit;
77
78			my $clean = 0;
79			my $config = 'conf/config.yml';
80			my $debug = 0;
81	dpavlin	510	my $only_filter;
82	dpavlin	492	my $force_set = 0;
83	dpavlin	507	my $stats = 0;
84	dpavlin	516	my $validate_path;
85	dpavlin	301
86			GetOptions(
87			"limit=i" => \$limit,
88			"offset=i" => \$offset,
89			"clean" => \$clean,
90	dpavlin	510	"one=s" => \$only_filter,
91			"only=s" => \$only_filter,
92	dpavlin	301	"config" => \$config,
93			"debug" => \$debug,
94	dpavlin	492	"force-set" => \$force_set,
95	dpavlin	507	"stats" => \$stats,
96	dpavlin	516	"validate=s" => \$validate_path,
97	dpavlin	301	);
98
99			$config = LoadFile($config);
100
101			print "config = ",Dumper($config) if ($debug);
102
103	dpavlin	210	die "no databases in config file!\n" unless ($config->{databases});
104	dpavlin	431
105			my $log = _new WebPAC::Common()->_get_logger();
106	dpavlin	509	$log->info( "-" x 79 );
107	dpavlin	431
108	dpavlin	516	my $validate;
109			$validate = new WebPAC::Validate(
110			path => $validate_path,
111			) if ($validate_path);
112
113	dpavlin	430	my $use_indexer = $config->{use_indexer} \|\| 'hyperestraier';
114	dpavlin	509	if ($stats) {
115			$log->debug("option --stats disables update of indexing engine...");
116			$use_indexer = undef;
117			} else {
118			$log->info("using $use_indexer indexing engine...");
119			}
120	dpavlin	141
121	dpavlin	213	my $total_rows = 0;
122	dpavlin	389	my $start_t = time();
123	dpavlin	213
124	dpavlin	529	my @links;
125			my $indexer;
126
127	dpavlin	210	while (my ($database, $db_config) = each %{ $config->{databases} }) {
128	dpavlin	141
129	dpavlin	512	my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter);
130	dpavlin	510	next if ($only_database && $database !~ m/$only_database/i);
131	dpavlin	335
132	dpavlin	509	if ($use_indexer) {
133			my $indexer_config = $config->{$use_indexer} \|\| $log->logdie("can't find '$use_indexer' part in confguration");
134			$indexer_config->{database} = $database;
135			$indexer_config->{clean} = $clean;
136			$indexer_config->{label} = $db_config->{name};
137	dpavlin	431
138	dpavlin	509	if ($use_indexer eq 'hyperestraier') {
139	dpavlin	255
140	dpavlin	509	# open Hyper Estraier database
141			use WebPAC::Output::Estraier '0.10';
142			$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
143
144			} elsif ($use_indexer eq 'kinosearch') {
145	dpavlin	430
146	dpavlin	509	# open KinoSearch
147			use WebPAC::Output::KinoSearch;
148			$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path});
149			$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } );
150	dpavlin	431
151	dpavlin	509	} else {
152			$log->logdie("unknown use_indexer: $use_indexer");
153			}
154
155			$log->logide("can't continue without valid indexer") unless ($indexer);
156	dpavlin	430	}
157
158
159	dpavlin	255	#
160			# now WebPAC::Store
161			#
162	dpavlin	210	my $abs_path = abs_path($0);
163			$abs_path =~ s#/[^/]*$#/#;
164	dpavlin	141
165	dpavlin	210	my $db_path = $config->{webpac}->{db_path} . '/' . $database;
166	dpavlin	74
167	dpavlin	301	if ($clean) {
168	dpavlin	511	$log->info("creating new database '$database' in $db_path");
169	dpavlin	301	rmtree( $db_path ) \|\| $log->warn("can't remove $db_path: $!");
170			} else {
171	dpavlin	511	$log->info("working on database '$database' in $db_path");
172	dpavlin	301	}
173	dpavlin	255
174	dpavlin	210	my $db = new WebPAC::Store(
175			path => $db_path,
176	dpavlin	217	database => $database,
177	dpavlin	301	debug => $debug,
178	dpavlin	210	);
179	dpavlin	74
180	dpavlin	233
181	dpavlin	213	#
182			# now, iterate through input formats
183			#
184	dpavlin	74
185	dpavlin	213	my @inputs;
186			if (ref($db_config->{input}) eq 'ARRAY') {
187			@inputs = @{ $db_config->{input} };
188	dpavlin	255	} elsif ($db_config->{input}) {
189			push @inputs, $db_config->{input};
190	dpavlin	213	} else {
191	dpavlin	255	$log->info("database $database doesn't have inputs defined");
192	dpavlin	213	}
193	dpavlin	74
194	dpavlin	286	my @supported_inputs = keys %{ $config->{webpac}->{inputs} };
195
196	dpavlin	213	foreach my $input (@inputs) {
197	dpavlin	233
198	dpavlin	512	next if ($only_input && ($input->{name} !~ m#$only_input#i && $input->{type} !~ m#$only_input#i));
199	dpavlin	510
200	dpavlin	233	my $type = lc($input->{type});
201
202	dpavlin	286	die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs));
203	dpavlin	233
204			my $lookup = new WebPAC::Lookup(
205			lookup_file => $input->{lookup},
206	dpavlin	520	) if ($input->{lookup});
207	dpavlin	233
208	dpavlin	286	my $input_module = $config->{webpac}->{inputs}->{$type};
209	dpavlin	74
210	dpavlin	523	$log->info("working on input '$input->{name}' in $input->{path} [type: $input->{type}] using $input_module",
211			$input->{lookup} ? "lookup '$input->{lookup}'" : ""
212			);
213	dpavlin	286
214	dpavlin	287	my $input_db = new WebPAC::Input(
215			module => $input_module,
216	dpavlin	213	code_page => $config->{webpac}->{webpac_encoding},
217	dpavlin	301	limit => $limit \|\| $input->{limit},
218			offset => $offset,
219	dpavlin	251	lookup => $lookup,
220	dpavlin	416	recode => $input->{recode},
221	dpavlin	507	stats => $stats,
222	dpavlin	287	);
223	dpavlin	286	$log->logdie("can't create input using $input_module") unless ($input);
224	dpavlin	113
225	dpavlin	287	my $maxmfn = $input_db->open(
226	dpavlin	285	path => $input->{path},
227	dpavlin	213	code_page => $input->{encoding}, # database encoding
228	dpavlin	523	%{ $input },
229	dpavlin	213	);
230	dpavlin	113
231	dpavlin	213	my $n = new WebPAC::Normalize::XML(
232			# filter => { 'foo' => sub { shift } },
233			db => $db,
234	dpavlin	520	lookup_regex => $lookup ? $lookup->regex : undef,
235	dpavlin	213	lookup => $lookup,
236	dpavlin	221	prefix => $input->{name},
237	dpavlin	213	);
238	dpavlin	113
239	dpavlin	492	my $rules;
240	dpavlin	269	my $normalize_path = $input->{normalize}->{path};
241	dpavlin	210
242	dpavlin	492	if ($force_set) {
243			my $new_norm_path = $normalize_path;
244			$new_norm_path =~ s/\.xml$/.pl/;
245			if (-e $new_norm_path) {
246	dpavlin	493	$log->debug("--force-set replaced $normalize_path with $new_norm_path");
247	dpavlin	492	$normalize_path = $new_norm_path;
248			} else {
249	dpavlin	493	$log->debug("--force-set failed on $new_norm_path, fallback to $normalize_path");
250	dpavlin	492	}
251			}
252
253	dpavlin	269	if ($normalize_path =~ m/\.xml$/i) {
254			$n->open(
255			tag => $input->{normalize}->{tag},
256	dpavlin	492	xml_file => $normalize_path,
257	dpavlin	269	);
258			} elsif ($normalize_path =~ m/\.(?:yml\|yaml)$/i) {
259			$n->open_yaml(
260			path => $normalize_path,
261			tag => $input->{normalize}->{tag},
262			);
263	dpavlin	492	} elsif ($normalize_path =~ m/\.(?:pl)$/i) {
264			$n = undef;
265			$log->info("using WebPAC::Normalize::Set to process $normalize_path");
266			$rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!";
267	dpavlin	269	}
268
269	dpavlin	290	foreach my $pos ( 0 ... $input_db->size ) {
270	dpavlin	210
271	dpavlin	287	my $row = $input_db->fetch \|\| next;
272	dpavlin	213
273	dpavlin	291	my $mfn = $row->{'000'}->[0];
274	dpavlin	213
275	dpavlin	291	if (! $mfn \|\| $mfn !~ m#^\d+$#) {
276	dpavlin	290	$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
277			$mfn = $pos;
278	dpavlin	291	push @{ $row->{'000'} }, $pos;
279	dpavlin	290	}
280
281	dpavlin	516
282			if ($validate) {
283			my @errors = $validate->validate_errors( $row );
284			$log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors);
285			}
286
287	dpavlin	508
288			my $ds;
289			if ($n) {
290			$ds = $n->data_structure($row);
291			} else {
292			$ds = WebPAC::Normalize::Set::data_structure(
293	dpavlin	492	row => $row,
294			rules => $rules,
295	dpavlin	520	lookup => $lookup ? $lookup->lookup_hash : undef,
296	dpavlin	492	);
297	dpavlin	213
298	dpavlin	508	$db->save_ds(
299			id => $mfn,
300			ds => $ds,
301			prefix => $input->{name},
302	dpavlin	509	) if ($ds && !$stats);
303	dpavlin	508	}
304
305	dpavlin	430	$indexer->add(
306	dpavlin	291	id => $input->{name} . "/" . $mfn,
307	dpavlin	213	ds => $ds,
308	dpavlin	430	type => $config->{$use_indexer}->{type},
309	dpavlin	509	) if ($indexer);
310	dpavlin	213
311			$total_rows++;
312			}
313
314	dpavlin	507	$log->info("statistics of fields usage:\n", $input_db->stats) if ($stats);
315
316	dpavlin	210	};
317
318	dpavlin	509	eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
319	dpavlin	434
320	dpavlin	389	my $dt = time() - $start_t;
321	dpavlin	511	$log->info("$total_rows records ", $indexer ? "indexed " : "",
322			sprintf("in %.2f sec [%.2f rec/sec]",
323	dpavlin	389	$dt, ($total_rows / $dt)
324			)
325			);
326	dpavlin	255
327	dpavlin	431	#
328			# add Hyper Estraier links to other databases
329			#
330			if (ref($db_config->{links}) eq 'ARRAY') {
331			foreach my $link (@{ $db_config->{links} }) {
332			if ($use_indexer eq 'hyperestraier') {
333	dpavlin	529	$log->info("saving link $database -> $link->{to} [$link->{credit}]");
334			push @links, {
335	dpavlin	430	from => $database,
336			to => $link->{to},
337			credit => $link->{credit},
338	dpavlin	529	};
339	dpavlin	431	} else {
340			$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
341	dpavlin	430	}
342	dpavlin	255	}
343			}
344
345	dpavlin	210	}
346
347	dpavlin	529	foreach my $link (@links) {
348			$log->info("adding link $link->{from} -> $link->{to} [$link->{credit}]");
349			$indexer->add_link( %{ $link } );
350			}