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::Output::TT;
use YAML qw/LoadFile/;
use Getopt::Long;
use File::Path;
use Time::HiRes qw/time/;

=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

=back

=cut

my $offset;
my $limit;

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

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

$config = LoadFile($config);

print "config = ",Dumper($config) if ($debug);

die "no databases in config file!\n" unless ($config->{databases});
my $use_indexer = $config->{use_indexer} || 'hyperestraier';

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 $log = _new WebPAC::Common()->_get_logger();

        my $indexer;

        if ($use_indexer eq 'hyperestraier') {

                #
                # open Hyper Estraier database
                #

                use WebPAC::Output::Estraier '0.10';
                my $est_config = $config->{hyperestraier} || $log->logdie("can't find 'hyperestraier' part in confguration");
                $est_config->{database} = $database;
                $est_config->{clean} = $clean;
                $est_config->{label} = $db_config->{name};

                $indexer = new WebPAC::Output::Estraier( %{ $est_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},
                );
                $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 $normalize_path = $input->{normalize}->{path};

                if ($normalize_path =~ m/\.xml$/i) {
                        $n->open(
                                tag => $input->{normalize}->{tag},
                                xml_file => $input->{normalize}->{path},
                        );
                } elsif ($normalize_path =~ m/\.(?:yml|yaml)$/i) {
                        $n->open_yaml(
                                path => $normalize_path,
                                tag => $input->{normalize}->{tag},
                        );
                }

                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 = $n->data_structure($row);

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

                        $total_rows++;
                }

        };

        my $dt = time() - $start_t;
        $log->info("$total_rows records indexed in " .
                sprintf("%.2f sec [%.2f rec/sec]",
                        $dt, ($total_rows / $dt)
                )
        );

        if ($use_indexer eq 'hyperestraier') {
                #
                # add Hyper Estraier links to other databases
                #
                if (ref($db_config->{links}) eq 'ARRAY') {
                        foreach my $link (@{ $db_config->{links} }) {
                                $log->info("adding link $database -> $link->{to} [$link->{credit}]");
                                $indexer->add_link(
                                        from => $database,
                                        to => $link->{to},
                                        credit => $link->{credit},
                                );
                        }
                }
        } else {
                $log->warn("links not implemented for $use_indexer");
        }

}

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			use WebPAC::Output::TT;
16	dpavlin	141	use YAML qw/LoadFile/;
17	dpavlin	301	use Getopt::Long;
18			use File::Path;
19	dpavlin	389	use Time::HiRes qw/time/;
20	dpavlin	74
21	dpavlin	301	=head1 NAME
22	dpavlin	76
23	dpavlin	301	run.pl - start WebPAC indexing
24	dpavlin	141
25	dpavlin	301	B<this command will probably go away. Don't get used to it!>
26	dpavlin	141
27	dpavlin	301	Options:
28
29			=over 4
30
31			=item --offset 42
32
33			start loading (all) databases at offset 42
34
35			=item --limit 100
36
37			limit loading to 100 records
38
39			=item --clean
40
41			remove database and Hyper Estraier index before indexing
42
43	dpavlin	423	=item --only=database_name
44	dpavlin	335
45	dpavlin	423	reindex just single database (legacy name is --one)
46	dpavlin	335
47	dpavlin	301	=item --config conf/config.yml
48
49			path to YAML configuration file
50
51			=back
52
53			=cut
54
55			my $offset;
56			my $limit;
57
58			my $clean = 0;
59			my $config = 'conf/config.yml';
60			my $debug = 0;
61	dpavlin	423	my $only_db_name;
62	dpavlin	301
63			GetOptions(
64			"limit=i" => \$limit,
65			"offset=i" => \$offset,
66			"clean" => \$clean,
67	dpavlin	423	"one=s" => \$only_db_name,
68			"only=s" => \$only_db_name,
69	dpavlin	301	"config" => \$config,
70			"debug" => \$debug,
71			);
72
73			$config = LoadFile($config);
74
75			print "config = ",Dumper($config) if ($debug);
76
77	dpavlin	210	die "no databases in config file!\n" unless ($config->{databases});
78	dpavlin	430	my $use_indexer = $config->{use_indexer} \|\| 'hyperestraier';
79	dpavlin	141
80	dpavlin	213	my $total_rows = 0;
81	dpavlin	389	my $start_t = time();
82	dpavlin	213
83	dpavlin	210	while (my ($database, $db_config) = each %{ $config->{databases} }) {
84	dpavlin	141
85	dpavlin	423	next if ($only_db_name && $database !~ m/$only_db_name/i);
86	dpavlin	335
87	dpavlin	255	my $log = _new WebPAC::Common()->_get_logger();
88
89	dpavlin	430	my $indexer;
90	dpavlin	255
91	dpavlin	430	if ($use_indexer eq 'hyperestraier') {
92	dpavlin	255
93	dpavlin	430	#
94			# open Hyper Estraier database
95			#
96	dpavlin	255
97	dpavlin	430	use WebPAC::Output::Estraier '0.10';
98			my $est_config = $config->{hyperestraier} \|\| $log->logdie("can't find 'hyperestraier' part in confguration");
99			$est_config->{database} = $database;
100			$est_config->{clean} = $clean;
101			$est_config->{label} = $db_config->{name};
102
103			$indexer = new WebPAC::Output::Estraier( %{ $est_config } );
104			} else {
105			$log->logdie("unknown use_indexer: $use_indexer");
106			}
107
108			$log->logide("can't continue without valid indexer") unless ($indexer);
109
110	dpavlin	255	#
111			# now WebPAC::Store
112			#
113	dpavlin	210	my $abs_path = abs_path($0);
114			$abs_path =~ s#/[^/]*$#/#;
115	dpavlin	141
116	dpavlin	210	my $db_path = $config->{webpac}->{db_path} . '/' . $database;
117	dpavlin	74
118	dpavlin	301	if ($clean) {
119			$log->info("creating new database $database in $db_path");
120			rmtree( $db_path ) \|\| $log->warn("can't remove $db_path: $!");
121			} else {
122	dpavlin	391	$log->debug("working on $database in $db_path");
123	dpavlin	301	}
124	dpavlin	255
125	dpavlin	210	my $db = new WebPAC::Store(
126			path => $db_path,
127	dpavlin	217	database => $database,
128	dpavlin	301	debug => $debug,
129	dpavlin	210	);
130	dpavlin	74
131	dpavlin	233
132	dpavlin	213	#
133			# now, iterate through input formats
134			#
135	dpavlin	74
136	dpavlin	213	my @inputs;
137			if (ref($db_config->{input}) eq 'ARRAY') {
138			@inputs = @{ $db_config->{input} };
139	dpavlin	255	} elsif ($db_config->{input}) {
140			push @inputs, $db_config->{input};
141	dpavlin	213	} else {
142	dpavlin	255	$log->info("database $database doesn't have inputs defined");
143	dpavlin	213	}
144	dpavlin	74
145	dpavlin	286	my @supported_inputs = keys %{ $config->{webpac}->{inputs} };
146
147	dpavlin	213	foreach my $input (@inputs) {
148	dpavlin	233
149			my $type = lc($input->{type});
150
151	dpavlin	286	die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs));
152	dpavlin	233
153			my $lookup = new WebPAC::Lookup(
154			lookup_file => $input->{lookup},
155			);
156
157	dpavlin	286	my $input_module = $config->{webpac}->{inputs}->{$type};
158	dpavlin	74
159	dpavlin	391	$log->info("working on input '$input->{path}' [$input->{type}] using $input_module lookup '$input->{lookup}'");
160	dpavlin	286
161	dpavlin	287	my $input_db = new WebPAC::Input(
162			module => $input_module,
163	dpavlin	213	code_page => $config->{webpac}->{webpac_encoding},
164	dpavlin	301	limit => $limit \|\| $input->{limit},
165			offset => $offset,
166	dpavlin	251	lookup => $lookup,
167	dpavlin	416	recode => $input->{recode},
168	dpavlin	287	);
169	dpavlin	286	$log->logdie("can't create input using $input_module") unless ($input);
170	dpavlin	113
171	dpavlin	287	my $maxmfn = $input_db->open(
172	dpavlin	285	path => $input->{path},
173	dpavlin	213	code_page => $input->{encoding}, # database encoding
174			);
175	dpavlin	113
176	dpavlin	213	my $n = new WebPAC::Normalize::XML(
177			# filter => { 'foo' => sub { shift } },
178			db => $db,
179			lookup_regex => $lookup->regex,
180			lookup => $lookup,
181	dpavlin	221	prefix => $input->{name},
182	dpavlin	213	);
183	dpavlin	113
184	dpavlin	269	my $normalize_path = $input->{normalize}->{path};
185	dpavlin	210
186	dpavlin	269	if ($normalize_path =~ m/\.xml$/i) {
187			$n->open(
188			tag => $input->{normalize}->{tag},
189			xml_file => $input->{normalize}->{path},
190			);
191			} elsif ($normalize_path =~ m/\.(?:yml\|yaml)$/i) {
192			$n->open_yaml(
193			path => $normalize_path,
194			tag => $input->{normalize}->{tag},
195			);
196			}
197
198	dpavlin	290	foreach my $pos ( 0 ... $input_db->size ) {
199	dpavlin	210
200	dpavlin	287	my $row = $input_db->fetch \|\| next;
201	dpavlin	213
202	dpavlin	291	my $mfn = $row->{'000'}->[0];
203	dpavlin	213
204	dpavlin	291	if (! $mfn \|\| $mfn !~ m#^\d+$#) {
205	dpavlin	290	$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
206			$mfn = $pos;
207	dpavlin	291	push @{ $row->{'000'} }, $pos;
208	dpavlin	290	}
209
210	dpavlin	213	my $ds = $n->data_structure($row);
211
212	dpavlin	430	$indexer->add(
213	dpavlin	291	id => $input->{name} . "/" . $mfn,
214	dpavlin	213	ds => $ds,
215	dpavlin	430	type => $config->{$use_indexer}->{type},
216	dpavlin	213	);
217
218			$total_rows++;
219			}
220
221	dpavlin	210	};
222
223	dpavlin	389	my $dt = time() - $start_t;
224			$log->info("$total_rows records indexed in " .
225			sprintf("%.2f sec [%.2f rec/sec]",
226			$dt, ($total_rows / $dt)
227			)
228			);
229	dpavlin	255
230	dpavlin	430	if ($use_indexer eq 'hyperestraier') {
231			#
232			# add Hyper Estraier links to other databases
233			#
234			if (ref($db_config->{links}) eq 'ARRAY') {
235			foreach my $link (@{ $db_config->{links} }) {
236			$log->info("adding link $database -> $link->{to} [$link->{credit}]");
237			$indexer->add_link(
238			from => $database,
239			to => $link->{to},
240			credit => $link->{credit},
241			);
242			}
243	dpavlin	255	}
244	dpavlin	430	} else {
245			$log->warn("links not implemented for $use_indexer");
246	dpavlin	255	}
247
248	dpavlin	210	}
249