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 WebPAC::Output::Estraier '0.10';
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 $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();

        #
        # open Hyper Estraier database
        #

        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};

        my $est = new WebPAC::Output::Estraier( %{ $est_config } );

        #
        # 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);

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

        #
        # 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}]");
                        $est->add_link(
                                from => $database,
                                to => $link->{to},
                                credit => $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			use WebPAC::Output::TT;
16	dpavlin	401	use WebPAC::Output::Estraier '0.10';
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	74
22	dpavlin	301	=head1 NAME
23	dpavlin	76
24	dpavlin	301	run.pl - start WebPAC indexing
25	dpavlin	141
26	dpavlin	301	B<this command will probably go away. Don't get used to it!>
27	dpavlin	141
28	dpavlin	301	Options:
29
30			=over 4
31
32			=item --offset 42
33
34			start loading (all) databases at offset 42
35
36			=item --limit 100
37
38			limit loading to 100 records
39
40			=item --clean
41
42			remove database and Hyper Estraier index before indexing
43
44	dpavlin	423	=item --only=database_name
45	dpavlin	335
46	dpavlin	423	reindex just single database (legacy name is --one)
47	dpavlin	335
48	dpavlin	301	=item --config conf/config.yml
49
50			path to YAML configuration file
51
52			=back
53
54			=cut
55
56			my $offset;
57			my $limit;
58
59			my $clean = 0;
60			my $config = 'conf/config.yml';
61			my $debug = 0;
62	dpavlin	423	my $only_db_name;
63	dpavlin	301
64			GetOptions(
65			"limit=i" => \$limit,
66			"offset=i" => \$offset,
67			"clean" => \$clean,
68	dpavlin	423	"one=s" => \$only_db_name,
69			"only=s" => \$only_db_name,
70	dpavlin	301	"config" => \$config,
71			"debug" => \$debug,
72			);
73
74			$config = LoadFile($config);
75
76			print "config = ",Dumper($config) if ($debug);
77
78	dpavlin	210	die "no databases in config file!\n" unless ($config->{databases});
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			#
90			# open Hyper Estraier database
91			#
92
93			my $est_config = $config->{hyperestraier} \|\| $log->logdie("can't find 'hyperestraier' part in confguration");
94			$est_config->{database} = $database;
95	dpavlin	307	$est_config->{clean} = $clean;
96	dpavlin	401	$est_config->{label} = $db_config->{name};
97	dpavlin	255
98	dpavlin	307	my $est = new WebPAC::Output::Estraier( %{ $est_config } );
99	dpavlin	255
100			#
101			# now WebPAC::Store
102			#
103	dpavlin	210	my $abs_path = abs_path($0);
104			$abs_path =~ s#/[^/]*$#/#;
105	dpavlin	141
106	dpavlin	210	my $db_path = $config->{webpac}->{db_path} . '/' . $database;
107	dpavlin	74
108	dpavlin	301	if ($clean) {
109			$log->info("creating new database $database in $db_path");
110			rmtree( $db_path ) \|\| $log->warn("can't remove $db_path: $!");
111			} else {
112	dpavlin	391	$log->debug("working on $database in $db_path");
113	dpavlin	301	}
114	dpavlin	255
115	dpavlin	210	my $db = new WebPAC::Store(
116			path => $db_path,
117	dpavlin	217	database => $database,
118	dpavlin	301	debug => $debug,
119	dpavlin	210	);
120	dpavlin	74
121	dpavlin	233
122	dpavlin	213	#
123			# now, iterate through input formats
124			#
125	dpavlin	74
126	dpavlin	213	my @inputs;
127			if (ref($db_config->{input}) eq 'ARRAY') {
128			@inputs = @{ $db_config->{input} };
129	dpavlin	255	} elsif ($db_config->{input}) {
130			push @inputs, $db_config->{input};
131	dpavlin	213	} else {
132	dpavlin	255	$log->info("database $database doesn't have inputs defined");
133	dpavlin	213	}
134	dpavlin	74
135	dpavlin	286	my @supported_inputs = keys %{ $config->{webpac}->{inputs} };
136
137	dpavlin	213	foreach my $input (@inputs) {
138	dpavlin	233
139			my $type = lc($input->{type});
140
141	dpavlin	286	die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs));
142	dpavlin	233
143			my $lookup = new WebPAC::Lookup(
144			lookup_file => $input->{lookup},
145			);
146
147	dpavlin	286	my $input_module = $config->{webpac}->{inputs}->{$type};
148	dpavlin	74
149	dpavlin	391	$log->info("working on input '$input->{path}' [$input->{type}] using $input_module lookup '$input->{lookup}'");
150	dpavlin	286
151	dpavlin	287	my $input_db = new WebPAC::Input(
152			module => $input_module,
153	dpavlin	213	code_page => $config->{webpac}->{webpac_encoding},
154	dpavlin	301	limit => $limit \|\| $input->{limit},
155			offset => $offset,
156	dpavlin	251	lookup => $lookup,
157	dpavlin	416	recode => $input->{recode},
158	dpavlin	287	);
159	dpavlin	286	$log->logdie("can't create input using $input_module") unless ($input);
160	dpavlin	113
161	dpavlin	287	my $maxmfn = $input_db->open(
162	dpavlin	285	path => $input->{path},
163	dpavlin	213	code_page => $input->{encoding}, # database encoding
164			);
165	dpavlin	113
166	dpavlin	213	my $n = new WebPAC::Normalize::XML(
167			# filter => { 'foo' => sub { shift } },
168			db => $db,
169			lookup_regex => $lookup->regex,
170			lookup => $lookup,
171	dpavlin	221	prefix => $input->{name},
172	dpavlin	213	);
173	dpavlin	113
174	dpavlin	269	my $normalize_path = $input->{normalize}->{path};
175	dpavlin	210
176	dpavlin	269	if ($normalize_path =~ m/\.xml$/i) {
177			$n->open(
178			tag => $input->{normalize}->{tag},
179			xml_file => $input->{normalize}->{path},
180			);
181			} elsif ($normalize_path =~ m/\.(?:yml\|yaml)$/i) {
182			$n->open_yaml(
183			path => $normalize_path,
184			tag => $input->{normalize}->{tag},
185			);
186			}
187
188	dpavlin	290	foreach my $pos ( 0 ... $input_db->size ) {
189	dpavlin	210
190	dpavlin	287	my $row = $input_db->fetch \|\| next;
191	dpavlin	213
192	dpavlin	291	my $mfn = $row->{'000'}->[0];
193	dpavlin	213
194	dpavlin	291	if (! $mfn \|\| $mfn !~ m#^\d+$#) {
195	dpavlin	290	$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
196			$mfn = $pos;
197	dpavlin	291	push @{ $row->{'000'} }, $pos;
198	dpavlin	290	}
199
200	dpavlin	213	my $ds = $n->data_structure($row);
201
202			$est->add(
203	dpavlin	291	id => $input->{name} . "/" . $mfn,
204	dpavlin	213	ds => $ds,
205			type => $config->{hyperestraier}->{type},
206			);
207
208			$total_rows++;
209			}
210
211	dpavlin	210	};
212
213	dpavlin	389	my $dt = time() - $start_t;
214			$log->info("$total_rows records indexed in " .
215			sprintf("%.2f sec [%.2f rec/sec]",
216			$dt, ($total_rows / $dt)
217			)
218			);
219	dpavlin	255
220			#
221			# add Hyper Estraier links to other databases
222			#
223			if (ref($db_config->{links}) eq 'ARRAY') {
224			foreach my $link (@{ $db_config->{links} }) {
225			$log->info("adding link $database -> $link->{to} [$link->{credit}]");
226			$est->add_link(
227			from => $database,
228			to => $link->{to},
229			credit => $link->{credit},
230			);
231			}
232			}
233
234	dpavlin	210	}
235