webpac2/trunk/run.pl

#!/usr/bin/perl -w

use strict;

use Cwd qw/abs_path/;
use File::Temp qw/tempdir/;
use lib './lib';

use WebPAC::Common 0.02;
use WebPAC::Parser 0.04;
use WebPAC::Input 0.13;
use WebPAC::Store 0.11;
use WebPAC::Normalize 0.21;
use WebPAC::Output::TT;
use WebPAC::Validate 0.06;
use WebPAC::Output::MARC;
use WebPAC::Config;
use Getopt::Long;
use File::Path;
use Time::HiRes qw/time/;
use File::Slurp;
use Data::Dump qw/dump/;
use Storable qw/dclone/;

use Proc::Queue size => 1;
use POSIX ":sys_wait_h"; # imports WNOHANG

=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 --stats

disable indexing, modify_* in configuration 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>

=item --marc-normalize conf/normalize/mapping.pl

This option specifies normalisation file for MARC creation

=item --marc-output out/marc/test.marc

Optional path to output file

=item --marc-lint

By default turned on if C<--marc-normalize> is used. You can disable lint
messages with C<--no-marc-lint>.

=item --marc-dump

Force dump or input and marc record for debugging.

=item --parallel 4

Run databases in parallel (aproximatly same as number of processors in
machine if you want to use full load)

=item --only-links

Create just links

=item --merge

Create merged index of databases which have links

=back

=cut

my $offset;
my $limit;

my $clean = 0;
my $config_path;
my $debug = 0;
my $only_filter;
my $stats = 0;
my $validate_path;
my ($marc_normalize, $marc_output);
my $marc_lint = 1;
my $marc_dump = 0;
my $parallel = 0;
my $only_links = 0;
my $merge = 0;

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

GetOptions(
        "limit=i" => \$limit,
        "offset=i" => \$offset,
        "clean" => \$clean,
        "one=s" => \$only_filter,
        "only=s" => \$only_filter,
        "config" => \$config_path,
        "debug+" => \$debug,
        "stats" => \$stats,
        "validate=s" => \$validate_path,
        "marc-normalize=s" => \$marc_normalize,
        "marc-output=s" => \$marc_output,
        "marc-lint!" => \$marc_lint,
        "marc-dump!" => \$marc_dump,
        "parallel=i" => \$parallel,
        "only-links!" => \$only_links,
        "merge" => \$merge,
);

my $config = new WebPAC::Config( path => $config_path );

#print "config = ",dump($config) if ($debug);

die "no databases in config file!\n" unless ($config->databases);

$log->info( "-" x 79 );


my $estcmd_fh;
my $estcmd_path = './estcmd-merge.sh';
if ($merge) {
        open($estcmd_fh, '>', $estcmd_path) || $log->logdie("can't open $estcmd_path: $!");
        print $estcmd_fh 'cd /data/estraier/_node/ || exit 1',$/;
        print $estcmd_fh 'sudo /etc/init.d/hyperestraier stop',$/;
        $log->info("created merge batch file $estcmd_path");
}


my $validate;
$validate = new WebPAC::Validate(
        path => $validate_path,
) if ($validate_path);


my $use_indexer = $config->use_indexer;
if ($stats) {
        $log->debug("option --stats disables update of indexing engine...");
        $use_indexer = undef;
} else {
        $log->info("using $use_indexer indexing engine...");
}

# disable indexing when creating marc
$use_indexer = undef if ($marc_normalize);

# parse normalize files and create source files for lookup and normalization

my $parser = new WebPAC::Parser( config => $config );

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

my @links;

if ($parallel) {
        $log->info("Using $parallel processes for speedup");
        Proc::Queue::size($parallel);
}

sub create_ds_config {
        my ($db_config, $database, $input, $mfn) = @_;
        my $c = dclone( $db_config );
        $c->{_} = $database || $log->logconfess("need database");
        $c->{_mfn} = $mfn || $log->logconfess("need mfn");
        $c->{input} = $input || $log->logconfess("need input");
        return $c;
}

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 ($parallel) {
                my $f=fork;
                if(defined ($f) and $f==0) {
                        $log->info("Created processes $$ for speedup");
                } else {
                        next;
                }
        }

        my $indexer;
        if ($use_indexer) {

                my $cfg_name = $use_indexer;
                $cfg_name =~ s/\-.*$//;

                my $indexer_config = $config->get( $cfg_name ) || $log->logdie("can't find '$cfg_name' part in confguration");
                $indexer_config->{database} = $database;
                $indexer_config->{clean} = $clean;
                $indexer_config->{label} = $db_config->{name};

                # force clean if database has links
                $indexer_config->{clean} = 1 if ($db_config->{links});

                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 'hyperestraier-native') {

                        # open Hyper Estraier database
                        use WebPAC::Output::EstraierNative;
                        $indexer = new WebPAC::Output::EstraierNative( %{ $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);
        }


        #
        # store Hyper Estraier links to other databases
        #
        if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) {
                foreach my $link (@{ $db_config->{links} }) {
                        if ($use_indexer eq 'hyperestraier') {
                                if ($merge) {
                                        print $estcmd_fh 'sudo -u www-data estcmd merge ' . $database . ' ' . $link->{to},$/;
                                } else {
                                        $log->info("saving link $database -> $link->{to} [$link->{credit}]");
                                        push @links, sub {
                                                $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}]");
                        }
                }
        }
        next if ($only_links);


        #
        # now WebPAC::Store
        #
        my $abs_path = abs_path($0);
        $abs_path =~ s#/[^/]*$#/#;

        my $db_path = $config->webpac('db_path');

        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 $store = new WebPAC::Store(
                path => $db_path,
                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");
        }

        foreach my $input (@inputs) {

                my $input_name = $input->{name} || $log->logdie("input without a name isn't valid: ",dump($input));

                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(",", $config->webpac('inputs') ), " not '$type'!\n" unless (grep(/$type/, $config->webpac('inputs')));

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

                my @lookups = $parser->have_lookup_create($database, $input);

                $log->info("working on input '$input_name' in $input->{path} [type: $input->{type}] using $input_module",
                        @lookups ? " creating lookups: ".join(", ", @lookups) : ""
                );

                if ($stats) {
                        # disable modification of records if --stats is in use
                        delete($input->{modify_records});
                        delete($input->{modify_file});
                }

                my $input_db = new WebPAC::Input(
                        module => $input_module,
                        encoding => $config->webpac('webpac_encoding'),
                        limit => $limit || $input->{limit},
                        offset => $offset,
                        recode => $input->{recode},
                        stats => $stats,
                        modify_records => $input->{modify_records},
                        modify_file => $input->{modify_file},
                );
                $log->logdie("can't create input using $input_module") unless ($input);

                if (defined( $input->{lookup} )) {
                        $log->warn("$database/$input_name has depriciated lookup definition, removing it...");
                        delete( $input->{lookup} );
                }

                my $lookup_coderef;

                if (@lookups) {

                        my $rules = $parser->lookup_create_rules($database, $input) || $log->logdie("no rules found for $database/$input");

                        $lookup_coderef = sub {
                                my $rec = shift || die "need rec!";
                                my $mfn = $rec->{'000'}->[0] || die "need mfn in 000";

                                WebPAC::Normalize::data_structure(
                                        row => $rec,
                                        rules => "_debug(2); $rules _debug(0)",
                                        config => create_ds_config( $db_config, $database, $input, $mfn ),
                                );

                                warn "current lookup: ", dump(WebPAC::Normalize::_get_lookup());
                        };

                        WebPAC::Normalize::_set_lookup( undef );

                        $log->debug("created lookup_coderef using:\n$rules");

                };

                my $lookup_jar;

                my $maxmfn = $input_db->open(
                        path => $input->{path},
                        code_page => $input->{encoding},        # database encoding
                        lookup_coderef => $lookup_coderef,
                        lookup => $lookup_jar,
                        %{ $input },
                );

                my $lookup_data = WebPAC::Normalize::_get_lookup();

                if (defined( $lookup_data->{$database}->{$input_name} )) {
                        $log->debug("created following lookups: ", dump( $lookup_data ));

                        foreach my $key (keys %{ $lookup_data->{$database}->{$input_name} }) {
                                $store->save_lookup(
                                        database => $database,
                                        input => $input_name,
                                        key => $key,
                                        data => $lookup_data->{$database}->{$input_name}->{$key},
                                );
                        }
                }

                my $report_fh;
                if ($stats || $validate) {
                        my $path = "out/report/${database}-${input_name}.txt";
                        open($report_fh, '>', $path) || $log->logdie("can't open $path: $!");

                        print $report_fh "Report for database '$database' input '$input_name' records ",
                                $offset || 1, "-", $limit || $input->{limit} || $maxmfn, "\n\n";
                        $log->info("Generating report file $path");
                }

                my $marc;
                if ($marc_normalize) {
                        $marc = new WebPAC::Output::MARC(
                                path => $marc_output || "out/marc/${database}-${input_name}.marc",
                                lint => $marc_lint,
                                dump => $marc_dump,
                        );
                }

                my $rules = $parser->normalize_rules($database,$input_name) || $log->logdie("no normalize rules found for $database/$input_name");
                $log->debug("parsed normalize rules:\n$rules");

                # reset position in database
                $input_db->seek(1);

                # generate name of config key for indexer (strip everything after -)
                my $indexer_config = $use_indexer;
                $indexer_config =~ s/^(\w+)-?.*$/$1/g if ($indexer_config);

                my $lookup_hash;
                my $depends = $parser->depends($database,$input_name);
        
                if ($depends) {
                        $log->debug("$database/$input_name depends on: ", dump($depends)) if ($depends);
                        $log->logdie("parser->depends didn't return HASH") unless (ref($depends) eq 'HASH');

                        foreach my $db (keys %$depends) {
                                foreach my $i (keys %{$depends->{$db}}) {
                                        foreach my $k (keys %{$depends->{$db}->{$i}}) {
                                                $log->debug("loading lookup $db/$i");
                                                $lookup_hash->{$db}->{$i}->{$k} = $store->load_lookup(
                                                        database => $db,
                                                        input => $i,
                                                        key => $k,
                                                );
                                        }
                                }
                        }

                        $log->debug("lookup_hash = ", dump( $lookup_hash ));
                }


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

                        my $ds = WebPAC::Normalize::data_structure(
                                row => $row,
                                rules => $rules,
                                lookup => $lookup_hash,
                                config => create_ds_config( $db_config, $database, $input, $mfn ),
                                marc_encoding => 'utf-8',
                                load_ds_coderef => sub {
                                        my ($database,$input,$mfn) = @_;
                                        return $store->load_ds(
                                                database => $database,
                                                input => $input,
                                                id => $mfn,
                                        );
                                },
                        );

                        $store->save_ds(
                                database => $database,
                                input => $input_name,
                                id => $mfn,
                                ds => $ds,
                        ) if ($ds && !$stats);

                        $indexer->add(
                                id => "${input_name}/${mfn}",
                                ds => $ds,
                                type => $config->get($indexer_config)->{type},
                        ) if ($indexer && $ds);

                        if ($marc) {
                                my $i = 0;

                                while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) {
                                        $marc->add(
                                                id => $mfn . ( $i ? "/$i" : '' ),
                                                fields => $fields,
                                                leader => WebPAC::Normalize::marc_leader(),
                                                row => $row,
                                        );
                                        $i++;
                                }

                                $log->info("Created $i instances of MFN $mfn\n") if ($i > 1);
                        }

                        $total_rows++;
                }

                if ($validate) {
                        my $errors = $validate->report;
                        if ($errors) {
                                $log->info("validation errors:\n$errors\n" );
                                print $report_fh "$errors\n" if ($report_fh);
                        }
                }

                if ($stats) {
                        my $s = $input_db->stats;
                        $log->info("statistics of fields usage:\n$s");
                        print $report_fh "Statistics of fields usage:\n$s" if ($report_fh);
                }

                # close MARC file
                $marc->finish if ($marc);

                # close report
                close($report_fh) if ($report_fh)

        }

        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)
                )
        );


        # end forked process
        if ($parallel) {
                $log->info("parallel process $$ finished");
                exit(0);
        }

}

if ($parallel) {
        # wait all children to finish
        sleep(1) while wait != -1;
        $log->info("all parallel processes finished");
}

#
# handle links or merge after indexing
#

if ($merge) {
        print $estcmd_fh 'sudo /etc/init.d/hyperestraier start',$/;
        close($estcmd_fh);
        chmod 0700, $estcmd_path || $log->warn("can't chmod 0700 $estcmd_path: $!");
        system $estcmd_path;
} else {
        foreach my $link (@links) {
                $log->logdie("coderef in link ", Dumper($link), " is ", ref($link), " and not CODE") unless (ref($link) eq 'CODE');
                $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 lib './lib';
8
9	dpavlin	255	use WebPAC::Common 0.02;
10	dpavlin	698	use WebPAC::Parser 0.04;
11	dpavlin	707	use WebPAC::Input 0.13;
12	dpavlin	713	use WebPAC::Store 0.11;
13	dpavlin	731	use WebPAC::Normalize 0.21;
14	dpavlin	74	use WebPAC::Output::TT;
15	dpavlin	652	use WebPAC::Validate 0.06;
16	dpavlin	578	use WebPAC::Output::MARC;
17	dpavlin	684	use WebPAC::Config;
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	556	use Data::Dump qw/dump/;
23	dpavlin	595	use Storable qw/dclone/;
24	dpavlin	74
25	dpavlin	606	use Proc::Queue size => 1;
26			use POSIX ":sys_wait_h"; # imports WNOHANG
27
28	dpavlin	301	=head1 NAME
29	dpavlin	76
30	dpavlin	301	run.pl - start WebPAC indexing
31	dpavlin	141
32	dpavlin	301	B<this command will probably go away. Don't get used to it!>
33	dpavlin	141
34	dpavlin	301	Options:
35
36			=over 4
37
38			=item --offset 42
39
40			start loading (all) databases at offset 42
41
42			=item --limit 100
43
44			limit loading to 100 records
45
46			=item --clean
47
48			remove database and Hyper Estraier index before indexing
49
50	dpavlin	510	=item --only=database_name/input_filter
51	dpavlin	335
52	dpavlin	423	reindex just single database (legacy name is --one)
53	dpavlin	335
54	dpavlin	510	C</input_filter> is optional part which can be C<name>
55			or C<type> from input
56
57	dpavlin	301	=item --config conf/config.yml
58
59			path to YAML configuration file
60
61	dpavlin	507	=item --stats
62
63	dpavlin	638	disable indexing, modify_* in configuration and dump statistics about field
64			and subfield usage for each input
65	dpavlin	507
66	dpavlin	516	=item --validate path/to/validation_file
67
68			turn on extra validation of imput records, see L<WebPAC::Validation>
69
70	dpavlin	552	=item --marc-normalize conf/normalize/mapping.pl
71
72			This option specifies normalisation file for MARC creation
73
74			=item --marc-output out/marc/test.marc
75
76			Optional path to output file
77
78	dpavlin	556	=item --marc-lint
79
80			By default turned on if C<--marc-normalize> is used. You can disable lint
81	dpavlin	558	messages with C<--no-marc-lint>.
82	dpavlin	556
83	dpavlin	559	=item --marc-dump
84
85			Force dump or input and marc record for debugging.
86
87	dpavlin	606	=item --parallel 4
88
89			Run databases in parallel (aproximatly same as number of processors in
90			machine if you want to use full load)
91
92	dpavlin	607	=item --only-links
93
94			Create just links
95
96	dpavlin	608	=item --merge
97
98			Create merged index of databases which have links
99
100	dpavlin	301	=back
101
102			=cut
103
104			my $offset;
105			my $limit;
106
107			my $clean = 0;
108	dpavlin	684	my $config_path;
109	dpavlin	301	my $debug = 0;
110	dpavlin	510	my $only_filter;
111	dpavlin	507	my $stats = 0;
112	dpavlin	516	my $validate_path;
113	dpavlin	552	my ($marc_normalize, $marc_output);
114	dpavlin	556	my $marc_lint = 1;
115	dpavlin	559	my $marc_dump = 0;
116	dpavlin	606	my $parallel = 0;
117	dpavlin	607	my $only_links = 0;
118	dpavlin	608	my $merge = 0;
119	dpavlin	606
120	dpavlin	611	my $log = _new WebPAC::Common()->_get_logger();
121
122	dpavlin	301	GetOptions(
123			"limit=i" => \$limit,
124			"offset=i" => \$offset,
125			"clean" => \$clean,
126	dpavlin	510	"one=s" => \$only_filter,
127			"only=s" => \$only_filter,
128	dpavlin	684	"config" => \$config_path,
129	dpavlin	560	"debug+" => \$debug,
130	dpavlin	507	"stats" => \$stats,
131	dpavlin	516	"validate=s" => \$validate_path,
132	dpavlin	552	"marc-normalize=s" => \$marc_normalize,
133			"marc-output=s" => \$marc_output,
134	dpavlin	556	"marc-lint!" => \$marc_lint,
135	dpavlin	559	"marc-dump!" => \$marc_dump,
136	dpavlin	606	"parallel=i" => \$parallel,
137	dpavlin	607	"only-links!" => \$only_links,
138	dpavlin	608	"merge" => \$merge,
139	dpavlin	301	);
140
141	dpavlin	684	my $config = new WebPAC::Config( path => $config_path );
142	dpavlin	301
143	dpavlin	611	#print "config = ",dump($config) if ($debug);
144	dpavlin	301
145	dpavlin	684	die "no databases in config file!\n" unless ($config->databases);
146	dpavlin	431
147	dpavlin	509	$log->info( "-" x 79 );
148	dpavlin	431
149	dpavlin	608
150			my $estcmd_fh;
151			my $estcmd_path = './estcmd-merge.sh';
152			if ($merge) {
153			open($estcmd_fh, '>', $estcmd_path) \|\| $log->logdie("can't open $estcmd_path: $!");
154			print $estcmd_fh 'cd /data/estraier/_node/ \|\| exit 1',$/;
155			print $estcmd_fh 'sudo /etc/init.d/hyperestraier stop',$/;
156			$log->info("created merge batch file $estcmd_path");
157			}
158
159
160	dpavlin	516	my $validate;
161			$validate = new WebPAC::Validate(
162			path => $validate_path,
163			) if ($validate_path);
164
165	dpavlin	608
166	dpavlin	684	my $use_indexer = $config->use_indexer;
167	dpavlin	509	if ($stats) {
168			$log->debug("option --stats disables update of indexing engine...");
169			$use_indexer = undef;
170			} else {
171			$log->info("using $use_indexer indexing engine...");
172			}
173	dpavlin	141
174	dpavlin	552	# disable indexing when creating marc
175			$use_indexer = undef if ($marc_normalize);
176
177	dpavlin	698	# parse normalize files and create source files for lookup and normalization
178
179			my $parser = new WebPAC::Parser( config => $config );
180
181	dpavlin	213	my $total_rows = 0;
182	dpavlin	389	my $start_t = time();
183	dpavlin	213
184	dpavlin	529	my @links;
185
186	dpavlin	606	if ($parallel) {
187			$log->info("Using $parallel processes for speedup");
188			Proc::Queue::size($parallel);
189			}
190
191	dpavlin	707	sub create_ds_config {
192			my ($db_config, $database, $input, $mfn) = @_;
193			my $c = dclone( $db_config );
194			$c->{_} = $database \|\| $log->logconfess("need database");
195			$c->{_mfn} = $mfn \|\| $log->logconfess("need mfn");
196			$c->{input} = $input \|\| $log->logconfess("need input");
197			return $c;
198			}
199
200	dpavlin	684	while (my ($database, $db_config) = each %{ $config->databases }) {
201	dpavlin	141
202	dpavlin	512	my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter);
203	dpavlin	510	next if ($only_database && $database !~ m/$only_database/i);
204	dpavlin	335
205	dpavlin	606	if ($parallel) {
206			my $f=fork;
207			if(defined ($f) and $f==0) {
208			$log->info("Created processes $$ for speedup");
209			} else {
210			next;
211			}
212			}
213
214	dpavlin	607	my $indexer;
215	dpavlin	509	if ($use_indexer) {
216	dpavlin	627
217			my $cfg_name = $use_indexer;
218			$cfg_name =~ s/\-.*$//;
219
220	dpavlin	684	my $indexer_config = $config->get( $cfg_name ) \|\| $log->logdie("can't find '$cfg_name' part in confguration");
221	dpavlin	509	$indexer_config->{database} = $database;
222			$indexer_config->{clean} = $clean;
223			$indexer_config->{label} = $db_config->{name};
224	dpavlin	431
225	dpavlin	608	# force clean if database has links
226			$indexer_config->{clean} = 1 if ($db_config->{links});
227
228	dpavlin	509	if ($use_indexer eq 'hyperestraier') {
229	dpavlin	255
230	dpavlin	509	# open Hyper Estraier database
231			use WebPAC::Output::Estraier '0.10';
232			$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
233
234	dpavlin	627	} elsif ($use_indexer eq 'hyperestraier-native') {
235
236			# open Hyper Estraier database
237			use WebPAC::Output::EstraierNative;
238			$indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } );
239
240	dpavlin	509	} elsif ($use_indexer eq 'kinosearch') {
241	dpavlin	430
242	dpavlin	509	# open KinoSearch
243			use WebPAC::Output::KinoSearch;
244			$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path});
245			$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } );
246	dpavlin	431
247	dpavlin	509	} else {
248			$log->logdie("unknown use_indexer: $use_indexer");
249			}
250
251			$log->logide("can't continue without valid indexer") unless ($indexer);
252	dpavlin	430	}
253
254
255	dpavlin	255	#
256	dpavlin	607	# store Hyper Estraier links to other databases
257			#
258			if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) {
259			foreach my $link (@{ $db_config->{links} }) {
260			if ($use_indexer eq 'hyperestraier') {
261	dpavlin	608	if ($merge) {
262			print $estcmd_fh 'sudo -u www-data estcmd merge ' . $database . ' ' . $link->{to},$/;
263			} else {
264			$log->info("saving link $database -> $link->{to} [$link->{credit}]");
265			push @links, sub {
266			$log->info("adding link $database -> $link->{to} [$link->{credit}]");
267			$indexer->add_link(
268			from => $database,
269			to => $link->{to},
270			credit => $link->{credit},
271			);
272			};
273			}
274	dpavlin	607	} else {
275			$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
276			}
277			}
278			}
279			next if ($only_links);
280
281
282			#
283	dpavlin	255	# now WebPAC::Store
284			#
285	dpavlin	210	my $abs_path = abs_path($0);
286			$abs_path =~ s#/[^/]*$#/#;
287	dpavlin	141
288	dpavlin	713	my $db_path = $config->webpac('db_path');
289	dpavlin	74
290	dpavlin	301	if ($clean) {
291	dpavlin	511	$log->info("creating new database '$database' in $db_path");
292	dpavlin	301	rmtree( $db_path ) \|\| $log->warn("can't remove $db_path: $!");
293			} else {
294	dpavlin	511	$log->info("working on database '$database' in $db_path");
295	dpavlin	301	}
296	dpavlin	255
297	dpavlin	710	my $store = new WebPAC::Store(
298	dpavlin	210	path => $db_path,
299	dpavlin	301	debug => $debug,
300	dpavlin	210	);
301	dpavlin	74
302	dpavlin	233
303	dpavlin	213	#
304			# now, iterate through input formats
305			#
306	dpavlin	74
307	dpavlin	213	my @inputs;
308			if (ref($db_config->{input}) eq 'ARRAY') {
309			@inputs = @{ $db_config->{input} };
310	dpavlin	255	} elsif ($db_config->{input}) {
311			push @inputs, $db_config->{input};
312	dpavlin	213	} else {
313	dpavlin	255	$log->info("database $database doesn't have inputs defined");
314	dpavlin	213	}
315	dpavlin	74
316	dpavlin	213	foreach my $input (@inputs) {
317	dpavlin	233
318	dpavlin	710	my $input_name = $input->{name} \|\| $log->logdie("input without a name isn't valid: ",dump($input));
319	dpavlin	510
320	dpavlin	710	next if ($only_input && ($input_name !~ m#$only_input#i && $input->{type} !~ m#$only_input#i));
321
322	dpavlin	233	my $type = lc($input->{type});
323
324	dpavlin	684	die "I know only how to handle input types ", join(",", $config->webpac('inputs') ), " not '$type'!\n" unless (grep(/$type/, $config->webpac('inputs')));
325	dpavlin	233
326	dpavlin	684	my $input_module = $config->webpac('inputs')->{$type};
327	dpavlin	74
328	dpavlin	706	my @lookups = $parser->have_lookup_create($database, $input);
329
330	dpavlin	710	$log->info("working on input '$input_name' in $input->{path} [type: $input->{type}] using $input_module",
331	dpavlin	707	@lookups ? " creating lookups: ".join(", ", @lookups) : ""
332	dpavlin	523	);
333	dpavlin	286
334	dpavlin	638	if ($stats) {
335			# disable modification of records if --stats is in use
336			delete($input->{modify_records});
337			delete($input->{modify_file});
338			}
339
340	dpavlin	287	my $input_db = new WebPAC::Input(
341			module => $input_module,
342	dpavlin	684	encoding => $config->webpac('webpac_encoding'),
343	dpavlin	301	limit => $limit \|\| $input->{limit},
344			offset => $offset,
345	dpavlin	416	recode => $input->{recode},
346	dpavlin	507	stats => $stats,
347	dpavlin	600	modify_records => $input->{modify_records},
348	dpavlin	636	modify_file => $input->{modify_file},
349	dpavlin	287	);
350	dpavlin	286	$log->logdie("can't create input using $input_module") unless ($input);
351	dpavlin	113
352	dpavlin	702	if (defined( $input->{lookup} )) {
353	dpavlin	710	$log->warn("$database/$input_name has depriciated lookup definition, removing it...");
354	dpavlin	702	delete( $input->{lookup} );
355			}
356
357	dpavlin	707	my $lookup_coderef;
358
359			if (@lookups) {
360
361			my $rules = $parser->lookup_create_rules($database, $input) \|\| $log->logdie("no rules found for $database/$input");
362
363			$lookup_coderef = sub {
364			my $rec = shift \|\| die "need rec!";
365			my $mfn = $rec->{'000'}->[0] \|\| die "need mfn in 000";
366
367			WebPAC::Normalize::data_structure(
368			row => $rec,
369	dpavlin	731	rules => "_debug(2); $rules _debug(0)",
370	dpavlin	707	config => create_ds_config( $db_config, $database, $input, $mfn ),
371			);
372
373	dpavlin	731	warn "current lookup: ", dump(WebPAC::Normalize::_get_lookup());
374	dpavlin	707	};
375
376			WebPAC::Normalize::_set_lookup( undef );
377
378			$log->debug("created lookup_coderef using:\n$rules");
379
380			};
381
382	dpavlin	731	my $lookup_jar;
383
384	dpavlin	287	my $maxmfn = $input_db->open(
385	dpavlin	285	path => $input->{path},
386	dpavlin	213	code_page => $input->{encoding}, # database encoding
387	dpavlin	707	lookup_coderef => $lookup_coderef,
388	dpavlin	731	lookup => $lookup_jar,
389	dpavlin	523	%{ $input },
390	dpavlin	213	);
391	dpavlin	113
392	dpavlin	710	my $lookup_data = WebPAC::Normalize::_get_lookup();
393	dpavlin	707
394	dpavlin	712	if (defined( $lookup_data->{$database}->{$input_name} )) {
395			$log->debug("created following lookups: ", dump( $lookup_data ));
396	dpavlin	710
397	dpavlin	712	foreach my $key (keys %{ $lookup_data->{$database}->{$input_name} }) {
398	dpavlin	713	$store->save_lookup(
399			database => $database,
400			input => $input_name,
401			key => $key,
402			data => $lookup_data->{$database}->{$input_name}->{$key},
403			);
404	dpavlin	712	}
405	dpavlin	710	}
406
407	dpavlin	671	my $report_fh;
408			if ($stats \|\| $validate) {
409	dpavlin	710	my $path = "out/report/${database}-${input_name}.txt";
410	dpavlin	671	open($report_fh, '>', $path) \|\| $log->logdie("can't open $path: $!");
411
412	dpavlin	710	print $report_fh "Report for database '$database' input '$input_name' records ",
413	dpavlin	671	$offset \|\| 1, "-", $limit \|\| $input->{limit} \|\| $maxmfn, "\n\n";
414			$log->info("Generating report file $path");
415			}
416
417	dpavlin	731	my $marc;
418	dpavlin	552	if ($marc_normalize) {
419	dpavlin	731	$marc = new WebPAC::Output::MARC(
420			path => $marc_output \|\| "out/marc/${database}-${input_name}.marc",
421			lint => $marc_lint,
422			dump => $marc_dump,
423			);
424	dpavlin	552	}
425
426	dpavlin	731	my $rules = $parser->normalize_rules($database,$input_name) \|\| $log->logdie("no normalize rules found for $database/$input_name");
427			$log->debug("parsed normalize rules:\n$rules");
428	dpavlin	492
429	dpavlin	731	# reset position in database
430			$input_db->seek(1);
431	dpavlin	269
432	dpavlin	731	# generate name of config key for indexer (strip everything after -)
433			my $indexer_config = $use_indexer;
434			$indexer_config =~ s/^(\w+)-?.*$/$1/g if ($indexer_config);
435	dpavlin	210
436	dpavlin	731	my $lookup_hash;
437			my $depends = $parser->depends($database,$input_name);
438
439			if ($depends) {
440			$log->debug("$database/$input_name depends on: ", dump($depends)) if ($depends);
441			$log->logdie("parser->depends didn't return HASH") unless (ref($depends) eq 'HASH');
442	dpavlin	213
443	dpavlin	731	foreach my $db (keys %$depends) {
444			foreach my $i (keys %{$depends->{$db}}) {
445			foreach my $k (keys %{$depends->{$db}->{$i}}) {
446			$log->debug("loading lookup $db/$i");
447			$lookup_hash->{$db}->{$i}->{$k} = $store->load_lookup(
448			database => $db,
449			input => $i,
450			key => $k,
451			);
452			}
453			}
454			}
455	dpavlin	213
456	dpavlin	731	$log->debug("lookup_hash = ", dump( $lookup_hash ));
457			}
458	dpavlin	540
459	dpavlin	290
460	dpavlin	731	foreach my $pos ( 0 ... $input_db->size ) {
461	dpavlin	674
462	dpavlin	731	my $row = $input_db->fetch \|\| next;
463	dpavlin	516
464	dpavlin	731	my $mfn = $row->{'000'}->[0];
465	dpavlin	516
466	dpavlin	731	if (! $mfn \|\| $mfn !~ m#^\d+$#) {
467			$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
468			$mfn = $pos;
469			push @{ $row->{'000'} }, $pos;
470			}
471	dpavlin	213
472	dpavlin	508
473	dpavlin	731	if ($validate) {
474			if ( my $errors = $validate->validate_errors( $row, $input_db->dump ) ) {
475			$log->error( "MFN $mfn validation error:\n",
476			$validate->report_error( $errors )
477			);
478	dpavlin	539	}
479	dpavlin	731	}
480	dpavlin	213
481	dpavlin	731	my $ds = WebPAC::Normalize::data_structure(
482			row => $row,
483			rules => $rules,
484			lookup => $lookup_hash,
485			config => create_ds_config( $db_config, $database, $input, $mfn ),
486			marc_encoding => 'utf-8',
487			load_ds_coderef => sub {
488			my ($database,$input,$mfn) = @_;
489			return $store->load_ds(
490			database => $database,
491			input => $input,
492			id => $mfn,
493			);
494			},
495			);
496	dpavlin	507
497	dpavlin	731	$store->save_ds(
498			database => $database,
499			input => $input_name,
500			id => $mfn,
501			ds => $ds,
502			) if ($ds && !$stats);
503	dpavlin	210
504	dpavlin	731	$indexer->add(
505			id => "${input_name}/${mfn}",
506			ds => $ds,
507			type => $config->get($indexer_config)->{type},
508			) if ($indexer && $ds);
509	dpavlin	539
510	dpavlin	731	if ($marc) {
511			my $i = 0;
512	dpavlin	559
513	dpavlin	731	while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) {
514			$marc->add(
515			id => $mfn . ( $i ? "/$i" : '' ),
516			fields => $fields,
517			leader => WebPAC::Normalize::marc_leader(),
518			row => $row,
519			);
520			$i++;
521	dpavlin	580	}
522
523	dpavlin	731	$log->info("Created $i instances of MFN $mfn\n") if ($i > 1);
524	dpavlin	539	}
525
526	dpavlin	731	$total_rows++;
527			}
528	dpavlin	657
529	dpavlin	731	if ($validate) {
530			my $errors = $validate->report;
531			if ($errors) {
532			$log->info("validation errors:\n$errors\n" );
533			print $report_fh "$errors\n" if ($report_fh);
534	dpavlin	671	}
535	dpavlin	731	}
536	dpavlin	539
537	dpavlin	731	if ($stats) {
538			my $s = $input_db->stats;
539			$log->info("statistics of fields usage:\n$s");
540			print $report_fh "Statistics of fields usage:\n$s" if ($report_fh);
541	dpavlin	540	}
542
543	dpavlin	731	# close MARC file
544			$marc->finish if ($marc);
545
546			# close report
547			close($report_fh) if ($report_fh)
548
549	dpavlin	539	}
550
551	dpavlin	509	eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
552	dpavlin	434
553	dpavlin	389	my $dt = time() - $start_t;
554	dpavlin	511	$log->info("$total_rows records ", $indexer ? "indexed " : "",
555			sprintf("in %.2f sec [%.2f rec/sec]",
556	dpavlin	389	$dt, ($total_rows / $dt)
557			)
558			);
559	dpavlin	255
560
561	dpavlin	606	# end forked process
562			if ($parallel) {
563			$log->info("parallel process $$ finished");
564			exit(0);
565			}
566
567	dpavlin	210	}
568
569	dpavlin	606	if ($parallel) {
570			# wait all children to finish
571			sleep(1) while wait != -1;
572			$log->info("all parallel processes finished");
573			}
574
575	dpavlin	608	#
576			# handle links or merge after indexing
577			#
578	dpavlin	607
579	dpavlin	608	if ($merge) {
580			print $estcmd_fh 'sudo /etc/init.d/hyperestraier start',$/;
581			close($estcmd_fh);
582			chmod 0700, $estcmd_path \|\| $log->warn("can't chmod 0700 $estcmd_path: $!");
583			system $estcmd_path;
584			} else {
585			foreach my $link (@links) {
586			$log->logdie("coderef in link ", Dumper($link), " is ", ref($link), " and not CODE") unless (ref($link) eq 'CODE');
587			$link->();
588			}
589	dpavlin	529	}