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 0.08;
use WebPAC::Input 0.16;
use WebPAC::Store 0.14;
use WebPAC::Normalize 0.22;
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 Pod::Usage qw/pod2usage/;

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!>

=head1 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-lint

By default turned on if normalisation file has C<marc*> directives. 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_lint = 1;
my $marc_dump = 0;
my $parallel = 0;
my $only_links = 0;
my $merge = 0;
my $help;

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-lint!" => \$marc_lint,
        "marc-dump!" => \$marc_dump,
        "parallel=i" => \$parallel,
        "only-links!" => \$only_links,
        "merge" => \$merge,
        "help" => \$help,
);

pod2usage(-verbose => 2) if ($help);

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 $log_file = 'log';

if (-e $log_file ) {    # && -s $log_file > 5 * 1024 * 1024) {
        $log->info("moved old log with ", -s $log_file, " bytes to '${log_file}.old'");
        rename $log_file, "${log_file}.old" || $log->logwarn("can't rename $log_file to ${log_file}.old: $!");
}

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;
$stats ||= $validate;
if ($stats) {
        $log->debug("disabled indexing for stats collection");
        $use_indexer = undef;
} else {
        $log->info("using $use_indexer indexing engine...");
}

# 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 && $parser->have_rules( 'search', $database )) {

                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 => $rules,
                                        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 },
                        load_row => sub {
                                my $a = shift;
                                return $store->load_row(
                                        database => $database,
                                        input => $input_name,
                                        id => $a->{id},
                                );
                        },
                        save_row => sub {
                                my $a = shift;
                                return $store->save_row(
                                        database => $database,
                                        input => $input_name,
                                        id => $a->{id},
                                        row => $a->{row},
                                );
                        },

                );

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

                if (defined( $lookup_data->{$database}->{$input_name} )) {
                        $log->debug("created following lookups: ", sub { 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 $marc;
                if ($parser->have_rules( 'marc', $database, $input_name )) {
                        $marc = new WebPAC::Output::MARC(
                                path => "out/marc/${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}}) {
                                                my $t = time();
                                                $log->debug("loading lookup $db/$i");
                                                $lookup_hash->{$db}->{$i}->{$k} = $store->load_lookup(
                                                        database => $db,
                                                        input => $i,
                                                        key => $k,
                                                );
                                                $log->debug(sprintf("lookup $db/$i took %.2fs", time() - $t));
                                        }
                                }
                        }

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


                foreach my $pos ( 0 ... $input_db->size ) {

                        my $row = $input_db->fetch || next;

                        $total_rows++;

                        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_rec( $row, $input_db->dump_ascii ) ) {
                                        $log->error( "MFN $mfn validation error:\n",
                                                $validate->report_error( $errors )
                                        );
                                }
                                next;   # validation doesn't create any output
                        }

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

                        $log->debug("ds = ", sub { dump($ds) }) if ($ds);

                        $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);
                        }
                }

                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	755	use WebPAC::Parser 0.08;
11	dpavlin	774	use WebPAC::Input 0.16;
12	dpavlin	762	use WebPAC::Store 0.14;
13	dpavlin	736	use WebPAC::Normalize 0.22;
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	790	use Pod::Usage qw/pod2usage/;
25	dpavlin	74
26	dpavlin	606	use Proc::Queue size => 1;
27			use POSIX ":sys_wait_h"; # imports WNOHANG
28
29	dpavlin	301	=head1 NAME
30	dpavlin	76
31	dpavlin	301	run.pl - start WebPAC indexing
32	dpavlin	141
33	dpavlin	301	B<this command will probably go away. Don't get used to it!>
34	dpavlin	141
35	dpavlin	790	=head1 OPTIONS
36	dpavlin	301
37			=over 4
38
39			=item --offset 42
40
41			start loading (all) databases at offset 42
42
43			=item --limit 100
44
45			limit loading to 100 records
46
47			=item --clean
48
49			remove database and Hyper Estraier index before indexing
50
51	dpavlin	510	=item --only=database_name/input_filter
52	dpavlin	335
53	dpavlin	423	reindex just single database (legacy name is --one)
54	dpavlin	335
55	dpavlin	510	C</input_filter> is optional part which can be C<name>
56			or C<type> from input
57
58	dpavlin	301	=item --config conf/config.yml
59
60			path to YAML configuration file
61
62	dpavlin	507	=item --stats
63
64	dpavlin	638	disable indexing, modify_* in configuration and dump statistics about field
65			and subfield 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	556	=item --marc-lint
72
73	dpavlin	738	By default turned on if normalisation file has C<marc*> directives. You can disable lint
74	dpavlin	558	messages with C<--no-marc-lint>.
75	dpavlin	556
76	dpavlin	559	=item --marc-dump
77
78			Force dump or input and marc record for debugging.
79
80	dpavlin	606	=item --parallel 4
81
82			Run databases in parallel (aproximatly same as number of processors in
83			machine if you want to use full load)
84
85	dpavlin	607	=item --only-links
86
87			Create just links
88
89	dpavlin	608	=item --merge
90
91			Create merged index of databases which have links
92
93	dpavlin	301	=back
94
95			=cut
96
97			my $offset;
98			my $limit;
99
100			my $clean = 0;
101	dpavlin	684	my $config_path;
102	dpavlin	301	my $debug = 0;
103	dpavlin	510	my $only_filter;
104	dpavlin	507	my $stats = 0;
105	dpavlin	516	my $validate_path;
106	dpavlin	556	my $marc_lint = 1;
107	dpavlin	559	my $marc_dump = 0;
108	dpavlin	606	my $parallel = 0;
109	dpavlin	607	my $only_links = 0;
110	dpavlin	608	my $merge = 0;
111	dpavlin	790	my $help;
112	dpavlin	606
113	dpavlin	611	my $log = _new WebPAC::Common()->_get_logger();
114
115	dpavlin	301	GetOptions(
116			"limit=i" => \$limit,
117			"offset=i" => \$offset,
118			"clean" => \$clean,
119	dpavlin	510	"one=s" => \$only_filter,
120			"only=s" => \$only_filter,
121	dpavlin	684	"config" => \$config_path,
122	dpavlin	560	"debug+" => \$debug,
123	dpavlin	507	"stats" => \$stats,
124	dpavlin	516	"validate=s" => \$validate_path,
125	dpavlin	556	"marc-lint!" => \$marc_lint,
126	dpavlin	559	"marc-dump!" => \$marc_dump,
127	dpavlin	606	"parallel=i" => \$parallel,
128	dpavlin	607	"only-links!" => \$only_links,
129	dpavlin	608	"merge" => \$merge,
130	dpavlin	790	"help" => \$help,
131	dpavlin	301	);
132
133	dpavlin	790	pod2usage(-verbose => 2) if ($help);
134
135	dpavlin	684	my $config = new WebPAC::Config( path => $config_path );
136	dpavlin	301
137	dpavlin	611	#print "config = ",dump($config) if ($debug);
138	dpavlin	301
139	dpavlin	684	die "no databases in config file!\n" unless ($config->databases);
140	dpavlin	431
141	dpavlin	509	$log->info( "-" x 79 );
142	dpavlin	431
143	dpavlin	775	my $log_file = 'log';
144	dpavlin	608
145	dpavlin	775	if (-e $log_file ) { # && -s $log_file > 5 * 1024 * 1024) {
146			$log->info("moved old log with ", -s $log_file, " bytes to '${log_file}.old'");
147			rename $log_file, "${log_file}.old" \|\| $log->logwarn("can't rename $log_file to ${log_file}.old: $!");
148			}
149
150	dpavlin	608	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	769	$stats \|\|= $validate;
168	dpavlin	509	if ($stats) {
169	dpavlin	769	$log->debug("disabled indexing for stats collection");
170	dpavlin	509	$use_indexer = undef;
171			} else {
172			$log->info("using $use_indexer indexing engine...");
173			}
174	dpavlin	141
175	dpavlin	698	# parse normalize files and create source files for lookup and normalization
176
177			my $parser = new WebPAC::Parser( config => $config );
178
179	dpavlin	213	my $total_rows = 0;
180	dpavlin	389	my $start_t = time();
181	dpavlin	213
182	dpavlin	529	my @links;
183
184	dpavlin	606	if ($parallel) {
185			$log->info("Using $parallel processes for speedup");
186			Proc::Queue::size($parallel);
187			}
188
189	dpavlin	707	sub create_ds_config {
190			my ($db_config, $database, $input, $mfn) = @_;
191			my $c = dclone( $db_config );
192			$c->{_} = $database \|\| $log->logconfess("need database");
193			$c->{_mfn} = $mfn \|\| $log->logconfess("need mfn");
194			$c->{input} = $input \|\| $log->logconfess("need input");
195			return $c;
196			}
197
198	dpavlin	684	while (my ($database, $db_config) = each %{ $config->databases }) {
199	dpavlin	141
200	dpavlin	512	my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter);
201	dpavlin	510	next if ($only_database && $database !~ m/$only_database/i);
202	dpavlin	335
203	dpavlin	606	if ($parallel) {
204			my $f=fork;
205			if(defined ($f) and $f==0) {
206			$log->info("Created processes $$ for speedup");
207			} else {
208			next;
209			}
210			}
211
212	dpavlin	607	my $indexer;
213	dpavlin	755	if ($use_indexer && $parser->have_rules( 'search', $database )) {
214	dpavlin	627
215			my $cfg_name = $use_indexer;
216			$cfg_name =~ s/\-.*$//;
217
218	dpavlin	684	my $indexer_config = $config->get( $cfg_name ) \|\| $log->logdie("can't find '$cfg_name' part in confguration");
219	dpavlin	509	$indexer_config->{database} = $database;
220			$indexer_config->{clean} = $clean;
221			$indexer_config->{label} = $db_config->{name};
222	dpavlin	431
223	dpavlin	608	# force clean if database has links
224			$indexer_config->{clean} = 1 if ($db_config->{links});
225
226	dpavlin	509	if ($use_indexer eq 'hyperestraier') {
227	dpavlin	255
228	dpavlin	509	# open Hyper Estraier database
229			use WebPAC::Output::Estraier '0.10';
230			$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
231
232	dpavlin	627	} elsif ($use_indexer eq 'hyperestraier-native') {
233
234			# open Hyper Estraier database
235			use WebPAC::Output::EstraierNative;
236			$indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } );
237
238	dpavlin	509	} elsif ($use_indexer eq 'kinosearch') {
239	dpavlin	430
240	dpavlin	509	# open KinoSearch
241			use WebPAC::Output::KinoSearch;
242			$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path});
243			$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } );
244	dpavlin	431
245	dpavlin	509	} else {
246			$log->logdie("unknown use_indexer: $use_indexer");
247			}
248
249			$log->logide("can't continue without valid indexer") unless ($indexer);
250	dpavlin	430	}
251
252
253	dpavlin	255	#
254	dpavlin	607	# store Hyper Estraier links to other databases
255			#
256			if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) {
257			foreach my $link (@{ $db_config->{links} }) {
258			if ($use_indexer eq 'hyperestraier') {
259	dpavlin	608	if ($merge) {
260			print $estcmd_fh 'sudo -u www-data estcmd merge ' . $database . ' ' . $link->{to},$/;
261			} else {
262			$log->info("saving link $database -> $link->{to} [$link->{credit}]");
263			push @links, sub {
264			$log->info("adding link $database -> $link->{to} [$link->{credit}]");
265			$indexer->add_link(
266			from => $database,
267			to => $link->{to},
268			credit => $link->{credit},
269			);
270			};
271			}
272	dpavlin	607	} else {
273			$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
274			}
275			}
276			}
277			next if ($only_links);
278
279
280			#
281	dpavlin	255	# now WebPAC::Store
282			#
283	dpavlin	210	my $abs_path = abs_path($0);
284			$abs_path =~ s#/[^/]*$#/#;
285	dpavlin	141
286	dpavlin	713	my $db_path = $config->webpac('db_path');
287	dpavlin	74
288	dpavlin	301	if ($clean) {
289	dpavlin	511	$log->info("creating new database '$database' in $db_path");
290	dpavlin	301	rmtree( $db_path ) \|\| $log->warn("can't remove $db_path: $!");
291			} else {
292	dpavlin	511	$log->info("working on database '$database' in $db_path");
293	dpavlin	301	}
294	dpavlin	255
295	dpavlin	710	my $store = new WebPAC::Store(
296	dpavlin	210	path => $db_path,
297	dpavlin	301	debug => $debug,
298	dpavlin	210	);
299	dpavlin	74
300	dpavlin	233
301	dpavlin	213	#
302			# now, iterate through input formats
303			#
304	dpavlin	74
305	dpavlin	213	my @inputs;
306			if (ref($db_config->{input}) eq 'ARRAY') {
307			@inputs = @{ $db_config->{input} };
308	dpavlin	255	} elsif ($db_config->{input}) {
309			push @inputs, $db_config->{input};
310	dpavlin	213	} else {
311	dpavlin	255	$log->info("database $database doesn't have inputs defined");
312	dpavlin	213	}
313	dpavlin	74
314	dpavlin	213	foreach my $input (@inputs) {
315	dpavlin	233
316	dpavlin	710	my $input_name = $input->{name} \|\| $log->logdie("input without a name isn't valid: ",dump($input));
317	dpavlin	510
318	dpavlin	710	next if ($only_input && ($input_name !~ m#$only_input#i && $input->{type} !~ m#$only_input#i));
319
320	dpavlin	233	my $type = lc($input->{type});
321
322	dpavlin	684	die "I know only how to handle input types ", join(",", $config->webpac('inputs') ), " not '$type'!\n" unless (grep(/$type/, $config->webpac('inputs')));
323	dpavlin	233
324	dpavlin	684	my $input_module = $config->webpac('inputs')->{$type};
325	dpavlin	74
326	dpavlin	706	my @lookups = $parser->have_lookup_create($database, $input);
327
328	dpavlin	710	$log->info("working on input '$input_name' in $input->{path} [type: $input->{type}] using $input_module",
329	dpavlin	707	@lookups ? " creating lookups: ".join(", ", @lookups) : ""
330	dpavlin	523	);
331	dpavlin	286
332	dpavlin	638	if ($stats) {
333			# disable modification of records if --stats is in use
334			delete($input->{modify_records});
335			delete($input->{modify_file});
336			}
337
338	dpavlin	287	my $input_db = new WebPAC::Input(
339			module => $input_module,
340	dpavlin	684	encoding => $config->webpac('webpac_encoding'),
341	dpavlin	301	limit => $limit \|\| $input->{limit},
342			offset => $offset,
343	dpavlin	416	recode => $input->{recode},
344	dpavlin	507	stats => $stats,
345	dpavlin	600	modify_records => $input->{modify_records},
346	dpavlin	636	modify_file => $input->{modify_file},
347	dpavlin	287	);
348	dpavlin	286	$log->logdie("can't create input using $input_module") unless ($input);
349	dpavlin	113
350	dpavlin	702	if (defined( $input->{lookup} )) {
351	dpavlin	710	$log->warn("$database/$input_name has depriciated lookup definition, removing it...");
352	dpavlin	702	delete( $input->{lookup} );
353			}
354
355	dpavlin	707	my $lookup_coderef;
356
357			if (@lookups) {
358
359			my $rules = $parser->lookup_create_rules($database, $input) \|\| $log->logdie("no rules found for $database/$input");
360
361			$lookup_coderef = sub {
362			my $rec = shift \|\| die "need rec!";
363			my $mfn = $rec->{'000'}->[0] \|\| die "need mfn in 000";
364
365			WebPAC::Normalize::data_structure(
366			row => $rec,
367	dpavlin	732	rules => $rules,
368	dpavlin	707	config => create_ds_config( $db_config, $database, $input, $mfn ),
369			);
370
371	dpavlin	732	#warn "current lookup: ", dump(WebPAC::Normalize::_get_lookup());
372	dpavlin	707	};
373
374			WebPAC::Normalize::_set_lookup( undef );
375
376			$log->debug("created lookup_coderef using:\n$rules");
377
378			};
379
380	dpavlin	731	my $lookup_jar;
381
382	dpavlin	287	my $maxmfn = $input_db->open(
383	dpavlin	285	path => $input->{path},
384	dpavlin	213	code_page => $input->{encoding}, # database encoding
385	dpavlin	707	lookup_coderef => $lookup_coderef,
386	dpavlin	731	lookup => $lookup_jar,
387	dpavlin	523	%{ $input },
388	dpavlin	762	load_row => sub {
389			my $a = shift;
390			return $store->load_row(
391			database => $database,
392			input => $input_name,
393			id => $a->{id},
394			);
395			},
396			save_row => sub {
397			my $a = shift;
398			return $store->save_row(
399			database => $database,
400			input => $input_name,
401			id => $a->{id},
402			row => $a->{row},
403			);
404			},
405
406	dpavlin	213	);
407	dpavlin	113
408	dpavlin	710	my $lookup_data = WebPAC::Normalize::_get_lookup();
409	dpavlin	707
410	dpavlin	712	if (defined( $lookup_data->{$database}->{$input_name} )) {
411	dpavlin	765	$log->debug("created following lookups: ", sub { dump( $lookup_data ) } );
412	dpavlin	710
413	dpavlin	712	foreach my $key (keys %{ $lookup_data->{$database}->{$input_name} }) {
414	dpavlin	713	$store->save_lookup(
415			database => $database,
416			input => $input_name,
417			key => $key,
418			data => $lookup_data->{$database}->{$input_name}->{$key},
419			);
420	dpavlin	712	}
421	dpavlin	710	}
422
423	dpavlin	671	my $report_fh;
424			if ($stats \|\| $validate) {
425	dpavlin	710	my $path = "out/report/${database}-${input_name}.txt";
426	dpavlin	671	open($report_fh, '>', $path) \|\| $log->logdie("can't open $path: $!");
427
428	dpavlin	710	print $report_fh "Report for database '$database' input '$input_name' records ",
429	dpavlin	671	$offset \|\| 1, "-", $limit \|\| $input->{limit} \|\| $maxmfn, "\n\n";
430			$log->info("Generating report file $path");
431			}
432
433	dpavlin	731	my $marc;
434	dpavlin	755	if ($parser->have_rules( 'marc', $database, $input_name )) {
435	dpavlin	731	$marc = new WebPAC::Output::MARC(
436	dpavlin	738	path => "out/marc/${database}-${input_name}.marc",
437	dpavlin	731	lint => $marc_lint,
438			dump => $marc_dump,
439			);
440	dpavlin	552	}
441
442	dpavlin	731	my $rules = $parser->normalize_rules($database,$input_name) \|\| $log->logdie("no normalize rules found for $database/$input_name");
443			$log->debug("parsed normalize rules:\n$rules");
444	dpavlin	492
445	dpavlin	731	# reset position in database
446			$input_db->seek(1);
447	dpavlin	269
448	dpavlin	731	# generate name of config key for indexer (strip everything after -)
449			my $indexer_config = $use_indexer;
450			$indexer_config =~ s/^(\w+)-?.*$/$1/g if ($indexer_config);
451	dpavlin	210
452	dpavlin	731	my $lookup_hash;
453			my $depends = $parser->depends($database,$input_name);
454
455			if ($depends) {
456			$log->debug("$database/$input_name depends on: ", dump($depends)) if ($depends);
457			$log->logdie("parser->depends didn't return HASH") unless (ref($depends) eq 'HASH');
458	dpavlin	213
459	dpavlin	731	foreach my $db (keys %$depends) {
460			foreach my $i (keys %{$depends->{$db}}) {
461			foreach my $k (keys %{$depends->{$db}->{$i}}) {
462	dpavlin	762	my $t = time();
463	dpavlin	731	$log->debug("loading lookup $db/$i");
464			$lookup_hash->{$db}->{$i}->{$k} = $store->load_lookup(
465			database => $db,
466			input => $i,
467			key => $k,
468			);
469	dpavlin	762	$log->debug(sprintf("lookup $db/$i took %.2fs", time() - $t));
470	dpavlin	731	}
471			}
472			}
473	dpavlin	213
474	dpavlin	762	$log->debug("lookup_hash = ", sub { dump( $lookup_hash ) });
475	dpavlin	731	}
476	dpavlin	540
477	dpavlin	290
478	dpavlin	731	foreach my $pos ( 0 ... $input_db->size ) {
479	dpavlin	674
480	dpavlin	731	my $row = $input_db->fetch \|\| next;
481	dpavlin	516
482	dpavlin	769	$total_rows++;
483
484	dpavlin	731	my $mfn = $row->{'000'}->[0];
485	dpavlin	516
486	dpavlin	731	if (! $mfn \|\| $mfn !~ m#^\d+$#) {
487			$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
488			$mfn = $pos;
489			push @{ $row->{'000'} }, $pos;
490			}
491	dpavlin	213
492	dpavlin	508
493	dpavlin	731	if ($validate) {
494	dpavlin	774	if ( my $errors = $validate->validate_rec( $row, $input_db->dump_ascii ) ) {
495	dpavlin	731	$log->error( "MFN $mfn validation error:\n",
496			$validate->report_error( $errors )
497			);
498	dpavlin	539	}
499	dpavlin	769	next; # validation doesn't create any output
500	dpavlin	731	}
501	dpavlin	213
502	dpavlin	731	my $ds = WebPAC::Normalize::data_structure(
503			row => $row,
504			rules => $rules,
505			lookup => $lookup_hash,
506			config => create_ds_config( $db_config, $database, $input, $mfn ),
507			marc_encoding => 'utf-8',
508	dpavlin	736	load_row_coderef => sub {
509	dpavlin	731	my ($database,$input,$mfn) = @_;
510	dpavlin	736	return $store->load_row(
511	dpavlin	731	database => $database,
512			input => $input,
513			id => $mfn,
514			);
515			},
516			);
517	dpavlin	507
518	dpavlin	765	$log->debug("ds = ", sub { dump($ds) }) if ($ds);
519	dpavlin	736
520	dpavlin	731	$store->save_ds(
521			database => $database,
522			input => $input_name,
523			id => $mfn,
524			ds => $ds,
525			) if ($ds && !$stats);
526	dpavlin	210
527	dpavlin	731	$indexer->add(
528			id => "${input_name}/${mfn}",
529			ds => $ds,
530			type => $config->get($indexer_config)->{type},
531			) if ($indexer && $ds);
532	dpavlin	539
533	dpavlin	731	if ($marc) {
534			my $i = 0;
535	dpavlin	559
536	dpavlin	731	while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) {
537			$marc->add(
538			id => $mfn . ( $i ? "/$i" : '' ),
539			fields => $fields,
540			leader => WebPAC::Normalize::marc_leader(),
541			row => $row,
542			);
543			$i++;
544	dpavlin	580	}
545
546	dpavlin	731	$log->info("Created $i instances of MFN $mfn\n") if ($i > 1);
547	dpavlin	539	}
548	dpavlin	731	}
549	dpavlin	657
550	dpavlin	731	if ($validate) {
551			my $errors = $validate->report;
552			if ($errors) {
553			$log->info("validation errors:\n$errors\n" );
554			print $report_fh "$errors\n" if ($report_fh);
555	dpavlin	671	}
556	dpavlin	731	}
557	dpavlin	539
558	dpavlin	731	if ($stats) {
559			my $s = $input_db->stats;
560			$log->info("statistics of fields usage:\n$s");
561			print $report_fh "Statistics of fields usage:\n$s" if ($report_fh);
562	dpavlin	540	}
563
564	dpavlin	731	# close MARC file
565			$marc->finish if ($marc);
566
567			# close report
568			close($report_fh) if ($report_fh)
569
570	dpavlin	539	}
571
572	dpavlin	509	eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
573	dpavlin	434
574	dpavlin	389	my $dt = time() - $start_t;
575	dpavlin	511	$log->info("$total_rows records ", $indexer ? "indexed " : "",
576			sprintf("in %.2f sec [%.2f rec/sec]",
577	dpavlin	389	$dt, ($total_rows / $dt)
578			)
579			);
580	dpavlin	255
581
582	dpavlin	606	# end forked process
583			if ($parallel) {
584			$log->info("parallel process $$ finished");
585			exit(0);
586			}
587
588	dpavlin	210	}
589
590	dpavlin	606	if ($parallel) {
591			# wait all children to finish
592			sleep(1) while wait != -1;
593			$log->info("all parallel processes finished");
594			}
595
596	dpavlin	608	#
597			# handle links or merge after indexing
598			#
599	dpavlin	607
600	dpavlin	608	if ($merge) {
601			print $estcmd_fh 'sudo /etc/init.d/hyperestraier start',$/;
602			close($estcmd_fh);
603			chmod 0700, $estcmd_path \|\| $log->warn("can't chmod 0700 $estcmd_path: $!");
604			system $estcmd_path;
605			} else {
606			foreach my $link (@links) {
607			$log->logdie("coderef in link ", Dumper($link), " is ", ref($link), " and not CODE") unless (ref($link) eq 'CODE');
608			$link->();
609			}
610	dpavlin	529	}