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 0.11;
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>

You can use special variables C<$database> and $C<$input> in this parametar
like C<--validate 'conf/validate/$database-$input'> to construct filename

=item --validate-delimiters path/to/validate_delimiters_file

this option is used with C<--validate> to turn on extra validation of
delimiters. If file is non existant, it will be created on first run.

=item --marc-generate

Generate MARC file. This will automatically be on if file contains C<marc*> directives.
You can use this option as C<--no-marc-generate> to disable MARC generation.

=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 $validate_delimiters_path;
my $marc_generate = 1;
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,
        "validate-delimiters=s" => \$validate_delimiters_path,
        "marc-generate!" => \$marc_generate,
        "marc-lint!" => \$marc_lint,
        "marc-dump!" => \$marc_dump,
        "parallel=i" => \$parallel,
        "only-links!" => \$only_links,
        "merge" => \$merge,
        "help" => \$help,
);

$marc_generate = 0 if ( $validate_delimiters_path );

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,
        delimiters => $config->webpac('delimiters'),
        delimiters_path => $validate_delimiters_path,
) if ($validate_path || $validate_delimiters_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 $marc;
                if ($marc_generate && $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::_get_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);
                        }

                        print $report_fh "\nAll possible subfields/delimiter templates:\n", $validate->delimiters_templates( report => 1, current_input => 1 ), "\n\n";

                        # must be last thing that touches $validate for this input
                        $validate->reset;
                }

                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");
}

# save new delimiters if needed
$validate->save_delimiters_templates if ( $validate );

#
# 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	833	use WebPAC::Validate 0.11;
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	856	You can use special variables C<$database> and $C<$input> in this parametar
72			like C<--validate 'conf/validate/$database-$input'> to construct filename
73
74	dpavlin	837	=item --validate-delimiters path/to/validate_delimiters_file
75
76			this option is used with C<--validate> to turn on extra validation of
77			delimiters. If file is non existant, it will be created on first run.
78
79	dpavlin	803	=item --marc-generate
80
81			Generate MARC file. This will automatically be on if file contains C<marc*> directives.
82			You can use this option as C<--no-marc-generate> to disable MARC generation.
83
84	dpavlin	556	=item --marc-lint
85
86	dpavlin	738	By default turned on if normalisation file has C<marc*> directives. You can disable lint
87	dpavlin	558	messages with C<--no-marc-lint>.
88	dpavlin	556
89	dpavlin	559	=item --marc-dump
90
91			Force dump or input and marc record for debugging.
92
93	dpavlin	606	=item --parallel 4
94
95			Run databases in parallel (aproximatly same as number of processors in
96			machine if you want to use full load)
97
98	dpavlin	607	=item --only-links
99
100			Create just links
101
102	dpavlin	608	=item --merge
103
104			Create merged index of databases which have links
105
106	dpavlin	301	=back
107
108			=cut
109
110			my $offset;
111			my $limit;
112
113			my $clean = 0;
114	dpavlin	684	my $config_path;
115	dpavlin	301	my $debug = 0;
116	dpavlin	510	my $only_filter;
117	dpavlin	507	my $stats = 0;
118	dpavlin	516	my $validate_path;
119	dpavlin	837	my $validate_delimiters_path;
120	dpavlin	803	my $marc_generate = 1;
121	dpavlin	556	my $marc_lint = 1;
122	dpavlin	559	my $marc_dump = 0;
123	dpavlin	606	my $parallel = 0;
124	dpavlin	607	my $only_links = 0;
125	dpavlin	608	my $merge = 0;
126	dpavlin	790	my $help;
127	dpavlin	606
128	dpavlin	611	my $log = _new WebPAC::Common()->_get_logger();
129
130	dpavlin	301	GetOptions(
131			"limit=i" => \$limit,
132			"offset=i" => \$offset,
133			"clean" => \$clean,
134	dpavlin	510	"one=s" => \$only_filter,
135			"only=s" => \$only_filter,
136	dpavlin	684	"config" => \$config_path,
137	dpavlin	560	"debug+" => \$debug,
138	dpavlin	507	"stats" => \$stats,
139	dpavlin	516	"validate=s" => \$validate_path,
140	dpavlin	837	"validate-delimiters=s" => \$validate_delimiters_path,
141	dpavlin	803	"marc-generate!" => \$marc_generate,
142	dpavlin	556	"marc-lint!" => \$marc_lint,
143	dpavlin	559	"marc-dump!" => \$marc_dump,
144	dpavlin	606	"parallel=i" => \$parallel,
145	dpavlin	607	"only-links!" => \$only_links,
146	dpavlin	608	"merge" => \$merge,
147	dpavlin	790	"help" => \$help,
148	dpavlin	301	);
149
150	dpavlin	851	$marc_generate = 0 if ( $validate_delimiters_path );
151
152	dpavlin	790	pod2usage(-verbose => 2) if ($help);
153
154	dpavlin	684	my $config = new WebPAC::Config( path => $config_path );
155	dpavlin	301
156	dpavlin	611	#print "config = ",dump($config) if ($debug);
157	dpavlin	301
158	dpavlin	684	die "no databases in config file!\n" unless ($config->databases);
159	dpavlin	431
160	dpavlin	509	$log->info( "-" x 79 );
161	dpavlin	431
162	dpavlin	775	my $log_file = 'log';
163	dpavlin	608
164	dpavlin	775	if (-e $log_file ) { # && -s $log_file > 5 * 1024 * 1024) {
165			$log->info("moved old log with ", -s $log_file, " bytes to '${log_file}.old'");
166			rename $log_file, "${log_file}.old" \|\| $log->logwarn("can't rename $log_file to ${log_file}.old: $!");
167			}
168
169	dpavlin	608	my $estcmd_fh;
170			my $estcmd_path = './estcmd-merge.sh';
171			if ($merge) {
172			open($estcmd_fh, '>', $estcmd_path) \|\| $log->logdie("can't open $estcmd_path: $!");
173			print $estcmd_fh 'cd /data/estraier/_node/ \|\| exit 1',$/;
174			print $estcmd_fh 'sudo /etc/init.d/hyperestraier stop',$/;
175			$log->info("created merge batch file $estcmd_path");
176			}
177
178	dpavlin	516	my $validate;
179			$validate = new WebPAC::Validate(
180			path => $validate_path,
181	dpavlin	833	delimiters => $config->webpac('delimiters'),
182	dpavlin	837	delimiters_path => $validate_delimiters_path,
183			) if ($validate_path \|\| $validate_delimiters_path);
184	dpavlin	516
185	dpavlin	684	my $use_indexer = $config->use_indexer;
186	dpavlin	769	$stats \|\|= $validate;
187	dpavlin	509	if ($stats) {
188	dpavlin	769	$log->debug("disabled indexing for stats collection");
189	dpavlin	509	$use_indexer = undef;
190			} else {
191			$log->info("using $use_indexer indexing engine...");
192			}
193	dpavlin	141
194	dpavlin	698	# parse normalize files and create source files for lookup and normalization
195
196			my $parser = new WebPAC::Parser( config => $config );
197
198	dpavlin	213	my $total_rows = 0;
199	dpavlin	389	my $start_t = time();
200	dpavlin	213
201	dpavlin	529	my @links;
202
203	dpavlin	606	if ($parallel) {
204			$log->info("Using $parallel processes for speedup");
205			Proc::Queue::size($parallel);
206			}
207
208	dpavlin	707	sub create_ds_config {
209			my ($db_config, $database, $input, $mfn) = @_;
210			my $c = dclone( $db_config );
211			$c->{_} = $database \|\| $log->logconfess("need database");
212			$c->{_mfn} = $mfn \|\| $log->logconfess("need mfn");
213			$c->{input} = $input \|\| $log->logconfess("need input");
214			return $c;
215			}
216
217	dpavlin	684	while (my ($database, $db_config) = each %{ $config->databases }) {
218	dpavlin	141
219	dpavlin	512	my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter);
220	dpavlin	510	next if ($only_database && $database !~ m/$only_database/i);
221	dpavlin	335
222	dpavlin	606	if ($parallel) {
223			my $f=fork;
224			if(defined ($f) and $f==0) {
225			$log->info("Created processes $$ for speedup");
226			} else {
227			next;
228			}
229			}
230
231	dpavlin	607	my $indexer;
232	dpavlin	755	if ($use_indexer && $parser->have_rules( 'search', $database )) {
233	dpavlin	627
234			my $cfg_name = $use_indexer;
235			$cfg_name =~ s/\-.*$//;
236
237	dpavlin	684	my $indexer_config = $config->get( $cfg_name ) \|\| $log->logdie("can't find '$cfg_name' part in confguration");
238	dpavlin	509	$indexer_config->{database} = $database;
239			$indexer_config->{clean} = $clean;
240			$indexer_config->{label} = $db_config->{name};
241	dpavlin	431
242	dpavlin	608	# force clean if database has links
243			$indexer_config->{clean} = 1 if ($db_config->{links});
244
245	dpavlin	509	if ($use_indexer eq 'hyperestraier') {
246	dpavlin	255
247	dpavlin	509	# open Hyper Estraier database
248			use WebPAC::Output::Estraier '0.10';
249			$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
250
251	dpavlin	627	} elsif ($use_indexer eq 'hyperestraier-native') {
252
253			# open Hyper Estraier database
254			use WebPAC::Output::EstraierNative;
255			$indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } );
256
257	dpavlin	509	} elsif ($use_indexer eq 'kinosearch') {
258	dpavlin	430
259	dpavlin	509	# open KinoSearch
260			use WebPAC::Output::KinoSearch;
261			$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path});
262			$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } );
263	dpavlin	431
264	dpavlin	509	} else {
265			$log->logdie("unknown use_indexer: $use_indexer");
266			}
267
268			$log->logide("can't continue without valid indexer") unless ($indexer);
269	dpavlin	430	}
270
271
272	dpavlin	255	#
273	dpavlin	607	# store Hyper Estraier links to other databases
274			#
275			if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) {
276			foreach my $link (@{ $db_config->{links} }) {
277			if ($use_indexer eq 'hyperestraier') {
278	dpavlin	608	if ($merge) {
279			print $estcmd_fh 'sudo -u www-data estcmd merge ' . $database . ' ' . $link->{to},$/;
280			} else {
281			$log->info("saving link $database -> $link->{to} [$link->{credit}]");
282			push @links, sub {
283			$log->info("adding link $database -> $link->{to} [$link->{credit}]");
284			$indexer->add_link(
285			from => $database,
286			to => $link->{to},
287			credit => $link->{credit},
288			);
289			};
290			}
291	dpavlin	607	} else {
292			$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
293			}
294			}
295			}
296			next if ($only_links);
297
298
299			#
300	dpavlin	255	# now WebPAC::Store
301			#
302	dpavlin	210	my $abs_path = abs_path($0);
303			$abs_path =~ s#/[^/]*$#/#;
304	dpavlin	141
305	dpavlin	713	my $db_path = $config->webpac('db_path');
306	dpavlin	74
307	dpavlin	301	if ($clean) {
308	dpavlin	511	$log->info("creating new database '$database' in $db_path");
309	dpavlin	301	rmtree( $db_path ) \|\| $log->warn("can't remove $db_path: $!");
310			} else {
311	dpavlin	511	$log->info("working on database '$database' in $db_path");
312	dpavlin	301	}
313	dpavlin	255
314	dpavlin	710	my $store = new WebPAC::Store(
315	dpavlin	210	path => $db_path,
316	dpavlin	301	debug => $debug,
317	dpavlin	210	);
318	dpavlin	74
319	dpavlin	233
320	dpavlin	213	#
321			# now, iterate through input formats
322			#
323	dpavlin	74
324	dpavlin	213	my @inputs;
325			if (ref($db_config->{input}) eq 'ARRAY') {
326			@inputs = @{ $db_config->{input} };
327	dpavlin	255	} elsif ($db_config->{input}) {
328			push @inputs, $db_config->{input};
329	dpavlin	213	} else {
330	dpavlin	255	$log->info("database $database doesn't have inputs defined");
331	dpavlin	213	}
332	dpavlin	74
333	dpavlin	213	foreach my $input (@inputs) {
334	dpavlin	233
335	dpavlin	710	my $input_name = $input->{name} \|\| $log->logdie("input without a name isn't valid: ",dump($input));
336	dpavlin	510
337	dpavlin	710	next if ($only_input && ($input_name !~ m#$only_input#i && $input->{type} !~ m#$only_input#i));
338
339	dpavlin	233	my $type = lc($input->{type});
340
341	dpavlin	684	die "I know only how to handle input types ", join(",", $config->webpac('inputs') ), " not '$type'!\n" unless (grep(/$type/, $config->webpac('inputs')));
342	dpavlin	233
343	dpavlin	684	my $input_module = $config->webpac('inputs')->{$type};
344	dpavlin	74
345	dpavlin	706	my @lookups = $parser->have_lookup_create($database, $input);
346
347	dpavlin	710	$log->info("working on input '$input_name' in $input->{path} [type: $input->{type}] using $input_module",
348	dpavlin	707	@lookups ? " creating lookups: ".join(", ", @lookups) : ""
349	dpavlin	523	);
350	dpavlin	286
351	dpavlin	638	if ($stats) {
352			# disable modification of records if --stats is in use
353			delete($input->{modify_records});
354			delete($input->{modify_file});
355			}
356
357	dpavlin	287	my $input_db = new WebPAC::Input(
358			module => $input_module,
359	dpavlin	684	encoding => $config->webpac('webpac_encoding'),
360	dpavlin	301	limit => $limit \|\| $input->{limit},
361			offset => $offset,
362	dpavlin	416	recode => $input->{recode},
363	dpavlin	507	stats => $stats,
364	dpavlin	600	modify_records => $input->{modify_records},
365	dpavlin	636	modify_file => $input->{modify_file},
366	dpavlin	287	);
367	dpavlin	286	$log->logdie("can't create input using $input_module") unless ($input);
368	dpavlin	113
369	dpavlin	702	if (defined( $input->{lookup} )) {
370	dpavlin	710	$log->warn("$database/$input_name has depriciated lookup definition, removing it...");
371	dpavlin	702	delete( $input->{lookup} );
372			}
373
374	dpavlin	707	my $lookup_coderef;
375
376			if (@lookups) {
377
378			my $rules = $parser->lookup_create_rules($database, $input) \|\| $log->logdie("no rules found for $database/$input");
379
380			$lookup_coderef = sub {
381			my $rec = shift \|\| die "need rec!";
382			my $mfn = $rec->{'000'}->[0] \|\| die "need mfn in 000";
383
384			WebPAC::Normalize::data_structure(
385			row => $rec,
386	dpavlin	732	rules => $rules,
387	dpavlin	707	config => create_ds_config( $db_config, $database, $input, $mfn ),
388			);
389
390	dpavlin	732	#warn "current lookup: ", dump(WebPAC::Normalize::_get_lookup());
391	dpavlin	707	};
392
393			WebPAC::Normalize::_set_lookup( undef );
394
395			$log->debug("created lookup_coderef using:\n$rules");
396
397			};
398
399	dpavlin	731	my $lookup_jar;
400
401	dpavlin	287	my $maxmfn = $input_db->open(
402	dpavlin	285	path => $input->{path},
403	dpavlin	213	code_page => $input->{encoding}, # database encoding
404	dpavlin	707	lookup_coderef => $lookup_coderef,
405	dpavlin	731	lookup => $lookup_jar,
406	dpavlin	523	%{ $input },
407	dpavlin	762	load_row => sub {
408			my $a = shift;
409			return $store->load_row(
410			database => $database,
411			input => $input_name,
412			id => $a->{id},
413			);
414			},
415			save_row => sub {
416			my $a = shift;
417			return $store->save_row(
418			database => $database,
419			input => $input_name,
420			id => $a->{id},
421			row => $a->{row},
422			);
423			},
424
425	dpavlin	213	);
426	dpavlin	113
427	dpavlin	710	my $lookup_data = WebPAC::Normalize::_get_lookup();
428	dpavlin	707
429	dpavlin	712	if (defined( $lookup_data->{$database}->{$input_name} )) {
430	dpavlin	765	$log->debug("created following lookups: ", sub { dump( $lookup_data ) } );
431	dpavlin	710
432	dpavlin	712	foreach my $key (keys %{ $lookup_data->{$database}->{$input_name} }) {
433	dpavlin	713	$store->save_lookup(
434			database => $database,
435			input => $input_name,
436			key => $key,
437			data => $lookup_data->{$database}->{$input_name}->{$key},
438			);
439	dpavlin	712	}
440	dpavlin	710	}
441
442	dpavlin	671	my $report_fh;
443			if ($stats \|\| $validate) {
444	dpavlin	710	my $path = "out/report/${database}-${input_name}.txt";
445	dpavlin	671	open($report_fh, '>', $path) \|\| $log->logdie("can't open $path: $!");
446
447	dpavlin	710	print $report_fh "Report for database '$database' input '$input_name' records ",
448	dpavlin	671	$offset \|\| 1, "-", $limit \|\| $input->{limit} \|\| $maxmfn, "\n\n";
449			$log->info("Generating report file $path");
450			}
451
452	dpavlin	731	my $marc;
453	dpavlin	803	if ($marc_generate && $parser->have_rules( 'marc', $database, $input_name )) {
454	dpavlin	731	$marc = new WebPAC::Output::MARC(
455	dpavlin	738	path => "out/marc/${database}-${input_name}.marc",
456	dpavlin	731	lint => $marc_lint,
457			dump => $marc_dump,
458			);
459	dpavlin	552	}
460
461	dpavlin	731	my $rules = $parser->normalize_rules($database,$input_name) \|\| $log->logdie("no normalize rules found for $database/$input_name");
462			$log->debug("parsed normalize rules:\n$rules");
463	dpavlin	492
464	dpavlin	731	# reset position in database
465			$input_db->seek(1);
466	dpavlin	269
467	dpavlin	731	# generate name of config key for indexer (strip everything after -)
468			my $indexer_config = $use_indexer;
469			$indexer_config =~ s/^(\w+)-?.*$/$1/g if ($indexer_config);
470	dpavlin	210
471	dpavlin	731	my $lookup_hash;
472			my $depends = $parser->depends($database,$input_name);
473
474			if ($depends) {
475			$log->debug("$database/$input_name depends on: ", dump($depends)) if ($depends);
476			$log->logdie("parser->depends didn't return HASH") unless (ref($depends) eq 'HASH');
477	dpavlin	213
478	dpavlin	731	foreach my $db (keys %$depends) {
479			foreach my $i (keys %{$depends->{$db}}) {
480			foreach my $k (keys %{$depends->{$db}->{$i}}) {
481	dpavlin	762	my $t = time();
482	dpavlin	731	$log->debug("loading lookup $db/$i");
483			$lookup_hash->{$db}->{$i}->{$k} = $store->load_lookup(
484			database => $db,
485			input => $i,
486			key => $k,
487			);
488	dpavlin	762	$log->debug(sprintf("lookup $db/$i took %.2fs", time() - $t));
489	dpavlin	731	}
490			}
491			}
492	dpavlin	213
493	dpavlin	762	$log->debug("lookup_hash = ", sub { dump( $lookup_hash ) });
494	dpavlin	731	}
495	dpavlin	540
496	dpavlin	290
497	dpavlin	731	foreach my $pos ( 0 ... $input_db->size ) {
498	dpavlin	674
499	dpavlin	731	my $row = $input_db->fetch \|\| next;
500	dpavlin	516
501	dpavlin	769	$total_rows++;
502
503	dpavlin	731	my $mfn = $row->{'000'}->[0];
504	dpavlin	516
505	dpavlin	731	if (! $mfn \|\| $mfn !~ m#^\d+$#) {
506			$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
507			$mfn = $pos;
508			push @{ $row->{'000'} }, $pos;
509			}
510	dpavlin	213
511	dpavlin	508
512	dpavlin	731	if ($validate) {
513	dpavlin	774	if ( my $errors = $validate->validate_rec( $row, $input_db->dump_ascii ) ) {
514	dpavlin	731	$log->error( "MFN $mfn validation error:\n",
515			$validate->report_error( $errors )
516			);
517	dpavlin	539	}
518	dpavlin	769	next; # validation doesn't create any output
519	dpavlin	731	}
520	dpavlin	213
521	dpavlin	731	my $ds = WebPAC::Normalize::data_structure(
522			row => $row,
523			rules => $rules,
524			lookup => $lookup_hash,
525			config => create_ds_config( $db_config, $database, $input, $mfn ),
526			marc_encoding => 'utf-8',
527	dpavlin	736	load_row_coderef => sub {
528	dpavlin	731	my ($database,$input,$mfn) = @_;
529	dpavlin	736	return $store->load_row(
530	dpavlin	731	database => $database,
531			input => $input,
532			id => $mfn,
533			);
534			},
535			);
536	dpavlin	507
537	dpavlin	765	$log->debug("ds = ", sub { dump($ds) }) if ($ds);
538	dpavlin	736
539	dpavlin	731	$store->save_ds(
540			database => $database,
541			input => $input_name,
542			id => $mfn,
543			ds => $ds,
544			) if ($ds && !$stats);
545	dpavlin	210
546	dpavlin	731	$indexer->add(
547			id => "${input_name}/${mfn}",
548			ds => $ds,
549			type => $config->get($indexer_config)->{type},
550			) if ($indexer && $ds);
551	dpavlin	539
552	dpavlin	731	if ($marc) {
553			my $i = 0;
554	dpavlin	559
555	dpavlin	731	while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) {
556			$marc->add(
557			id => $mfn . ( $i ? "/$i" : '' ),
558			fields => $fields,
559	dpavlin	814	leader => WebPAC::Normalize::_get_marc_leader(),
560	dpavlin	731	row => $row,
561			);
562			$i++;
563	dpavlin	580	}
564
565	dpavlin	731	$log->info("Created $i instances of MFN $mfn\n") if ($i > 1);
566	dpavlin	539	}
567	dpavlin	731	}
568	dpavlin	657
569	dpavlin	731	if ($validate) {
570			my $errors = $validate->report;
571			if ($errors) {
572			$log->info("validation errors:\n$errors\n" );
573			print $report_fh "$errors\n" if ($report_fh);
574	dpavlin	671	}
575	dpavlin	833
576	dpavlin	851	print $report_fh "\nAll possible subfields/delimiter templates:\n", $validate->delimiters_templates( report => 1, current_input => 1 ), "\n\n";
577
578			# must be last thing that touches $validate for this input
579			$validate->reset;
580	dpavlin	731	}
581	dpavlin	539
582	dpavlin	731	if ($stats) {
583			my $s = $input_db->stats;
584			$log->info("statistics of fields usage:\n$s");
585			print $report_fh "Statistics of fields usage:\n$s" if ($report_fh);
586	dpavlin	540	}
587
588	dpavlin	731	# close MARC file
589			$marc->finish if ($marc);
590
591			# close report
592	dpavlin	850	close($report_fh) if ($report_fh);
593	dpavlin	539	}
594
595	dpavlin	509	eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
596	dpavlin	434
597	dpavlin	389	my $dt = time() - $start_t;
598	dpavlin	511	$log->info("$total_rows records ", $indexer ? "indexed " : "",
599			sprintf("in %.2f sec [%.2f rec/sec]",
600	dpavlin	389	$dt, ($total_rows / $dt)
601			)
602			);
603	dpavlin	255
604
605	dpavlin	606	# end forked process
606			if ($parallel) {
607			$log->info("parallel process $$ finished");
608			exit(0);
609			}
610
611	dpavlin	210	}
612
613	dpavlin	606	if ($parallel) {
614			# wait all children to finish
615			sleep(1) while wait != -1;
616			$log->info("all parallel processes finished");
617			}
618
619	dpavlin	837	# save new delimiters if needed
620	dpavlin	845	$validate->save_delimiters_templates if ( $validate );
621	dpavlin	837
622	dpavlin	608	#
623			# handle links or merge after indexing
624			#
625	dpavlin	607
626	dpavlin	608	if ($merge) {
627			print $estcmd_fh 'sudo /etc/init.d/hyperestraier start',$/;
628			close($estcmd_fh);
629			chmod 0700, $estcmd_path \|\| $log->warn("can't chmod 0700 $estcmd_path: $!");
630			system $estcmd_path;
631			} else {
632			foreach my $link (@links) {
633			$log->logdie("coderef in link ", Dumper($link), " is ", ref($link), " and not CODE") unless (ref($link) eq 'CODE');
634			$link->();
635			}
636	dpavlin	529	}