use WebPAC::Common 0.02;
use WebPAC::Parser 0.08;
use WebPAC::Input 0.16;
-use WebPAC::Store 0.14;
+use WebPAC::Store 0.15;
use WebPAC::Normalize 0.22;
use WebPAC::Output::TT;
use WebPAC::Validate 0.11;
use Storable qw/dclone/;
use Pod::Usage qw/pod2usage/;
-use Proc::Queue size => 1;
use POSIX ":sys_wait_h"; # imports WNOHANG
=head1 NAME
"clean" => \$clean,
"one=s" => \$only_filter,
"only=s" => \$only_filter,
- "config" => \$config_path,
+ "config=s" => \$config_path,
"debug+" => \$debug,
"stats" => \$stats,
"validate=s" => \$validate_path,
my $config = new WebPAC::Config( path => $config_path );
+WebPAC::Normalize::_debug( $debug - 1 ) if $debug > 1;
+
#print "config = ",dump($config) if ($debug);
die "no databases in config file!\n" unless ($config->databases);
if ($stats) {
$log->debug("disabled indexing for stats collection");
$use_indexer = undef;
-} else {
+} elsif ( $use_indexer ) {
$log->info("using $use_indexer indexing engine...");
}
# parse normalize files and create source files for lookup and normalization
-my $parser = new WebPAC::Parser( config => $config );
+my ($only_database,$only_input) = split(m#/#, $only_filter) if $only_filter;
+
+my $parser = new WebPAC::Parser(
+ config => $config,
+ only_database => $only_database,
+ only_input => $only_input,
+);
my $total_rows = 0;
my $start_t = time();
my @links;
if ($parallel) {
+ eval 'use Proc::Queue size => 1;';
+ die $@ if $@;
$log->info("Using $parallel processes for speedup");
Proc::Queue::size($parallel);
}
foreach my $database ( sort keys %{ $config->databases } ) {
my $db_config = $config->databases->{$database};
- my ($only_database,$only_input) = split(m#/#, $only_filter) if ($only_filter);
next if ($only_database && $database !~ m/$only_database/i);
if ($parallel) {
if ($use_indexer eq 'hyperestraier') {
# open Hyper Estraier database
- use WebPAC::Output::Estraier '0.10';
+ require WebPAC::Output::Estraier;
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
} elsif ($use_indexer eq 'hyperestraier-native') {
# open Hyper Estraier database
- use WebPAC::Output::EstraierNative;
+ require 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 } );
+ die "no longer supported";
} else {
$log->logdie("unknown use_indexer: $use_indexer");
}
- $log->logide("can't continue without valid indexer") unless ($indexer);
+ $log->logdie("can't continue without valid indexer") unless ($indexer);
}
#
# now WebPAC::Store
#
- my $abs_path = abs_path($0);
- $abs_path =~ s#/[^/]*$#/#;
+ my $store = new WebPAC::Store({
+ debug => $debug,
+ });
- 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");
- }
+ #
+ # prepare output
+ #
+ my @outputs = force_array( $db_config->{output}, sub {
+ $log->error("Database $database doesn't have any outputs defined. Do you want to remove it from configuration?" );
+ } );
- my $store = new WebPAC::Store(
- path => $db_path,
- debug => $debug,
- );
+ my @output_modules;
+
+ foreach my $output ( @outputs ) {
+
+#warn '## output = ',dump( $output );
+
+ my $module = $output->{module} || $log->logdie("need module in output section of $database");
+ $module = 'WebPAC::Output::' . $module unless $module =~ m/::/;
+
+ $log->debug("loading output module $module");
+ eval "require $module";
+
+ # add database to arugemnts for output filter
+ $output->{database} = $database;
+ $output->{clean} = $clean;
+
+ $log->debug("calling $module->new(",dump( $output ),")");
+ my $out = new $module->new( $output );
+ if ( $out->init ) {
+ push @output_modules, $out;
+ } else {
+ $log->warn("SKIPPED $module");
+ }
+ }
#
# now, iterate through input formats
#
- my @inputs;
- if (ref($db_config->{input}) eq 'ARRAY') {
- @inputs = @{ $db_config->{input} };
- } elsif ($db_config->{input}) {
- push @inputs, $db_config->{input};
- } else {
+
+ my @inputs = force_array( $db_config->{input}, sub {
$log->info("database $database doesn't have inputs defined");
- }
+ } );
+
+ if ( -e 'out/debug' ) { # FIXME flag?
+ my $out;
+ foreach my $i ( @inputs ) {
+ warn dump( $i );
+ next unless defined $i->{normalize};
+ warn dump( $i->{normalize} );
+ foreach my $normalize ( @{ $i->{normalize} } ) {
+ my $path = $normalize->{path};
+ $out .= qq/\n##\n## $path\n##\n\n/;
+ $out .= read_file( $path );
+ }
+ }
+ my $all = "out/debug/all-normalize.pl";
+ write_file( $all, $out );
+ warn "### all normalize for this input saved to: $all";
+ };
foreach my $input (@inputs) {
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')));
+ # FIXME check if input module exists
+ my $input_module = $input->{module};
- my $input_module = $config->webpac('inputs')->{$type};
+ if ( ! $input_module ) {
+ if ( grep(/$type/, $config->webpac('inputs')) ) {
+ $input_module = $config->webpac('inputs')->{$type};
+ } else {
+ $log->logdie("I know only how to handle input types ", join(",", $config->webpac('inputs') ), " not '$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",
+ $log->info("working on $database/$input_name with $input_module on $input->{path}",
@lookups ? " creating lookups: ".join(", ", @lookups) : ""
);
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},
+ input_config => $input,
);
$log->logdie("can't create input using $input_module") unless ($input);
my $maxmfn = $input_db->open(
path => $input->{path},
- code_page => $input->{encoding}, # database encoding
+ input_encoding => $input->{encoding}, # database encoding
lookup_coderef => $lookup_coderef,
lookup => $lookup_jar,
%{ $input },
);
}
- my $rules = $parser->normalize_rules($database,$input_name) || $log->logdie("no normalize rules found for $database/$input_name");
+ my $rules = $parser->normalize_rules($database,$input_name);
+ $log->logwarn("no normalize rules for $database/$input_name") unless $rules;
+
$log->debug("parsed normalize rules:\n$rules");
# reset position in database
}
+ # setup input name for all output filters
+ foreach my $out ( @output_modules ) {
+ if ( $out->can('input') ) {
+ $out->input( $input_name );
+ } else {
+ $log->warn("output filter ",ref($out)," doesn't support input name");
+ }
+ }
+
+
foreach my $pos ( 0 ... $input_db->size ) {
my $row = $input_db->fetch || next;
my $mfn = $row->{'000'}->[0];
- if (! $mfn || $mfn !~ m#^\d+$#) {
+ if (! $mfn || $mfn !~ m{^\d+$}) {
$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
$mfn = $pos;
push @{ $row->{'000'} }, $pos;
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 $store->load_row(
+ if ($rules) {
+
+ 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) = @_;
+#warn "### load_row($database,$input,$mfn) from data_structure\n";
+ 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,
+ input => $input_name,
id => $mfn,
- );
- },
- );
+ ds => $ds,
+ ) if !$stats;
- $log->debug("ds = ", sub { dump($ds) }) if ($ds);
+ $indexer->add(
+ id => "${input_name}/${mfn}",
+ ds => $ds,
+ type => $config->get($indexer_config)->{type},
+ ) if $indexer;
- $store->save_ds(
- database => $database,
- input => $input_name,
- id => $mfn,
- ds => $ds,
- ) if ($ds && !$stats);
+ foreach my $out ( @output_modules ) {
+ $out->add( $mfn, $ds ) if $out->can('add');
+ }
- $indexer->add(
- id => "${input_name}/${mfn}",
- ds => $ds,
- type => $config->get($indexer_config)->{type},
- ) if ($indexer && $ds);
+ } else {
+ $log->warn("record $pos didn't produce any output after normalization rules!") unless $marc;
+ }
+ }
if ($marc) {
my $i = 0;
- while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) {
+ while (my $fields = WebPAC::Normalize::MARC::_get_marc_fields( fetch_next => 1 ) ) {
$marc->add(
id => $mfn . ( $i ? "/$i" : '' ),
fields => $fields,
- leader => WebPAC::Normalize::_get_marc_leader(),
+ leader => WebPAC::Normalize::MARC::_get_marc_leader(),
row => $row,
);
$i++;
$log->info("Created $i instances of MFN $mfn\n") if ($i > 1);
}
+
}
if ($validate) {
eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
+ foreach my $out ( @output_modules ) {
+ $out->finish if $out->can('finish');
+ }
+
my $dt = time() - $start_t;
$log->info("$total_rows records ", $indexer ? "indexed " : "",
sprintf("in %.2f sec [%.2f rec/sec]",
}
# save new delimiters if needed
-$validate->save_delimiters_templates if ( $validate );
+$validate->save_delimiters_templates if ( $validate_delimiters_path );
#
# handle links or merge after indexing