use lib './lib';
use WebPAC::Common 0.02;
-use WebPAC::Lookup 0.03;
-use WebPAC::Input 0.11;
-use WebPAC::Store 0.03;
+use WebPAC::Parser 0.04;
+use WebPAC::Input 0.13;
+use WebPAC::Store 0.11;
use WebPAC::Normalize 0.11;
use WebPAC::Output::TT;
use WebPAC::Validate 0.06;
use WebPAC::Output::MARC;
-use YAML qw/LoadFile/;
+use WebPAC::Config;
use Getopt::Long;
use File::Path;
use Time::HiRes qw/time/;
my $limit;
my $clean = 0;
-my $config = 'conf/config.yml';
+my $config_path;
my $debug = 0;
my $only_filter;
my $stats = 0;
my $log = _new WebPAC::Common()->_get_logger();
-my $hostname = `hostname`;
-chomp($hostname);
-$hostname =~ s/\..+$//;
-if (-e "conf/$hostname.yml") {
- $config = "conf/$hostname.yml";
- $log->info("using host configuration file: $config");
-}
-
GetOptions(
"limit=i" => \$limit,
"offset=i" => \$offset,
"clean" => \$clean,
"one=s" => \$only_filter,
"only=s" => \$only_filter,
- "config" => \$config,
+ "config" => \$config_path,
"debug+" => \$debug,
"stats" => \$stats,
"validate=s" => \$validate_path,
"merge" => \$merge,
);
-$config = LoadFile($config);
+my $config = new WebPAC::Config( path => $config_path );
#print "config = ",dump($config) if ($debug);
-die "no databases in config file!\n" unless ($config->{databases});
+die "no databases in config file!\n" unless ($config->databases);
$log->info( "-" x 79 );
) if ($validate_path);
-my $use_indexer = $config->{use_indexer} || 'hyperestraier';
+my $use_indexer = $config->use_indexer;
if ($stats) {
$log->debug("option --stats disables update of indexing engine...");
$use_indexer = undef;
# disable indexing when creating marc
$use_indexer = undef if ($marc_normalize);
+# parse normalize files and create source files for lookup and normalization
+
+my $parser = new WebPAC::Parser( config => $config );
+
my $total_rows = 0;
my $start_t = time();
Proc::Queue::size($parallel);
}
-while (my ($database, $db_config) = each %{ $config->{databases} }) {
+sub create_ds_config {
+ my ($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);
my $cfg_name = $use_indexer;
$cfg_name =~ s/\-.*$//;
- my $indexer_config = $config->{$cfg_name} || $log->logdie("can't find '$cfg_name' part in confguration");
+ 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};
my $abs_path = abs_path($0);
$abs_path =~ s#/[^/]*$#/#;
- my $db_path = $config->{webpac}->{db_path} . '/' . $database;
+ my $db_path = $config->webpac('db_path');
if ($clean) {
$log->info("creating new database '$database' in $db_path");
$log->info("working on database '$database' in $db_path");
}
- my $db = new WebPAC::Store(
+ my $store = new WebPAC::Store(
path => $db_path,
- database => $database,
debug => $debug,
);
$log->info("database $database doesn't have inputs defined");
}
- my @supported_inputs = keys %{ $config->{webpac}->{inputs} };
-
foreach my $input (@inputs) {
- next if ($only_input && ($input->{name} !~ m#$only_input#i && $input->{type} !~ m#$only_input#i));
+ 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(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs));
+ die "I know only how to handle input types ", join(",", $config->webpac('inputs') ), " not '$type'!\n" unless (grep(/$type/, $config->webpac('inputs')));
- my $lookup;
- if ($input->{lookup}) {
- $lookup = new WebPAC::Lookup(
- lookup_file => $input->{lookup},
- );
- delete( $input->{lookup} );
- }
+ my $input_module = $config->webpac('inputs')->{$type};
- 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",
- $input->{lookup} ? "lookup '$input->{lookup}'" : ""
+ $log->info("working on input '$input_name' in $input->{path} [type: $input->{type}] using $input_module",
+ @lookups ? " creating lookups: ".join(", ", @lookups) : ""
);
if ($stats) {
my $input_db = new WebPAC::Input(
module => $input_module,
- encoding => $config->{webpac}->{webpac_encoding},
+ encoding => $config->webpac('webpac_encoding'),
limit => $limit || $input->{limit},
offset => $offset,
- lookup_coderef => sub {
- my $rec = shift || return;
- $lookup->add( $rec );
- },
recode => $input->{recode},
stats => $stats,
modify_records => $input->{modify_records},
);
$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;
+ 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,
+ lookup => $lookup,
+ config => create_ds_config( $db_config, $database, $input, $mfn ),
+ );
+
+ warn "current lookup = ", dump($lookup) if ($lookup);
+ };
+
+ WebPAC::Normalize::_set_lookup( undef );
+
+ $log->debug("created lookup_coderef using:\n$rules");
+
+ };
+
my $maxmfn = $input_db->open(
path => $input->{path},
code_page => $input->{encoding}, # database encoding
+ lookup_coderef => $lookup_coderef,
%{ $input },
);
+ my $lookup_data = WebPAC::Normalize::_get_lookup();
+
+ if (defined( $lookup_data->{$database}->{$input_name} )) {
+ $log->debug("created following lookups: ", dump( $lookup_data ));
+
+ foreach my $key (keys %{ $lookup_data->{$database}->{$input_name} }) {
+ $store->save_lookup(
+ database => $database,
+ input => $input_name,
+ key => $key,
+ data => $lookup_data->{$database}->{$input_name}->{$key},
+ );
+ }
+ }
+
my $report_fh;
if ($stats || $validate) {
- my $path = "out/report/" . $database . '-' . $input->{name} . '.txt';
+ 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 ",
+ print $report_fh "Report for database '$database' input '$input_name' records ",
$offset || 1, "-", $limit || $input->{limit} || $maxmfn, "\n\n";
$log->info("Generating report file $path");
}
if ($marc_normalize) {
@norm_array = ( {
path => $marc_normalize,
- output => $marc_output || 'out/marc/' . $database . '-' . $input->{name} . '.marc',
+ output => $marc_output || "out/marc/${database}-${input_name}.marc",
} );
}
}
}
- my $ds_config = dclone($db_config);
-
- # default values -> database key
- $ds_config->{_} = $database;
-
- # current mfn
- $ds_config->{_mfn} = $mfn;
-
- # attach current input
- $ds_config->{input} = $input;
-
my $ds = WebPAC::Normalize::data_structure(
row => $row,
rules => $rules,
lookup => $lookup ? $lookup->lookup_hash : undef,
- config => $ds_config,
+ config => create_ds_config( $db_config, $database, $input, $mfn ),
marc_encoding => 'utf-8',
);
- $db->save_ds(
+ $store->save_ds(
+ database => $database,
+ input => $input_name,
id => $mfn,
ds => $ds,
- prefix => $input->{name},
) if ($ds && !$stats);
$indexer->add(
- id => $input->{name} . "/" . $mfn,
+ id => "${input_name}/${mfn}",
ds => $ds,
- type => $config->{$indexer_config}->{type},
+ type => $config->get($indexer_config)->{type},
) if ($indexer && $ds);
if ($marc) {
projects / webpac2 / blobdiff