--- trunk/run.pl 2005/12/15 14:12:00 251 +++ trunk/run.pl 2006/02/26 23:21:50 416 @@ -7,49 +7,116 @@ use Data::Dumper; use lib './lib'; +use WebPAC::Common 0.02; use WebPAC::Lookup; -use WebPAC::Input::ISIS; +use WebPAC::Input 0.03; use WebPAC::Store 0.03; use WebPAC::Normalize::XML; use WebPAC::Output::TT; -use WebPAC::Output::Estraier 0.02; +use WebPAC::Output::Estraier '0.10'; use YAML qw/LoadFile/; -use LWP::Simple; +use Getopt::Long; +use File::Path; +use Time::HiRes qw/time/; -my $limit = shift @ARGV; +=head1 NAME -my $config = LoadFile('conf/config.yml'); +run.pl - start WebPAC indexing -print "config = ",Dumper($config); +B + +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 --one=database_name + +reindex just single database + +=item --config conf/config.yml + +path to YAML configuration file + +=back + +=cut + +my $offset; +my $limit; + +my $clean = 0; +my $config = 'conf/config.yml'; +my $debug = 0; +my $one_db_name; + +GetOptions( + "limit=i" => \$limit, + "offset=i" => \$offset, + "clean" => \$clean, + "one=s" => \$one_db_name, + "config" => \$config, + "debug" => \$debug, +); + +$config = LoadFile($config); + +print "config = ",Dumper($config) if ($debug); die "no databases in config file!\n" unless ($config->{databases}); my $total_rows = 0; +my $start_t = time(); while (my ($database, $db_config) = each %{ $config->{databases} }) { + next if ($one_db_name && $database !~ m/$one_db_name/i); + + my $log = _new WebPAC::Common()->_get_logger(); + + # + # open Hyper Estraier database + # + + my $est_config = $config->{hyperestraier} || $log->logdie("can't find 'hyperestraier' part in confguration"); + $est_config->{database} = $database; + $est_config->{clean} = $clean; + $est_config->{label} = $db_config->{name}; + + my $est = new WebPAC::Output::Estraier( %{ $est_config } ); + + # + # now WebPAC::Store + # my $abs_path = abs_path($0); $abs_path =~ s#/[^/]*$#/#; my $db_path = $config->{webpac}->{db_path} . '/' . $database; + if ($clean) { + $log->info("creating new database $database in $db_path"); + rmtree( $db_path ) || $log->warn("can't remove $db_path: $!"); + } else { + $log->debug("working on $database in $db_path"); + } + my $db = new WebPAC::Store( path => $db_path, database => $database, - debug => 1, + debug => $debug, ); - my $log = $db->_get_logger; - $log->info("working on $database in $db_path"); - - my $est_config = $config->{hyperestraier} || $log->logdie("can't find 'hyperestraier' part in confguration"); - $est_config->{database} = $database; - - $log->info("using HyperEstraier URL $est_config->{masterurl}"); - - my $est = new WebPAC::Output::Estraier( - %{ $est_config }, - ); # # now, iterate through input formats @@ -58,35 +125,43 @@ my @inputs; if (ref($db_config->{input}) eq 'ARRAY') { @inputs = @{ $db_config->{input} }; - } else { + } elsif ($db_config->{input}) { push @inputs, $db_config->{input}; + } else { + $log->info("database $database doesn't have inputs defined"); } + my @supported_inputs = keys %{ $config->{webpac}->{inputs} }; + foreach my $input (@inputs) { my $type = lc($input->{type}); - die "I know only how to handle input type isis, not '$type'!\n" unless ($type eq 'isis'); + die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs)); my $lookup = new WebPAC::Lookup( lookup_file => $input->{lookup}, ); - $log->info("working on input $input->{path} [$input->{type}]"); + my $input_module = $config->{webpac}->{inputs}->{$type}; + + $log->info("working on input '$input->{path}' [$input->{type}] using $input_module lookup '$input->{lookup}'"); - my $isis = new WebPAC::Input::ISIS( + my $input_db = new WebPAC::Input( + module => $input_module, code_page => $config->{webpac}->{webpac_encoding}, - limit_mfn => $input->{limit}, + limit => $limit || $input->{limit}, + offset => $offset, lookup => $lookup, + recode => $input->{recode}, ); + $log->logdie("can't create input using $input_module") unless ($input); - my $maxmfn = $isis->open( - filename => $input->{path}, + my $maxmfn = $input_db->open( + path => $input->{path}, code_page => $input->{encoding}, # database encoding ); -$log->info( Dumper($lookup->{_lookup_data}) ); - my $n = new WebPAC::Normalize::XML( # filter => { 'foo' => sub { shift } }, db => $db, @@ -95,21 +170,36 @@ prefix => $input->{name}, ); - $n->open( - tag => $input->{normalize}->{tag}, - xml_file => $input->{normalize}->{path}, - ); + my $normalize_path = $input->{normalize}->{path}; - for ( 0 ... $isis->size ) { + if ($normalize_path =~ m/\.xml$/i) { + $n->open( + tag => $input->{normalize}->{tag}, + xml_file => $input->{normalize}->{path}, + ); + } elsif ($normalize_path =~ m/\.(?:yml|yaml)$/i) { + $n->open_yaml( + path => $normalize_path, + tag => $input->{normalize}->{tag}, + ); + } - my $row = $isis->fetch || next; + foreach my $pos ( 0 ... $input_db->size ) { - my $mfn = $row->{'000'}->[0] || die "can't find MFN"; + my $row = $input_db->fetch || next; + + my $mfn = $row->{'000'}->[0]; + + if (! $mfn || $mfn !~ m#^\d+$#) { + $log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos"); + $mfn = $pos; + push @{ $row->{'000'} }, $pos; + } my $ds = $n->data_structure($row); $est->add( - id => $input->{name} . "#" . $mfn, + id => $input->{name} . "/" . $mfn, ds => $ds, type => $config->{hyperestraier}->{type}, ); @@ -119,6 +209,26 @@ }; - $log->info("$total_rows records indexed"); + my $dt = time() - $start_t; + $log->info("$total_rows records indexed in " . + sprintf("%.2f sec [%.2f rec/sec]", + $dt, ($total_rows / $dt) + ) + ); + + # + # add Hyper Estraier links to other databases + # + if (ref($db_config->{links}) eq 'ARRAY') { + foreach my $link (@{ $db_config->{links} }) { + $log->info("adding link $database -> $link->{to} [$link->{credit}]"); + $est->add_link( + from => $database, + to => $link->{to}, + credit => $link->{credit}, + ); + } + } + }