--- trunk/run.pl 2005/11/25 00:22:58 141 +++ trunk/run.pl 2006/06/29 15:29:32 539 @@ -7,108 +7,312 @@ use Data::Dumper; use lib './lib'; +use WebPAC::Common 0.02; use WebPAC::Lookup; -use WebPAC::Input::ISIS; -use WebPAC::DB 0.02; -use WebPAC::Normalize::XML; +use WebPAC::Input 0.03; +use WebPAC::Store 0.03; +use WebPAC::Normalize; use WebPAC::Output::TT; -use WebPAC::Output::Estraier; +use WebPAC::Validate; use YAML qw/LoadFile/; +use Getopt::Long; +use File::Path; +use Time::HiRes qw/time/; +use File::Slurp; -my $limit = shift @ARGV; +=head1 NAME -my $config = LoadFile('conf/config.yml'); +run.pl - start WebPAC indexing -print "config = ",Dumper($config); +B -my $type = lc($config->{input}->{type}); +Options: -die "I know only how to handle input type isis, not '$type'!\n" unless ($type eq 'isis'); +=over 4 -my $abs_path = abs_path($0); -$abs_path =~ s#/[^/]*$#/#; +=item --offset 42 -my $lookup = new WebPAC::Lookup( - lookup_file => $config->{input}->{lookup}, -); +start loading (all) databases at offset 42 +=item --limit 100 +limit loading to 100 records -my $isis = new WebPAC::Input::ISIS( - code_page => $config->{webpac}->{webpac_encoding}, - limit_mfn => $config->{input}->{limit}, -); +=item --clean -my $maxmfn = $isis->open( - filename => $config->{input}->{path}, - code_page => $config->{input}->{encoding}, # database encoding -); +remove database and Hyper Estraier index before indexing -my $path = './db/'; +=item --only=database_name/input_filter -my $db = new WebPAC::DB( - path => $config->{webpac}->{db_path}, -); +reindex just single database (legacy name is --one) -my $n = new WebPAC::Normalize::XML( -# filter => { 'foo' => sub { shift } }, - db => $db, - lookup_regex => $lookup->regex, - lookup => $lookup, -); +C is optional part which can be C +or C from input -$n->open( - tag => $config->{normalize}->{tag}, - xml_file => $config->{normalize}->{path}, -); +=item --config conf/config.yml -my $out = new WebPAC::Output::TT( - include_path => $config->{webpac}->{template_path}, - filters => { foo => sub { shift } }, -); +path to YAML configuration file + +=item --stats + +disable indexing 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 + +=back + +=cut + +my $offset; +my $limit; -my $est = new WebPAC::Output::Estraier( - %{ $config->{hyperestraier} } +my $clean = 0; +my $config = 'conf/config.yml'; +my $debug = 0; +my $only_filter; +my $stats = 0; +my $validate_path; + +GetOptions( + "limit=i" => \$limit, + "offset=i" => \$offset, + "clean" => \$clean, + "one=s" => \$only_filter, + "only=s" => \$only_filter, + "config" => \$config, + "debug" => \$debug, + "stats" => \$stats, + "validate=s" => \$validate_path, ); -my $total_rows = 0; +$config = LoadFile($config); -for ( 0 ... $isis->size ) { +print "config = ",Dumper($config) if ($debug); - my $row = $isis->fetch || next; +die "no databases in config file!\n" unless ($config->{databases}); - my $mfn = $row->{'000'}->[0] || die "can't find MFN"; +my $log = _new WebPAC::Common()->_get_logger(); +$log->info( "-" x 79 ); - my $ds = $n->data_structure($row); +my $validate; +$validate = new WebPAC::Validate( + path => $validate_path, +) if ($validate_path); + +my $use_indexer = $config->{use_indexer} || 'hyperestraier'; +if ($stats) { + $log->debug("option --stats disables update of indexing engine..."); + $use_indexer = undef; +} else { + $log->info("using $use_indexer indexing engine..."); +} + +my $total_rows = 0; +my $start_t = time(); -# print STDERR Dumper($row, $ds); - -# my $html = $out->apply( -# template => 'html_ffzg.tt', -# data => $ds, -# ); -# -# # create test output -# -# my $file = sprintf('out/%02d.html', $mfn ); -# open(my $fh, '>', $file) or die "can't open $file: $!"; -# print $fh $html; -# close($fh); -# -# $html =~ s#\s*[\n\r]+\s*##gs; -# -# print STDERR $html; - - $est->add( - id => $mfn, - ds => $ds, - type => $config->{hyperestraier}->{type}, +my @links; +my $indexer; + +while (my ($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 ($use_indexer) { + my $indexer_config = $config->{$use_indexer} || $log->logdie("can't find '$use_indexer' part in confguration"); + $indexer_config->{database} = $database; + $indexer_config->{clean} = $clean; + $indexer_config->{label} = $db_config->{name}; + + if ($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 '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); + } + + + # + # 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 { + $log->info("working on database '$database' in $db_path"); + } + + my $db = new WebPAC::Store( + path => $db_path, + database => $database, + debug => $debug, ); - $total_rows++; -}; + # + # now, iterate through input formats + # + + my @inputs; + if (ref($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"); + } + + 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 $type = lc($input->{type}); + + 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 => $input->{lookup}, + ) if ($input->{lookup}); -my $log = $lookup->_get_logger; + my $input_module = $config->{webpac}->{inputs}->{$type}; + + $log->info("working on input '$input->{name}' in $input->{path} [type: $input->{type}] using $input_module", + $input->{lookup} ? "lookup '$input->{lookup}'" : "" + ); + + my $input_db = new WebPAC::Input( + module => $input_module, + code_page => $config->{webpac}->{webpac_encoding}, + limit => $limit || $input->{limit}, + offset => $offset, + lookup => $lookup, + recode => $input->{recode}, + stats => $stats, + ); + $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 + %{ $input }, + ); + + my @norm_array = ref($input->{normalize}) eq 'ARRAY' ? + @{ $input->{normalize} } : ( $input->{normalize} ); + + foreach my $normalize (@norm_array) { + + my $rules; + my $normalize_path = $normalize->{path} || $log->logdie("can't find normalize path in config"); + + $log->logdie("Found '$normalize_path' as normalization file which isn't supported any more!") unless ( $normalize_path =~ m!\.pl$!i ); + + my $rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!"; + + $log->info("Using $normalize_path for normalization..."); + + # reset position in database + $input_db->seek(1); + + foreach my $pos ( 0 ... $input_db->size ) { + + 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'} }, $pos; + } + + + if ($validate) { + my @errors = $validate->validate_errors( $row ); + $log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors); + } + + + my $ds = WebPAC::Normalize::data_structure( + row => $row, + rules => $rules, + lookup => $lookup ? $lookup->lookup_hash : undef, + ); + + $db->save_ds( + id => $mfn, + ds => $ds, + prefix => $input->{name}, + ) if ($ds && !$stats); + + $indexer->add( + id => $input->{name} . "/" . $mfn, + ds => $ds, + type => $config->{$use_indexer}->{type}, + ) if ($indexer); + + $total_rows++; + } + + $log->info("statistics of fields usage:\n", $input_db->stats) if ($stats); + + }; + + } + + 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 / $dt) + ) + ); -$log->info("$total_rows records indexed"); + # + # add Hyper Estraier links to other databases + # + if (ref($db_config->{links}) eq 'ARRAY') { + foreach my $link (@{ $db_config->{links} }) { + if ($use_indexer eq 'hyperestraier') { + $log->info("saving link $database -> $link->{to} [$link->{credit}]"); + push @links, { + from => $database, + to => $link->{to}, + credit => $link->{credit}, + }; + } else { + $log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]"); + } + } + } + +} + +foreach my $link (@links) { + $log->info("adding link $link->{from} -> $link->{to} [$link->{credit}]"); + $indexer->add_link( %{ $link } ); +}