--- trunk/run.pl 2006/11/03 19:41:28 768 +++ trunk/run.pl 2007/11/03 12:34:26 977 @@ -8,11 +8,11 @@ use WebPAC::Common 0.02; use WebPAC::Parser 0.08; -use WebPAC::Input 0.14; -use WebPAC::Store 0.14; +use WebPAC::Input 0.16; +use WebPAC::Store 0.15; use WebPAC::Normalize 0.22; use WebPAC::Output::TT; -use WebPAC::Validate 0.06; +use WebPAC::Validate 0.11; use WebPAC::Output::MARC; use WebPAC::Config; use Getopt::Long; @@ -21,6 +21,7 @@ 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 @@ -31,7 +32,7 @@ B -Options: +=head1 OPTIONS =over 4 @@ -67,6 +68,19 @@ turn on extra validation of imput records, see L +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 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 directives. You can disable lint @@ -102,11 +116,14 @@ 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 $log = _new WebPAC::Common()->_get_logger(); @@ -116,17 +133,24 @@ "clean" => \$clean, "one=s" => \$only_filter, "only=s" => \$only_filter, - "config" => \$config_path, + "config=s" => \$config_path, "debug+" => \$debug, "stats" => \$stats, "validate=s" => \$validate_path, + "validate-delimiters=s" => \$validate_delimiters_path, + "marc-generate!" => \$marc_generate, "marc-lint!" => \$marc_lint, "marc-dump!" => \$marc_dump, "parallel=i" => \$parallel, "only-links!" => \$only_links, "merge" => \$merge, + "help" => \$help, ); +$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); @@ -135,6 +159,12 @@ $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'; @@ -145,24 +175,29 @@ $log->info("created merge batch file $estcmd_path"); } - my $validate; $validate = new WebPAC::Validate( - path => $validate_path, -) if ($validate_path); - + delimiters => $config->webpac('delimiters'), +) if ($validate_path || $validate_delimiters_path); my $use_indexer = $config->use_indexer; +$stats ||= $validate; if ($stats) { - $log->debug("option --stats disables update of indexing engine..."); + $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(); @@ -183,9 +218,9 @@ return $c; } -while (my ($database, $db_config) = each %{ $config->databases }) { +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 ($parallel) { @@ -214,27 +249,24 @@ if ($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); } @@ -268,36 +300,49 @@ # # now WebPAC::Store # - my $abs_path = abs_path($0); - $abs_path =~ s#/[^/]*$#/#; + my $store = new WebPAC::Store({ + debug => $debug, + }); + + + # + # 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 $db_path = $config->webpac('db_path'); + my @output_modules; - if ($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"); - } + foreach my $output ( @outputs ) { - my $store = new WebPAC::Store( - path => $db_path, - debug => $debug, - ); +#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; + + $log->debug("calling $module->new(",dump( $output ),")"); + my $out = new $module->new( $output ); + $out->init; + + push @output_modules, $out; + } # # now, iterate through input formats # - my @inputs; - if (ref($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"); - } + } ); foreach my $input (@inputs) { @@ -307,9 +352,16 @@ 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); @@ -332,6 +384,7 @@ 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); @@ -416,10 +469,15 @@ 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 ( $validate ) { + $validate->read_validate_file( $validate->fill_in( $validate_path, database => $database, input => $input_name ) ) if ( $validate_path ); + $validate->read_validate_delimiters_file( $validate->fill_in( $validate_delimiters_path, database => $database, input => $input_name ) ) if ( $validate_delimiters_path ); + } } my $marc; - if ($parser->have_rules( 'marc', $database, $input_name )) { + if ($marc_generate && $parser->have_rules( 'marc', $database, $input_name )) { $marc = new WebPAC::Output::MARC( path => "out/marc/${database}-${input_name}.marc", lint => $marc_lint, @@ -463,13 +521,25 @@ } + # 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; + $total_rows++; + 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'} }, $pos; @@ -477,11 +547,12 @@ if ($validate) { - if ( my $errors = $validate->validate_rec( $row, $input_db->dump ) ) { + 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 $ds = WebPAC::Normalize::data_structure( @@ -492,6 +563,7 @@ marc_encoding => 'utf-8', load_row_coderef => sub { my ($database,$input,$mfn) = @_; +warn "### load_row($database,$input,$mfn) from data_structure\n"; return $store->load_row( database => $database, input => $input, @@ -500,20 +572,30 @@ }, ); - $log->debug("ds = ", sub { dump($ds) }) if ($ds); + $log->debug("ds = ", sub { dump($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); + $store->save_ds( + database => $database, + input => $input_name, + id => $mfn, + ds => $ds, + ) if !$stats; + + $indexer->add( + id => "${input_name}/${mfn}", + ds => $ds, + type => $config->get($indexer_config)->{type}, + ) if $indexer; + + foreach my $out ( @output_modules ) { + $out->add( $mfn, $ds ) if $out->can('add'); + } + + } else { + $log->warn("record $pos didn't produce any output after normalization rules!") unless $marc; + } if ($marc) { my $i = 0; @@ -522,7 +604,7 @@ $marc->add( id => $mfn . ( $i ? "/$i" : '' ), fields => $fields, - leader => WebPAC::Normalize::marc_leader(), + leader => WebPAC::Normalize::_get_marc_leader(), row => $row, ); $i++; @@ -531,7 +613,6 @@ $log->info("Created $i instances of MFN $mfn\n") if ($i > 1); } - $total_rows++; } if ($validate) { @@ -540,6 +621,11 @@ $log->info("validation errors:\n$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) { @@ -552,12 +638,15 @@ $marc->finish if ($marc); # close report - close($report_fh) if ($report_fh) - + close($report_fh) if ($report_fh); } 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]", @@ -580,6 +669,9 @@ $log->info("all parallel processes finished"); } +# save new delimiters if needed +$validate->save_delimiters_templates if ( $validate_delimiters_path ); + # # handle links or merge after indexing #