--- trunk/run.pl 2006/06/29 23:19:26 547 +++ trunk/run.pl 2006/09/06 19:25:22 636 @@ -4,22 +4,26 @@ use Cwd qw/abs_path/; use File::Temp qw/tempdir/; -use Data::Dumper; use lib './lib'; use WebPAC::Common 0.02; -use WebPAC::Lookup; -use WebPAC::Input 0.03; +use WebPAC::Lookup 0.03; +use WebPAC::Input 0.11; use WebPAC::Store 0.03; -use WebPAC::Normalize; +use WebPAC::Normalize 0.11; use WebPAC::Output::TT; use WebPAC::Validate; +use WebPAC::Output::MARC; use YAML qw/LoadFile/; use Getopt::Long; use File::Path; use Time::HiRes qw/time/; use File::Slurp; -use MARC::Record 2.0; # need 2.0 for utf-8 encoding see marcpm.sf.net +use Data::Dump qw/dump/; +use Storable qw/dclone/; + +use Proc::Queue size => 1; +use POSIX ":sys_wait_h"; # imports WNOHANG =head1 NAME @@ -63,6 +67,36 @@ turn on extra validation of imput records, see L +=item --marc-normalize conf/normalize/mapping.pl + +This option specifies normalisation file for MARC creation + +=item --marc-output out/marc/test.marc + +Optional path to output file + +=item --marc-lint + +By default turned on if C<--marc-normalize> is used. You can disable lint +messages with C<--no-marc-lint>. + +=item --marc-dump + +Force dump or input and marc record for debugging. + +=item --parallel 4 + +Run databases in parallel (aproximatly same as number of processors in +machine if you want to use full load) + +=item --only-links + +Create just links + +=item --merge + +Create merged index of databases which have links + =back =cut @@ -76,6 +110,22 @@ my $only_filter; my $stats = 0; my $validate_path; +my ($marc_normalize, $marc_output); +my $marc_lint = 1; +my $marc_dump = 0; +my $parallel = 0; +my $only_links = 0; +my $merge = 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, @@ -84,25 +134,43 @@ "one=s" => \$only_filter, "only=s" => \$only_filter, "config" => \$config, - "debug" => \$debug, + "debug+" => \$debug, "stats" => \$stats, "validate=s" => \$validate_path, + "marc-normalize=s" => \$marc_normalize, + "marc-output=s" => \$marc_output, + "marc-lint!" => \$marc_lint, + "marc-dump!" => \$marc_dump, + "parallel=i" => \$parallel, + "only-links!" => \$only_links, + "merge" => \$merge, ); $config = LoadFile($config); -print "config = ",Dumper($config) if ($debug); +#print "config = ",dump($config) if ($debug); die "no databases in config file!\n" unless ($config->{databases}); -my $log = _new WebPAC::Common()->_get_logger(); $log->info( "-" x 79 ); + +my $estcmd_fh; +my $estcmd_path = './estcmd-merge.sh'; +if ($merge) { + open($estcmd_fh, '>', $estcmd_path) || $log->logdie("can't open $estcmd_path: $!"); + print $estcmd_fh 'cd /data/estraier/_node/ || exit 1',$/; + print $estcmd_fh 'sudo /etc/init.d/hyperestraier stop',$/; + $log->info("created merge batch file $estcmd_path"); +} + + 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..."); @@ -111,29 +179,61 @@ $log->info("using $use_indexer indexing engine..."); } +# disable indexing when creating marc +$use_indexer = undef if ($marc_normalize); + my $total_rows = 0; my $start_t = time(); my @links; -my $indexer; + +if ($parallel) { + $log->info("Using $parallel processes for speedup"); + Proc::Queue::size($parallel); +} 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 ($parallel) { + my $f=fork; + if(defined ($f) and $f==0) { + $log->info("Created processes $$ for speedup"); + } else { + next; + } + } + + my $indexer; if ($use_indexer) { - my $indexer_config = $config->{$use_indexer} || $log->logdie("can't find '$use_indexer' part in confguration"); + + my $cfg_name = $use_indexer; + $cfg_name =~ s/\-.*$//; + + my $indexer_config = $config->{$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}; + # force clean if database has links + $indexer_config->{clean} = 1 if ($db_config->{links}); + 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 'hyperestraier-native') { + + # open Hyper Estraier database + use WebPAC::Output::EstraierNative; + $indexer = new WebPAC::Output::EstraierNative( %{ $indexer_config } ); + + $use_indexer = 'hyperestraier'; + } elsif ($use_indexer eq 'kinosearch') { # open KinoSearch @@ -150,6 +250,33 @@ # + # store Hyper Estraier links to other databases + # + if (ref($db_config->{links}) eq 'ARRAY' && $use_indexer) { + foreach my $link (@{ $db_config->{links} }) { + if ($use_indexer eq 'hyperestraier') { + if ($merge) { + print $estcmd_fh 'sudo -u www-data estcmd merge ' . $database . ' ' . $link->{to},$/; + } else { + $log->info("saving link $database -> $link->{to} [$link->{credit}]"); + push @links, sub { + $log->info("adding link $database -> $link->{to} [$link->{credit}]"); + $indexer->add_link( + from => $database, + to => $link->{to}, + credit => $link->{credit}, + ); + }; + } + } else { + $log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]"); + } + } + } + next if ($only_links); + + + # # now WebPAC::Store # my $abs_path = abs_path($0); @@ -194,9 +321,13 @@ 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 $lookup; + if ($input->{lookup}) { + $lookup = new WebPAC::Lookup( + lookup_file => $input->{lookup}, + ); + delete( $input->{lookup} ); + } my $input_module = $config->{webpac}->{inputs}->{$type}; @@ -206,12 +337,17 @@ my $input_db = new WebPAC::Input( module => $input_module, - code_page => $config->{webpac}->{webpac_encoding}, + encoding => $config->{webpac}->{webpac_encoding}, limit => $limit || $input->{limit}, offset => $offset, - lookup => $lookup, + lookup_coderef => sub { + my $rec = shift || return; + $lookup->add( $rec ); + }, recode => $input->{recode}, stats => $stats, + modify_records => $input->{modify_records}, + modify_file => $input->{modify_file}, ); $log->logdie("can't create input using $input_module") unless ($input); @@ -224,6 +360,13 @@ my @norm_array = ref($input->{normalize}) eq 'ARRAY' ? @{ $input->{normalize} } : ( $input->{normalize} ); + if ($marc_normalize) { + @norm_array = ( { + path => $marc_normalize, + output => $marc_output || 'out/marc/' . $database . '-' . $input->{name} . '.marc', + } ); + } + foreach my $normalize (@norm_array) { my $normalize_path = $normalize->{path} || $log->logdie("can't find normalize path in config"); @@ -234,13 +377,11 @@ $log->info("Using $normalize_path for normalization..."); - my $marc_fh; - if (my $path = $normalize->{marc21}) { - open($marc_fh, '>', $path) || - $log->logdie("can't open MARC output $path: $!"); - - $log->info("Creating MARC export file $path\n"); - } + my $marc = new WebPAC::Output::MARC( + path => $normalize->{output}, + lint => $marc_lint, + dump => $marc_dump, + ) if ($normalize->{output}); # reset position in database $input_db->seek(1); @@ -263,11 +404,22 @@ $log->error( "MFN $mfn validation errors:\n", join("\n", @errors) ) if (@errors); } - + 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, marc_encoding => 'utf-8', ); @@ -283,11 +435,20 @@ type => $config->{$use_indexer}->{type}, ) if ($indexer && $ds); - if ($marc_fh) { - my $marc = new MARC::Record; - $marc->encoding( 'utf-8' ); - $marc->add_fields( WebPAC::Normalize::_get_marc_fields() ); - print $marc_fh $marc->as_usmarc; + if ($marc) { + my $i = 0; + + while (my $fields = WebPAC::Normalize::_get_marc_fields( fetch_next => 1 ) ) { + $marc->add( + id => $mfn . ( $i ? "/$i" : '' ), + fields => $fields, + leader => WebPAC::Normalize::marc_leader(), + row => $row, + ); + $i++; + } + + $log->info("Created $i instances of MFN $mfn\n") if ($i > 1); } $total_rows++; @@ -296,7 +457,7 @@ $log->info("statistics of fields usage:\n", $input_db->stats) if ($stats); # close MARC file - close($marc_fh) if ($marc_fh); + $marc->finish if ($marc); } @@ -311,27 +472,33 @@ ) ); - # - # 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}]"); - } - } + + # end forked process + if ($parallel) { + $log->info("parallel process $$ finished"); + exit(0); } } -foreach my $link (@links) { - $log->info("adding link $link->{from} -> $link->{to} [$link->{credit}]"); - $indexer->add_link( %{ $link } ); +if ($parallel) { + # wait all children to finish + sleep(1) while wait != -1; + $log->info("all parallel processes finished"); +} + +# +# handle links or merge after indexing +# + +if ($merge) { + print $estcmd_fh 'sudo /etc/init.d/hyperestraier start',$/; + close($estcmd_fh); + chmod 0700, $estcmd_path || $log->warn("can't chmod 0700 $estcmd_path: $!"); + system $estcmd_path; +} else { + foreach my $link (@links) { + $log->logdie("coderef in link ", Dumper($link), " is ", ref($link), " and not CODE") unless (ref($link) eq 'CODE'); + $link->(); + } }