--- trunk/all2xml.pl 2003/02/23 00:00:51 21 +++ trunk2/all2all.pl 2004/09/26 15:11:36 471 @@ -1,275 +1,300 @@ #!/usr/bin/perl -w -use strict; -use OpenIsis; -use Getopt::Std; -use Data::Dumper; -use XML::Simple; -use Text::Unaccent 1.02; # 1.01 won't compile on my platform, -use Text::Iconv; -use Config::IniFiles; +=head1 NAME -$|=1; - -my $config_file = $0; -$config_file =~ s/\.pl$/.conf/; -die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file); - -my $config; - -use index_DBI; # there is no other, right now ;-) - -my $index = new index_DBI(); # open index - -my %opts; - -# usage: -# -d directory name -# -m multiple directories -# -q quiet -# -s run swish - -getopts('d:m:qs', \%opts); - -my $db_dir; - -#die "usage: $0 -d [database_dir] -m [database1,database2] " if (! %opts); - -#print Dumper($config->{indexer}); -#print "-" x 70,"\n"; - -Text::Iconv->raise_error(1); # Conversion errors raise exceptions - -#my $isis_codepage = Text::Iconv->new($config->{isis_codepage},'UTF8'); -#my $index_codepage = Text::Iconv->new($config->{isis_codepage},$config->{index_codepage}); -my $isis_codepage; -my $index_codepage; -my $cludge_codepage = Text::Iconv->new('UTF8','ISO8859-1'); -my $xml_codepage; - -sub isis2xml { - - use xmlify; - - my $row = shift @_; - my $add_xml = shift @_; - - my $xml; - - use parse_format; - - my $html = ""; # html formatted display output - - my %field_usage; # counter for usage of each field - - foreach my $field (keys %{$config->{indexer}}) { - - $field_usage{$field}++; - - my $swish_data = ""; - my $display_data = ""; - - foreach my $x (@{$config->{indexer}->{$field}->{isis}}) { - - my $format = $x->{content}; - my ($s,$d,$i) = (1,1,0); # swish, display default - $s = 0 if (lc($x->{type}) eq "display"); - $d = 0 if (lc($x->{type}) eq "swish"); - ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); -#print STDERR "## s: $s d: $d i: $i ## $format ##\n"; - - # FIX: this is ugly, UGLY, cludge string is returned - # in UTF8 encoding , but as if source charset - # is ISO8859-1 and not some other. This break other - # character encodings, so we convert it first - # back to ISO8859-1 - $format = $cludge_codepage->convert($format); - - my ($swish,$display) = parse_format($format,$row); -#print STDERR "s: $swish\nd: $display\n" if ($swish); - -#print STDERR "swish: $swish<-- display: $display<--\n"; - - # filter="name" ; filter this field through - # filter/[name].pm - my $filter; - if ($x->{filter}) { - $filter = "filter/".$x->{filter}.".pm"; - require $filter; - } - # type="swish" ; field for swish - if ($s && $swish) { - if ($filter) { - $swish_data .= join(" ",&filter($swish)); - } else { - $swish_data .= $swish if ($s && $swish); - } - } - - # type="display" ; field for display - if ($d && $display) { - if ($filter) { - $display_data .= join(" ",&filter($display)); - } else { - $display_data .= $display; - } - } - - # type="index" ; insert into index - if ($i && $display) { - my $index_data = $index_codepage->convert($display) || $display; - if ($filter) { - foreach my $d (&filter($index_data)) { - $index->insert($field, $d, $db_dir); - } - } else { - $index->insert($field, $index_data, $db_dir); - } - } - } +all2all.pl - basic script for all WebPAC needs +=cut -#print STDERR "s_d: $swish_data\nd_d: $display_data\n" if ($swish_data); - if ($display_data) { - $display_data = $isis_codepage->convert($display_data) || die "Can't convert '$display_data' !"; - # FIX: this is removed and replaced by html tag. - #$xml .= xmlify($field."_display", $display_data); - - if ($field eq "headline") { - $xml .= xmlify("headline", $display_data); - } else { - - # find field name (signular, plural) - my $field_name = ""; - if ($config->{indexer}->{$field}->{name_singular} && $field_usage{$field} == 1) { - $field_name = $config->{indexer}->{$field}->{name_singular}."#-#"; - } elsif ($config->{indexer}->{$field}->{name_plural}) { - $field_name = $config->{indexer}->{$field}->{name_plural}."#-#"; - } elsif ($config->{indexer}->{$field}->{name}) { - $field_name = $config->{indexer}->{$field}->{name}."#-#"; - } else { - print STDERR "WARNING: field '$field' doesn't have 'name' attribute!"; - } - if ($field_name) { - $html .= $xml_codepage->convert($field_name); - } - $html .= $display_data."###\n"; - } - } - if ($swish_data) { - my $i = Text::Iconv->new($config->{isis_codepage},'ISO8859-2'); - # remove extra spaces - $swish_data =~ s/ +/ /g; - $swish_data =~ s/ +$//g; - - $swish_data = $i->convert($swish_data); - $xml .= xmlify($field."_swish",unac_string('ISO8859-2',$swish_data)); - #$swish_data = $isis_codepage->convert($swish_data)."##" || $swish_data; - #$xml .= xmlify($field."_swish",unac_string($config->{isis_codepage},$swish_data)); - } +use strict; +use YAML; +use Carp; +use Getopt::Long; + +use lib './lib'; +use WebPAC; +use WebPAC::jsFind; +use WebPAC::Index; +use WebPAC::Tree; + +# options which can be changed via command line +# +my $code_page = 'ISO-8859-2'; +my ($limit_mfn, $start_mfn, $debug, $low_mem); +my $index_path = './out/index'; + +my $result = GetOptions( + "code_page=s" => \$code_page, + "limit_mfn=i" => \$limit_mfn, + "start_mfn=i" => \$start_mfn, + "debug!" => \$debug, + "low_mem!" => \$low_mem, +); + +# create WebPAC object +# +my $webpac = new WebPAC( + code_page => $code_page, + limit_mfn => $limit_mfn, + start_mfn => $start_mfn, + debug => $debug, + low_mem => $low_mem, +) || die; + +my $log = $webpac->_get_logger() || die "can't get logger"; + +$log->debug("creating WebPAC::jsFind object"); + +my $index = new WebPAC::jsFind( + index_path => $index_path, + keys => 10, +) || die; +my $thes; - } +$|=1; - # dump formatted output in - if ($html) { - $xml .= xmlify("html",$html); +my $maxmfn = $webpac->open_isis( + filename => shift @ARGV || '/data/hidra/THS/THS', + lookup => [ + { 'key' => 'd:v900', 'val' => 'v250^a' }, +# { 'eval' => '"v901^a" eq "Područje"', 'key' => 'pa:v561^4:v562^4:v461^1', 'val' => 'v900' }, +# { 'eval '=> '"v901^a" eq "Mikrotezaurus"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, +# { 'eval' => '"v901^a" eq "Deskriptor"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, + { 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' }, + { 'key' => '900_mfn:v900', 'val' => 'v000' }, + # tree structure + { 'eval' => 'length("v251") == 2 && "v800" =~ m/EUROVOC/ || "v800" =~ m/CROVOC/ && "v251" =~ m/^(H|HD|L|Z|P)$/', 'key' => 'root:v251', 'val' => 'v900' }, + { 'eval' => '"v251"', 'key' => 'code:v900', 'val' => 'v561^4:v251' }, + { 'eval' => '"v561^4" && "v562^4"', 'key' => 'code:v900', 'val' => 'v561^4:v562^4' }, + ], +); + +$log->debug("isis file ",$webpac->{'isis_filename'}," opened"); + +$log->info("rows: $maxmfn"); + +$webpac->open_import_xml(type => 'isis_hidra_ths'); + +if(1) { # XXX + +while (my $rec = $webpac->fetch_rec) { + + my @ds = $webpac->data_structure($rec); + + if (0 && $log->is_debug) { + $log->debug("rec = ",Dump($rec)); + $log->debug("ds = ",Dump(\@ds)); } - - if ($xml) { -#print STDERR "x: $xml\n"; - $xml .= $add_xml if ($add_xml); - return "\n$xml\n"; - } else { - return; - } -} -########################################################################## + next if (! @ds); -my $cfg = new Config::IniFiles( -file => $config_file ); + my $filename = $webpac->{'current_filename'} || $log->logdie("no current_filename in webpac object"); -foreach my $database ($cfg->Sections) { - - my $isis_db = $cfg -> val($database, 'isis_db'); - my $type = $cfg -> val($database, 'type'); - my $add_xml = $cfg -> val($database, 'xml'); - - # read configuration for this type - $config=XMLin("./import_xml/$type.xml", forcearray => [ 'isis' ], forcecontent => 1); - $isis_codepage = Text::Iconv->new($config->{isis_codepage},'UTF8'); - $index_codepage = Text::Iconv->new($config->{isis_codepage},$config->{index_codepage}); - $xml_codepage = Text::Iconv->new($cfg->val($database,'xml_codepage'),'UTF8'); - - my $db = OpenIsis::open( $isis_db ); - if (0) { -# # FIX -# if (! $db ) { - print STDERR "WARNING: can't open '$isis_db'\n"; - next ; + if ($filename) { + $webpac->output_file( + file => $filename, + template => 'html.tt', + data => \@ds, + headline => $webpac->{'headline'}, + ); + } else { + print $webpac->output( + template => 'text.tt', + data => \@ds, + headline => $webpac->{'headline'}, + ); } - my $max_rowid = OpenIsis::maxRowid( $db ); + my $headline = $webpac->{'headline'}; - print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; + my $f = $filename; + $f =~ s!out/!!; - my $path = $database; # was $isis_db - - my $last_p = 0; + # save into index + foreach my $ds (@ds) { + next if (! $ds->{'swish'}); + + # strip all non word characters from beginning or end + # of word + my $words = join(" ",@{$ds->{'swish'}}); + $words =~ s/^\W+//; + $words =~ s/\W*\s+\W*/ /g; + $words =~ s/\W+$//; + + $index->insert( + index_name => $ds->{'tag'}, + #path => $f, + path => $webpac->mfn, + headline => $headline, + words => $words, + ); + } -# { my $row_id = 4514; -# FIX - for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { - my $row = OpenIsis::read( $db, $row_id ); - if ($row && $row->{mfn}) { -#print STDERR "mfn: ",$row->{mfn},"\n"; - # output current process indicator - my $p = int($row->{mfn} * 100 / $max_rowid); - if ($p != $last_p) { - printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$row->{mfn},$max_rowid,"=" x ($p/2).">", $p ) if (! $opts{q}); - $last_p = $p; - } - - if (my $xml = isis2xml($row,$add_xml)) { -#print STDERR "--ret-->$xml\n"; - print "Path-Name: $path#".int($row->{mfn})."\n"; - print "Content-Length: ".(length($xml)+1)."\n"; - print "Document-Type: XML\n\n$xml\n"; - } + # save into sorted index (thesaurus) + foreach my $ds (@ds) { + next if (! $ds->{'index'}); + + $thes->{$ds->{'tag'}} ||= new WebPAC::Index; + + foreach my $h (@{$ds->{'index'}}) { + $thes->{$ds->{'tag'}}->insert( + headline => $h, + mfn => $webpac->mfn, + ); } } - print STDERR "\n"; -} -# call this to commit index -$index->close; +# print Dump(\@ds); -1; -__END__ -########################################################################## +} -=head1 NAME +foreach my $t (keys %{$thes}) { -isis2xml.pl - read isis file and dump XML + my @e = $thes->{$t}->elements; + if (! @e) { + $log->logwarn("no elements in sorted index $t?"); + next; + } -=head1 DESCRIPTION + my $file = "./out/bfilter/$t.txt"; + $log->info("saving sorted index $t to '$file' [".scalar(@e)." elements]"); -This command will read ISIS data file using OpenIsis perl module and -create XML file for usage with I -indexer. Dispite it's name, this script B -from isis files (isis allready has something like that). Output of this -script is tailor-made for SWISH-E. + $webpac->output_file( + file => $file, + template => 'index.tt', + data => \@e, + index_name => $t, + ); +} -=head1 AUTHOR +if (0 && $log->is_debug) { + $log->debug("lookup hash: ",Dump($webpac->{'lookup'})); + $log->debug("data hash: ",Dump($webpac->{'data'})); + foreach my $t (keys %{$thes}) { + $log->debug("thesaurus $t hash: ",Dump($thes->{$t})); + } +} -Dobrica Pavlinusic +} # XXX if(0) -=head1 COPYRIGHT +$log->debug("lookup hash: ",Dump($webpac->{'lookup'})); -GNU Public License (GPL) v2 or later +$log->info("creating tree"); -=head1 SEE ALSO +# +# define tree structure +# + +my $l = $webpac->{'lookup'} || $log->logconfess("can't find lookup"); + +my @tree = ({ + # level 0 + code_arr => sub { sort keys %{$l} }, + filter_code => sub { + my $t = shift; + return $t if ($t =~ s/root://); + }, + lookup_v900 => sub { shift @{$l->{"root:".$_[0]}} }, + lookup_term => sub { shift @{$l->{"d:".$_[1]}} }, + lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[1]}} }, + have_children => sub { return $l->{"a:".$_[0]."::"} }, + have_children_at_level => sub { + return unless (defined($l->{"code:".$_[1]})); + my $code = shift @{$l->{"code:".$_[1]}}; + print STDERR "## $_[1] -> $code\n"; + return undef unless($code); + return(9, $l->{"a:$code:"} ) if (defined($l->{"a:$code:"})); + }, + style => 'display: none', + },{ + # 1 + code_arr => sub { @{$_[0]} }, + filter_code => sub { shift }, # nop + lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, + lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, + lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, + have_children => sub { return $l->{"a:".$_[1].":"} }, + style => 'display: none', + },{ + # 2 + code_arr => sub { @{$_[0]} }, + filter_code => sub { shift }, + lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, + lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, + lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, + have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, + #style => 'display: none', + },{ + # 3 uži pojam + code_arr => sub { @{$_[0]} }, + filter_code => sub { shift }, + lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, + lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, + lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, + have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, + },{ + # 4 + code_arr => sub { @{$_[0]} }, + filter_code => sub { shift }, + lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, + lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, + lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, + have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, + },{ + # 5 + code_arr => sub { @{$_[0]} }, + filter_code => sub { shift }, + lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, + lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, + lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, + have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, + },{ + # 6 + code_arr => sub { @{$_[0]} }, + filter_code => sub { shift }, + lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, + lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, + lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, + have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, + },{ + # 7 + code_arr => sub { @{$_[0]} }, + filter_code => sub { shift }, + lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, + lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, + lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, + have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, + },{ + # 8 + code_arr => sub { @{$_[0]} }, + filter_code => sub { shift }, + lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, + lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, + lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, +# have_children => sub { return $l->{"a:".$_[1].":".$_[0]} }, + have_children => sub { 0 }, + },{ + # 9 - level which is never reached except explicitly + code_arr => sub { @{$_[0]} }, + filter_code => sub { shift }, + lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} }, + lookup_term => sub { shift @{$l->{"d:".$_[0]}} }, + lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} }, + have_children => sub { 0 }, + have_children_at_level => sub { defined($l->{"a:".$_[1].":".$_[0]}) && return (9,$l->{"a:".$_[1].":".$_[0]}) }, + },{ +}); + +my $tree = new WebPAC::Tree( + dir => './out', + html => 'browse.html', + template => './output_template/tree.tt', + js => 'tree-ids.js', + tree => \@tree, +); -SWISH-E web site at http://www.swish-e.org +$log->info("closing index"); +$index->close; -=cut +$log->info("elapsed time: ",$webpac->fmt_time(time() - $webpac->{'start_t'}));