--- trunk/all2xml.pl 2003/02/22 21:33:04 17
+++ trunk2/all2all.pl 2004/09/14 09:23:00 440
@@ -1,241 +1,438 @@
#!/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;
-
-$|=1;
-
-my $config_file = $0;
-$config_file =~ s/\.pl$/.conf/;
-die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file);
+=head1 NAME
-my $config;
+all2all.pl - basic script for all WebPAC needs
-use index_DBI; # there is no other, right now ;-)
+=cut
-my $index = new index_DBI(); # open index
+use strict;
+use Data::Dumper;
+use Carp;
+use Getopt::Long;
-my %opts;
+use lib './lib';
+use WebPAC;
+use WebPAC::jsFind;
+use WebPAC::Index;
+
+# 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;
-# usage:
-# -d directory name
-# -m multiple directories
-# -q quiet
-# -s run swish
+my $thes;
-getopts('d:m:qs', \%opts);
+$|=1;
-my $db_dir;
+my $maxmfn = $webpac->open_isis(
+ filename => shift @ARGV || '/data/hidra/THS/THS',
+ lookup => [
+ { 'key' => 'd:v900', 'val' => 'v250^a' },
+# { 'eval' => '"v901^a" eq "Područje"', 'key' => 'pa:v561^4:v562^4:v461^1', 'val' => 'v900' },
+# { 'eval '=> '"v901^a" eq "Mikrotezaurus"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' },
+# { 'eval' => '"v901^a" eq "Deskriptor"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' },
+ { 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' },
+ { 'key' => '900_mfn:v900', 'val' => 'v000' },
+ # tree structure
+ { 'eval' => 'length("v251") == 2', 'key' => 'root:v251', 'val' => 'v900' },
+ { 'eval' => '"v251"', 'key' => 'code:v900', 'val' => 'v561^4:v251' },
+ ],
+);
+
+$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) {
+ $log->debug("rec = ",Dumper($rec));
+ $log->debug("ds = ",Dumper(\@ds));
+ }
-#die "usage: $0 -d [database_dir] -m [database1,database2] " if (! %opts);
+ next if (! @ds);
-#print Dumper($config->{indexer});
-#print "-" x 70,"\n";
+ my $filename = $webpac->{'current_filename'} || $log->logdie("no current_filename in webpac object");
-Text::Iconv->raise_error(1); # Conversion errors raise exceptions
+ if ($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 $isis_codepage = Text::Iconv->new($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;
+ my $headline = $webpac->{'headline'};
-sub isis2xml {
+ my $f = $filename;
+ $f =~ s!out/!!;
- use xmlify;
+ # save into index
+ foreach my $ds (@ds) {
+ next if (! $ds->{'swish'});
+
+ $index->insert(
+ index_name => $ds->{'tag'},
+ path => $f,
+ headline => $headline,
+ words => join(" ",@{$ds->{'swish'}})
+ );
+ }
- my $row = shift @_;
- my $add_xml = shift @_;
+ # 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(
+ path => $f,
+ headline => $h,
+ );
+ }
+ }
- my $xml;
+# print Dumper(\@ds);
- use parse_format;
+}
- my $html = ""; # html formatted display output
+foreach my $t (keys %{$thes}) {
- my %field_usage; # counter for usage of each field
+ my @e = $thes->{$t}->elements;
+ if (! @e) {
+ $log->logwarn("no elements in sorted index $t?");
+ next;
+ }
- foreach my $field (keys %{$config->{indexer}}) {
+ my $file = "./out/bfilter/$t.txt";
+ $log->info("saving sorted index $t to '$file' [".scalar(@e)." elements]");
- $field_usage{$field}++;
+ $webpac->output_file(
+ file => $file,
+ template => 'index.tt',
+ data => \@e,
+ index_name => $t,
+ );
+}
- my $swish_data = "";
- my $display_data = "";
+if (0 && $log->is_debug) {
+ $log->debug("lookup hash: ",Dumper($webpac->{'lookup'}));
+ $log->debug("data hash: ",Dumper($webpac->{'data'}));
+ foreach my $t (keys %{$thes}) {
+ $log->debug("thesaurus $t hash: ",Dumper($thes->{$t}));
+ }
+}
- foreach my $x (@{$config->{indexer}->{$field}->{isis}}) {
+} # XXX if(0)
- 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 ## $format ##\n";
+$log->debug("lookup hash: ",Dumper($webpac->{'lookup'}));
- $format = $cludge_codepage->convert($format);
- my ($swish,$display) = parse_format($format,$row);
-#print STDERR "s: $swish\nd: $display\n" if ($swish);
+#
+# tree dump implementation which shouldn't be here :-)
+#
+
+
+my $file = 'out/browse.html';
+my $js_url = 'tree-ids.js';
+
+$log->info("creating '$file' for tree html");
+open(HTML, "> $file") || $log->logdie("can't open '$file'");
+
+print HTML qq{
+
+
+ Browse
+
+
+
+
+
+
+
+
+
+ search |
+ thesarus |
+ browse
+
+
+
+
+
+};
+
+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 { defined($l->{"a:".$_[0]."::"}) },
+ child_code => sub { return $_[0] },
+ style => 'display: none',
+ },{
+ # 1
+ code_arr => sub { @{$l->{"a:".$_[0]."::"}} },
+ filter_code => sub { shift }, # nop
+ lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
+ lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
+ lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
+ have_children => sub { defined($l->{"a:".$_[1].":"}) },
+ child_code => sub { return $_[1] },
+ style => 'display: none',
+ },{
+ # 2
+ code_arr => sub { @{$l->{"a:".$_[0].":"}} },
+ filter_code => sub { shift },
+ lookup_v900 => sub { shift },
+ lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
+ lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
+ have_children => sub { defined($l->{"a:".$_[2].":".$_[1]}) },
+ child_code => sub { return "a:".$_[2].":".$_[1] },
+ style => 'display: none',
+ },{
+ # 3 uži pojam
+ code_arr => sub { @{$l->{$_[0]}} },
+ filter_code => sub { shift },
+ lookup_v900 => sub {
+ my ($c,$p) = @_;
+ $p =~ s/^a:(..:....):.*$/$1/;
+ return "a:".$p.":".$c;
+ },
+ lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
+ lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
+ have_children => sub { defined($l->{$_[1]}) },
+ child_code => sub { return $_[1] },
+ },{
+ # 4
+ code_arr => sub { @{$l->{$_[0]}} },
+ filter_code => sub { shift },
+ lookup_v900 => sub {
+ my ($c,$p) = @_;
+ $p =~ s/^a:(..:....):.*$/$1/;
+ return "a:".$p.":".$c;
+ },
+ lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
+ lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
+ have_children => sub { defined($l->{$_[1]}) },
+ child_code => sub { return $_[1] },
+ },{
+ # 5
+ code_arr => sub { @{$l->{$_[0]}} },
+ filter_code => sub { shift },
+ lookup_v900 => sub {
+ my ($c,$p) = @_;
+ $p =~ s/^a:(..:....):.*$/$1/;
+ return "a:".$p.":".$c;
+ },
+ lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
+ lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
+ have_children => sub { defined($l->{$_[1]}) },
+ child_code => sub { return $_[1] },
+ },{
+ # 6
+ code_arr => sub { @{$l->{$_[0]}} },
+ filter_code => sub { shift },
+ lookup_v900 => sub {
+ my ($c,$p) = @_;
+ $p =~ s/^a:(..:....):.*$/$1/;
+ return "a:".$p.":".$c;
+ },
+ lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
+ lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
+ have_children => sub { defined($l->{$_[1]}) },
+ child_code => sub { return $_[1] },
+ },{
+ # 7
+ code_arr => sub { @{$l->{$_[0]}} },
+ filter_code => sub { shift },
+ lookup_v900 => sub {
+ my ($c,$p) = @_;
+ $p =~ s/^a:(..:....):.*$/$1/;
+ return "a:".$p.":".$c;
+ },
+ lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
+ lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
+ have_children => sub { defined($l->{$_[1]}) },
+ child_code => sub { return $_[1] },
+ },{
+ # 8
+ code_arr => sub { @{$l->{$_[0]}} },
+ filter_code => sub { shift },
+ lookup_v900 => sub {
+ my ($c,$p) = @_;
+ $p =~ s/^a:(..:....):.*$/$1/;
+ return "a:".$p.":".$c;
+ },
+ lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
+ lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
+# have_children => sub { defined($l->{$_[1]}) },
+# child_code => sub { return $_[1] },
+ have_children => sub { 0 },
+ child_code => sub { 0 },
+});
+
+my @show_ids;
+my @hide_ids;
+
+unroll(0,'');
+
+$log->debug("test filter: ",$tree[0]->{'filter_code'}->("root:99"));
+
+sub unroll {
+ my ($level,$start_code) = @_;
+
+ $log->logconfess("need level") unless (defined($level));
+
+ # all levels passed?
+ return if (! defined($tree[$level]));
+
+ $log->debug("unroll level $level, start code $start_code");
+
+ foreach my $code ($tree[$level]->{'code_arr'}->($start_code)) {
+
+ if ($code = $tree[$level]->{'filter_code'}->($code)) {
+
+ $log->debug("# $level filter passed code $code");
+
+ my $v900 = $tree[$level]->{'lookup_v900'}->($code,$start_code) || $log->warn("can't lookup_v900($code,$start_code)");
+ $log->debug("# $level lookup_v900($code,$start_code) = $v900");
+
+ my $term = $tree[$level]->{'lookup_term'}->($code,$v900) || $log->warn("can't lookup_term($code,$v900)");
+ $log->debug("# $level lookup_term($code,$v900) = $term");
-#print STDERR "swish: $swish<-- display: $display<--\n";
- # FIX: this is ugly, UGLY, cludge: OpenIsis return
- # UTF8 encoding of strings, but as if source charset
- # is ISO8859-1 and not some other. This breaks our
- # isis character encoding, so we convert it first
- # back to ISO8859-1 (which can actually be different
- # encoding in isis)
+ my $mfn = $tree[$level]->{'lookup_mfn'}->($code,$v900) || $log->warn("can't lookup_mfn($code,$v900)");
+ $log->debug("# $level lookup_mfn($code,$v900) = $mfn");
- $swish_data .= $swish if ($s && $swish);
- $display_data .= $display if ($d && $display);
+ $log->debug("$code -> $v900 : $term [$mfn]");
- # insert into index
- if ($i && $display) {
- my $index_data = $index_codepage->convert($display) || $display;
- $index->insert($field, $index_data, $db_dir);
+ my ($link_start,$link_end) = ('','');
+
+ my $have_children = $tree[$level]->{'have_children'}->($code,$v900,$start_code);
+ if ($have_children) {
+ ($link_start,$link_end) = (qq{},qq{});
+ } else {
+ $log->debug("# $level doesn't have_children($code,$v900,$start_code)");
}
- }
-
-#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);
+ my $mfn_link = "thes/$mfn.html";
+ if (-e "out/$mfn_link") {
+ print HTML " " x $level .
+ qq{- ${link_start}${term}${link_end}}.
+ qq{ »
\n};
} else {
+ $log->warn("file 'out/$mfn_link' doesn't exist, skipping");
+ }
- # 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}."#-#";
+ unless ($have_children) {
+ next;
+ }
+ my $style = $tree[$level]->{'style'};
+
+ print HTML " " x $level .
+ qq{\n \n};
+
+ if ($style) {
+ if ($style =~ m/display\s*:\s*none/i) {
+ push @hide_ids, "id$mfn";
} else {
- $field_name = $config->{indexer}->{$field}->{name}."#-#";
+ push @show_ids, "id$mfn";
}
- if ($field_name) {
- $html .= $xml_codepage->convert($field_name);
- }
- $html .= $display_data."###\n";
+ } else {
+ # default: show
+ push @show_ids, "id$mfn";
}
- }
- if ($swish_data) {
- my $i = Text::Iconv->new($config->{isis_codepage},'ISO8859-2');
- $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},$swish_data));
- }
+ unroll($level+1, $tree[$level]->{'child_code'}->($code,$v900,$start_code));
+
+ print HTML " " x $level . qq{
\n};
- }
-
- # dump formatted output in
- if ($html) {
- $xml .= xmlify("html",$html);
- }
-
- if ($xml) {
-#print STDERR "x: $xml\n";
- $xml .= $add_xml if ($add_xml);
- return "\n$xml\n";
- } else {
- return;
- }
-}
-
-##########################################################################
-
-my $cfg = new Config::IniFiles( -file => $config_file );
-
-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 ;
- }
-
- my $max_rowid = OpenIsis::maxRowid( $db );
-
- print STDERR "Reading database: $isis_db [$max_rowid rows]\n";
-
- my $path = $database; # was $isis_db
-
- my $last_p = 0;
-
-# { 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";
- }
}
}
- print STDERR "\n";
}
-# call this to commit index
-$index->close;
-
-1;
-__END__
-##########################################################################
-
-=head1 NAME
-
-isis2xml.pl - read isis file and dump XML
-
-=head1 DESCRIPTION
+print HTML qq{
+
+
+
+};
+
+close(HTML);
+
+
+my $js_file = "out/$js_url";
+$log->info("creating '$js_file' with arrays of shown and hidden ids");
+open(JS, ">", $js_file) || $log->logdie("can't open '$js_file': $!");
+print JS "var show = ['",join("','",@show_ids),"'];\n";
+print JS "var hide = ['",join("','",@hide_ids),"'];\n";
+close(JS);
-This command will read ISIS data file using OpenIsis perl module and
-create XML file for usage with I
-indexer. Dispite it's name, this script B
-from isis files (isis allready has something like that). Output of this
-script is tailor-made for SWISH-E.
-
-=head1 AUTHOR
-
-Dobrica Pavlinusic
-
-=head1 COPYRIGHT
-
-GNU Public License (GPL) v2 or later
-
-=head1 SEE ALSO
-
-SWISH-E web site at http://www.swish-e.org
+$log->info("closing index");
+$index->close;
-=cut
+$log->info("elapsed time: ",$webpac->fmt_time(time() - $webpac->{'start_t'}));