--- trunk/all2xml.pl 2003/02/22 21:33:04 17
+++ trunk2/all2all.pl 2004/09/11 23:32:41 425
@@ -1,241 +1,292 @@
#!/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);
-
-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;
+=head1 NAME
- my $row = shift @_;
- my $add_xml = shift @_;
+all2all.pl - basic script for all WebPAC needs
- my $xml;
+=cut
- use parse_format;
+use strict;
+use Data::Dumper;
+use Carp;
- my $html = ""; # html formatted display output
+use lib './lib';
+use WebPAC;
+use WebPAC::jsFind;
+use WebPAC::Index;
+
+my $webpac = new WebPAC(
+ code_page => 'ISO-8859-2',
+ limit_mfn => 500,
+# debug => 1,
+# low_mem => 1,
+) || die;
+
+my $log = $webpac->_get_logger() || die "can't get logger";
+
+$log->debug("creating WebPAC::jsFind object");
+
+my $index = new WebPAC::jsFind(
+ index_path => './out/index',
+ keys => 10,
+) || die;
- my %field_usage; # counter for usage of each field
+my $thes;
- foreach my $field (keys %{$config->{indexer}}) {
+$|=1;
- $field_usage{$field}++;
+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));
+ }
- my $swish_data = "";
- my $display_data = "";
+ next if (! @ds);
- foreach my $x (@{$config->{indexer}->{$field}->{isis}}) {
+ my $filename = $webpac->{'current_filename'} || $log->logdie("no current_filename in webpac object");
- 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 ## $format ##\n";
+ if ($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'},
+ );
+ }
- $format = $cludge_codepage->convert($format);
- my ($swish,$display) = parse_format($format,$row);
-#print STDERR "s: $swish\nd: $display\n" if ($swish);
+ my $headline = $webpac->{'headline'};
-#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 $f = $filename;
+ $f =~ s!out/!!;
- $swish_data .= $swish if ($s && $swish);
- $display_data .= $display if ($d && $display);
+ # 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'}})
+ );
+ }
- # insert into index
- if ($i && $display) {
- my $index_data = $index_codepage->convert($display) || $display;
- $index->insert($field, $index_data, $db_dir);
- }
+ # 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,
+ );
}
+ }
+# print Dumper(\@ds);
-#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 {
+foreach my $t (keys %{$thes}) {
- # 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}."#-#";
- } else {
- $field_name = $config->{indexer}->{$field}->{name}."#-#";
- }
- 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');
- $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));
- }
+ my @e = $thes->{$t}->elements;
+ if (! @e) {
+ $log->logwarn("no elements in sorted index $t?");
+ next;
+ }
+ my $file = "./out/bfilter/$t.txt";
+ $log->info("saving sorted index $t to '$file' [".scalar(@e)." elements]");
- }
+ $webpac->output_file(
+ file => $file,
+ template => 'index.tt',
+ data => \@e,
+ index_name => $t,
+ );
+}
- # dump formatted output in
- if ($html) {
- $xml .= xmlify("html",$html);
- }
-
- if ($xml) {
-#print STDERR "x: $xml\n";
- $xml .= $add_xml if ($add_xml);
- return "