--- trunk/all2xml.pl	2003/01/11 19:55:30	9
+++ trunk2/all2all.pl	2004/09/26 15:11:36	471
@@ -1,250 +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,
-require Unicode::Map8;
-use DBI;
-
-my $config=XMLin(undef, forcearray => [ 'isis' ], forcecontent => 1);
-my $dbh = DBI->connect("DBI:Pg:dbname=webpac","","") || die $DBI::errstr; # FIX
-# FIX; select relname from pg_class where relname like 'index_%' ;
-$dbh->begin_work || die $dbh->errstr();
-
-$dbh->do("delete from index_author") || die $dbh->errstr();
-$dbh->do("delete from index_title") || die $dbh->errstr();
-
-my %opts;
-
-# usage:
-# 	-d directory name
-# 	-m multiple directories
-# 	-q quiet
-# 	-s run swish
-
-getopts('d:m:qs', \%opts);
-
-my $db_dir = $opts{d} || "ps";	# FIX
-
-#die "usage: $0 -d [database_dir] -m [database1,database2] " if (! %opts);
-
-#print Dumper($config->{indexer});
-#print "-" x 70,"\n";
-
-# how to convert isis code page to UTF8?
-my $isis_map = Unicode::Map8->new($config->{isis_codepage}) || die;
-
-sub isis2xml {
-
-	my $row = shift @_;
-
-	my $xml;
-	$xml->{db_dir} = [ $db_dir ];	# FIX remove?
-
-	sub isis_sf {
-		my $row = shift @_;
-		my $isis_id = shift @_;
-		my $subfield = shift @_;
-		if ($row->{$isis_id}->[0]) {
-			my $sf = OpenIsis::subfields($row->{$isis_id}->[0]);
-			if (! defined $subfield || length($subfield) == 0) {
-				# subfield list undef, empty or no defined subfields for this record
-				my $all_sf = $row->{$isis_id}->[0];
-				$all_sf =~ s/\^./ /g;	nuke definirions
-				return $all_sf;
-			} elsif ($sf->{$subfield}) {
-				return $sf->{$subfield};
-			}
-		}
-	}
-
-	foreach my $field (keys %{$config->{indexer}}) {
+=head1 NAME
 
-		my $display_data = "";
-		my $swish_data = "";
-		my $index_data = "";
-
-		foreach my $x (@{$config->{indexer}->{$field}->{isis}}) {
-
-			my $display_tmp = "";
-			my $swish_tmp = "";
-			my $index_tmp = "";
-
-			my $format = $x->{content};
-			my $s = 1;	# swish only
-			my $d = 1;	# display only
-			my $i = 0;	# index only
-			$s = 0 if (lc($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";	
-			# parse format
-			my $prefix = "";
-			if ($format =~ s/^([^\d]+)//) {
-				$prefix = $1;
-			}
-			while ($format) {
-				if ($format =~ s/^(\d\d\d)(\w?)//) {
-					my $isis_tmp = isis_sf($row,$1,$2);
-					if ($isis_tmp) {
-#						$display_tmp .= $prefix . "/$1/$2/".$isis_tmp if ($d);
-						$display_tmp .= $prefix . $isis_tmp if ($d);
-						$swish_tmp .= $isis_tmp." " if ($s);
-						$index_tmp .= $prefix . $isis_tmp if ($i);
-#print STDERR " $isis_tmp <--\n";
-					}
-					$prefix = "";
-				} elsif ($format =~ s/^([^\d]+)//) {
-					$prefix = $1;
-				} else {
-					print STDERR "WARNING: unparsed format '$format'\n";
-					last;
-				};
-			}
-			# add suffix
-			$display_tmp .= $prefix if ($display_tmp);
-			$index_tmp .= $prefix if ($index_tmp);
-
-#			$display_data .= $display_tmp if ($display_tmp ne "");
-#			$swish_data .= $swish_tmp if ($swish_tmp ne "");
-			$display_data .= $display_tmp;
-			$swish_data .= $swish_tmp;
-			$index_data .= $index_tmp;
+all2all.pl - basic script for all WebPAC needs
 
-		}
-#print "--display:$display_data\n--swish:$swish_data\n";
-		#$xml->{$field."_display"} = $isis_map->tou($display_data)->utf8 if ($display_data);
-		#$xml->{$field."_swish"} = unac_string($config->{isis_codepage},$swish_data) if ($swish_data);
-		$xml->{$field."_display" } = [ $isis_map->tou($display_data)->utf8 ] if ($display_data);
-		$xml->{$field."_swish"} = [ unac_string($config->{isis_codepage},$swish_data) ] if ($swish_data);
-
-		# index
-		if ($index_data && $index_data ne "") {
-			my $sql = "select $field from index_$field where upper($field)=upper(?)";
-			my $sth = $dbh->prepare($sql) || die $dbh->errstr();
-			$sth->execute($index_data) || die "SQL: $sql; ".$dbh->errstr();
-#print STDERR "--->$index_data<---\n";
-			if (! $sth->fetchrow_hashref) {
-				my $sql = "insert into index_$field values (?)";
-				my $sth = $dbh->prepare($sql) || die $dbh->errstr();
-#print STDERR "$sql: $index_data<!----\n";
-				$sth->execute($index_data) || die "SQL: $sql; ".$dbh->errstr();
-			}
-		}
+=cut
 
+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;
+
+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) {
-		return XMLout($xml, rootname => 'xml', keeproot => 0, noattr => 0 );
-	} else {
-		return;
-	}
-}
 
-##########################################################################
+	next if (! @ds);
 
-my $last_tell=0;
+	my $filename = $webpac->{'current_filename'} || $log->logdie("no current_filename in webpac object");
 
-my @isis_dirs = ( '.' );	# use dirname as database name
-
-if ($opts{m}) {
-	@isis_dirs = split(/,/,$opts{m});
-}
-
-my @isis_dbs;
-
-foreach (@isis_dirs) {
-	if (-e $config->{isis_data}."/$db_dir/$_/LIBRI") {
-		push @isis_dbs,$config->{isis_data}."/$db_dir/$_/LIBRI/LIBRI";
-	}
-	if (-e $config->{isis_data}."/$db_dir/$_/PERI") {
-		push @isis_dbs,$config->{isis_data}."/$db_dir/$_/PERI/PERI";
-	}
-	if (-e $config->{isis_data}."/$db_dir/$_/AMS") {
-		push @isis_dbs,$config->{isis_data}."/$db_dir/$_/AMS/AMS";
-	}
-	if (-e $config->{isis_data}."/$db_dir/$_/ARTI") {
-#		push @isis_dbs,$config->{isis_data}."/$db_dir/$_/ARTI/ARTI";
+	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 => $webpac->{'headline'},
+		);
 	}
-}
-
-print STDERR "FATAL: Can't find isis database.\nPerhaps isis_data (".$config->{isis_data}.") has wrong value?\n" if (! @isis_dbs);
-
-my $db;
 
-foreach my $isis_db (@isis_dbs) {
+	my $headline = $webpac->{'headline'};
 
+	my $f = $filename;
+	$f =~ s!out/!!;
 
-	my $db = OpenIsis::open( $isis_db );
-	if (0) {
-#	# FIX
-#	if (! $db ) {
-		print STDERR "WARNING: can't open '$isis_db'\n";
-		next ;
+	# 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 $max_rowid = OpenIsis::maxRowid( $db );
-
-	print STDERR "Reading database: $isis_db [$max_rowid rows]\n";
-
-	my $last_p = 0;
-
-#	{ my $row_id = 1;
-# FIX
-	for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) {
-		my $row = OpenIsis::read( $db, $row_id );
-		if ($row && $row->{mfn}) {
-
-			# 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)) {
-				my $path = $isis_db;
-				$path =~ s#$config->{isis_data}/*##g;
-				my $out = "Path-Name: $path#".$row->{mfn}."\n";
-				$out .= "Content-Length: ".(length($xml)+1)."\n";
-				$out .= "Document-Type: XML\n\n$xml\n";
-				print $out;
-			}
+	# 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";
-}
 
-$dbh->commit || die $dbh->errstr();
+#	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<SWISH-E>
-indexer. Dispite it's name, this script B<isn't general xml generator>
-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 <dpavlin@rot13.org>
+} # 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'}));