--- trunk/all2xml.pl 2003/03/22 22:43:05 43
+++ trunk2/all2all.pl 2004/09/12 17:21:47 429
@@ -1,299 +1,409 @@
#!/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;
-use Encode;
-
-$|=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
+=head1 NAME
-my %opts;
+all2all.pl - basic script for all WebPAC needs
-# usage:
-# -d directory name
-# -m multiple directories
-# -q quiet
-# -s run swish
+=cut
-getopts('d:m:qs', \%opts);
+use strict;
+use Data::Dumper;
+use Carp;
-my $path; # this is name of database
+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;
-Text::Iconv->raise_error(1); # Conversion errors raise exceptions
+my $thes;
-# this is encoding of all files on disk, including import_xml/*.xml file and
-# filter/*.pm files! It will be used to store strings in perl internally!
-my $codepage = 'ISO-8859-2';
+$|=1;
-my $utf2cp = Text::Iconv->new('UTF-8',$codepage);
-# this function will convert data from XML files to local encoding
-sub x {
- return $utf2cp->convert($_[0]);
-}
+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));
+ }
-# decode isis import codepage
-my $isis2cp;
+ next if (! @ds);
-# outgoing xml must be in UTF-8
-my $cp2utf = Text::Iconv->new($codepage,'UTF-8');
+ my $filename = $webpac->{'current_filename'} || $log->logdie("no current_filename in webpac object");
-sub isis2xml {
+ if ($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'},
+ );
+ }
- use xmlify;
+ my $headline = $webpac->{'headline'};
- my $row = shift @_;
- my $add_xml = shift @_;
+ my $f = $filename;
+ $f =~ s!out/!!;
- my $xml;
+ # 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'}})
+ );
+ }
- use parse_format;
+ # 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 $html = ""; # html formatted display output
+# print Dumper(\@ds);
- my %field_usage; # counter for usage of each field
+}
- # sort subrouting using order="" attribute
- sub by_order {
- return 0 if (! $config->{indexer}->{$a}->{order});
- return 0 if (! $config->{indexer}->{$b}->{order});
+foreach my $t (keys %{$thes}) {
- return $config->{indexer}->{$a}->{order} <=>
- $config->{indexer}->{$b}->{order} ;
+ my @e = $thes->{$t}->elements;
+ if (! @e) {
+ $log->logwarn("no elements in sorted index $t?");
+ next;
}
- foreach my $field (sort by_order keys %{$config->{indexer}}) {
+ my $file = "./out/bfilter/$t.txt";
+ $log->info("saving sorted index $t to '$file' [".scalar(@e)." elements]");
- $field=x($field);
-
- $field_usage{$field}++;
+ $webpac->output_file(
+ file => $file,
+ template => 'index.tt',
+ data => \@e,
+ index_name => $t,
+ );
+}
- my $swish_data = "";
- my $display_data = "";
- my $line_delimiter;
+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->{$t}));
+ }
+}
- my ($swish,$display);
+} # XXX if(0)
- foreach my $x (@{$config->{indexer}->{$field}->{isis}}) {
+$log->debug("lookup hash: ",Dumper($webpac->{'lookup'}));
- my $format = x($x->{content});
- my $delimiter = x($x->{delimiter}) || ' ';
+#
+# 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
+
+
+
+
+
+
+
+
+
+
+
+
+};
+
+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");
- my $isis_i = 0; # isis repeatable offset
+ my $mfn = $tree[$level]->{'lookup_mfn'}->($code,$v900) || $log->warn("can't lookup_mfn($code,$v900)");
+ $log->debug("# $level lookup_mfn($code,$v900) = $mfn");
- my ($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");
+ $log->debug("$code -> $v900 : $term [$mfn]");
- # what will separate last line from this one?
- if ($display_data && $x->{append} && $x->{append} eq "1") {
- $line_delimiter = ' ';
- } elsif ($display_data) {
- $line_delimiter = '
';
+ my ($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)");
}
- # init vars so that we go into while...
- ($swish,$display) = (1,1);
-
- while ($swish || $display) {
- ($swish,$display) = parse_format($format,$row,$isis_i++,$isis2cp);
-
- # filter="name" ; filter this field through
- # filter/[name].pm
- my $filter = $x->{filter};
- if ($filter) {
- require "filter/".$filter.".pm";
- }
- # type="swish" ; field for swish
- if ($s && $swish) {
- if ($filter) {
- no strict 'refs';
- $swish_data .= join(" ",&$filter($swish));
- } else {
- $swish_data .= $swish;
- }
- }
-
- # type="display" ; field for display
- if ($d && $display) {
- if ($line_delimiter && $display_data) {
- $display_data .= $line_delimiter;
- undef $line_delimiter;
- }
- if ($filter) {
- no strict 'refs';
- $display_data .= join($delimiter,&$filter($display));
- } else {
- if ($display_data) {
- $display_data .= $delimiter.$display;
- } else {
- $display_data .= $display;
- }
- }
- }
-
- # type="index" ; insert into index
- if ($i && $display) {
- my $index_data = $display;
- if ($filter) {
- no strict 'refs';
- foreach my $d (&$filter($index_data)) {
- $index->insert($field, $d, $path);
- }
- } else {
- $index->insert($field, $index_data, $path);
- }
- }
+ 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");
}
- }
-
- if ($display_data) {
-
- if ($field eq "headline") {
- $xml .= xmlify("headline", $display_data);
- } else {
+ unless ($have_children) {
+ next;
+ }
+ my $style = $tree[$level]->{'style'};
- # 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}."#-#";
+ print HTML " " x $level .
+ qq{\n \n};
+
+ if ($style) {
+ if ($style =~ m/display\s*:\s*none/i) {
+ push @hide_ids, "mfn$mfn";
} else {
- print STDERR "WARNING: field '$field' doesn't have 'name' attribute!";
+ push @show_ids, "mfn$mfn";
}
- if ($field_name) {
- $html .= x($field_name);
- }
- $html .= $display_data."###\n";
- }
- }
- if ($swish_data) {
- # remove extra spaces
- $swish_data =~ s/ +/ /g;
- $swish_data =~ s/ +$//g;
-
- $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
- }
-
-
- }
-
- # dump formatted output in
- if ($html) {
- $xml .= xmlify("html",$html);
- }
-
- if ($xml) {
- $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') || die "$database doesn't have 'isis_db' defined!";
- my $type = $cfg -> val($database, 'type') || die "$database doesn't have 'type' defined";
- my $add_xml = $cfg -> val($database, 'xml'); # optional
-
- $config=XMLin("./import_xml/$type.xml", forcearray => [ 'isis' ], forcecontent => 1);
-
- $isis2cp = Text::Iconv->new($config->{isis_codepage},$codepage);
-
- 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}) {
- # 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 = $p;
+ } else {
+ # default: show
+ push @show_ids, "mfn$mfn";
}
- my $swishpath = $path."#".int($row->{mfn});
+ unroll($level+1, $tree[$level]->{'child_code'}->($code,$v900,$start_code));
+
+ print HTML " " x $level . qq{
\n};
- if (my $xml = $cp2utf->convert(isis2xml($row,$add_xml."$swishpath"))) {
- use bytes; # as opposed to chars
- print "Path-Name: $swishpath\n";
- print "Content-Length: ".(length($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
-
-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
+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': $!");
+print JS "var show = ['",join("','",@show_ids),"'];\n";
+print JS "var hide = ['",join("','",@hide_ids),"'];\n";
+close(JS);
-=head1 COPYRIGHT
-
-GNU Public License (GPL) v2 or later
-
-=head1 SEE ALSO
-
-SWISH-E web site at http://www.swish-e.org
+$log->info("closing index");
+$index->close;
-=cut
+$log->info("elapsed time: ",$webpac->fmt_time(time() - $webpac->{'start_t'}));