--- trunk2/all2all.pl 2004/09/11 23:32:41 425
+++ trunk2/all2all.pl 2004/10/30 23:04:37 560
@@ -7,19 +7,96 @@
=cut
use strict;
-use Data::Dumper;
+use locale;
+use YAML;
use Carp;
+use Getopt::Long;
+use Text::Unaccent 1.02;
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,
+);
+
+my $filter = {
+ 'CROVOC' => sub {
+ my $tmp = shift || return;
+ return undef unless ($tmp =~ s/\s*CROVOC.*$/ #C#/);
+ # remove repeating stars
+ # FIXME this should be fixed at right place, not here!
+ $tmp =~ s/(\s*#C#)+/ #C#/g;
+ return $tmp;
+ },
+ 'CROVOC_tree' => sub {
+ my $tmp = shift || return;
+ $tmp =~ s/\s*CROVOC.*$/ #C#/;
+ $tmp =~ s/\s*EUROVOC.*//;
+ return $tmp;
+ },
+ # TT filter
+ 'CROVOC_img' => sub {
+ my $tmp = shift;
+ $tmp =~ s/\s*#C#\s*/ /gis;
+ return $tmp;
+ }
+};
+
+## remove accented characters
+#
+sub unac {
+ my $string = shift || return;
+ $string = unac_string($code_page,$string);
+ $string =~ tr/ðÐ/dD/;
+ $string = unac_entities($string);
+ return $string;
+}
+sub unac_2 {
+ my $string = shift || return;
+ if (length($string) > 2) {
+ my $pr = substr($string,0,2);
+ $string = unac_string($code_page,substr($string,2));
+ $string =~ tr/ðÐ/dD/;
+ $string = $pr . $string;
+ }
+ $string = unac_entities($string);
+ return lc($string);
+}
+
+sub unac_entities {
+ my $ent = shift || return;
+
+ $ent =~ s/&(\w)(acute|cedil|circ|grave|ring|slash|tilde|uml);/$1/gi;
+ $ent =~ s/ð/d/g;
+ $ent =~ s/&E[tT][hH];/D/g;
+ $ent =~ s/&(\w\w)lig;/$1/gi;
+
+ return $ent;
+}
+
+# create WebPAC object
+#
my $webpac = new WebPAC(
- code_page => 'ISO-8859-2',
- limit_mfn => 500,
-# debug => 1,
-# low_mem => 1,
+ code_page => $code_page,
+ limit_mfn => $limit_mfn,
+ start_mfn => $start_mfn,
+ debug => $debug,
+ low_mem => $low_mem,
+ filter => $filter,
) || die;
my $log = $webpac->_get_logger() || die "can't get logger";
@@ -27,8 +104,8 @@
$log->debug("creating WebPAC::jsFind object");
my $index = new WebPAC::jsFind(
- index_path => './out/index',
- keys => 10,
+ index_path => $index_path,
+ keys => 62,
) || die;
my $thes;
@@ -38,15 +115,16 @@
my $maxmfn = $webpac->open_isis(
filename => shift @ARGV || '/data/hidra/THS/THS',
lookup => [
- { 'key' => 'd:v900', 'val' => 'v250^a' },
+ { 'key' => 'd:v900', 'val' => 'filter{CROVOC_tree}v250^a v800' },
# { 'eval' => '"v901^a" eq "Podruèje"', 'key' => 'pa:v561^4:v562^4:v461^1', 'val' => 'v900' },
# { 'eval '=> '"v901^a" eq "Mikrotezaurus"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' },
# { 'eval' => '"v901^a" eq "Deskriptor"', 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' },
{ 'key' => 'a:v561^4:v562^4:v461^1', 'val' => 'v900' },
{ 'key' => '900_mfn:v900', 'val' => 'v000' },
# tree structure
- { 'eval' => 'length("v251") == 2', 'key' => 'root:v251', 'val' => 'v900' },
+ { 'eval' => 'length("v251") == 2 && "v800" =~ m/EUROVOC/ || "v800" =~ m/CROVOC/ && "v251" =~ m/^(H|HD|L|Z|P)$/', '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' },
],
);
@@ -63,8 +141,8 @@
my @ds = $webpac->data_structure($rec);
if (0 && $log->is_debug) {
- $log->debug("rec = ",Dumper($rec));
- $log->debug("ds = ",Dumper(\@ds));
+ $log->debug("rec = ",Dump($rec));
+ $log->debug("ds = ",Dump(\@ds));
}
next if (! @ds);
@@ -95,11 +173,26 @@
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;
+ $words =~ s/\W+$//;
+
+ # first try to generate headline for this entry from index
+ my $h = $ds->{'index'}->[0];
+ # then, from display
+ $h ||= $ds->{'display'}->[0];
+ # and as last resport, fallback to headline
+ $h ||= $headline;
+
$index->insert(
index_name => $ds->{'tag'},
- path => $f,
- headline => $headline,
- words => join(" ",@{$ds->{'swish'}})
+ #path => $f,
+ path => $webpac->mfn,
+ headline => $h,
+ words => unac($words),
);
}
@@ -107,17 +200,18 @@
foreach my $ds (@ds) {
next if (! $ds->{'index'});
- $thes->{$ds->{'tag'}} ||= new WebPAC::Index;
+ $thes->{$ds->{'tag'}} ||= new WebPAC::Index( name => $ds->{'tag'} );
foreach my $h (@{$ds->{'index'}}) {
$thes->{$ds->{'tag'}}->insert(
- path => $f,
+ sort_by => unac_2($h),
+ mfn => $webpac->mfn,
headline => $h,
);
}
}
-# print Dumper(\@ds);
+# print Dump(\@ds);
}
@@ -141,35 +235,23 @@
}
if (0 && $log->is_debug) {
- $log->debug("lookup hash: ",Dumper($webpac->{'lookup'}));
- $log->debug("data hash: ",Dumper($webpac->{'data'}));
+ $log->debug("lookup hash: ",Dump($webpac->{'lookup'}));
+ $log->debug("data hash: ",Dump($webpac->{'data'}));
foreach my $t (keys %{$thes}) {
- $log->debug("thesaurus $t hash: ",Dumper($thes->{$t}));
+ $log->debug("thesaurus $t hash: ",Dump($thes->{$t}));
}
}
} # XXX if(0)
-$log->info("lookup hash: ",Dumper($webpac->{'lookup'}));
+#$log->debug("lookup hash: ",Dump($webpac->{'lookup'}));
+
+$log->info("creating tree");
#
-# tree dump implementation which shouldn't be here :-)
+# define tree structure
#
-my $file = 'out/browse.html';
-open(HTML, "> $file") || $log->logdie("can't open '$file'");
-
-print HTML qq{
-
-
- Browse
-
-
-
-
-
-};
-
my $l = $webpac->{'lookup'} || $log->logconfess("can't find lookup");
my @tree = ({
@@ -182,111 +264,114 @@
lookup_v900 => sub { shift @{$l->{"root:".$_[0]}} },
lookup_term => sub { shift @{$l->{"d:".$_[1]}} },
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[1]}} },
- have_children => sub { $l->{"a:".$_[0]."::" } },
- child_code => sub { return $_[0] },
+ 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 { @{$l->{"a:".$_[0]."::"}} },
+ 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 { $l->{"a:".$_[1].":" } },
- child_code => sub { return $_[1] },
+ have_children => sub { return $l->{"a:".$_[1].":"} },
style => 'display: none',
},{
# 2
- code_arr => sub { @{$l->{"a:".$_[0].":"}} },
+ code_arr => sub { @{$_[0]} },
filter_code => sub { shift },
- lookup_v900 => sub { shift },
+ lookup_v900 => sub { shift @{$l->{"code:".$_[0]}} },
lookup_term => sub { shift @{$l->{"d:".$_[0]}} },
lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[0]}} },
- have_children => sub { $l->{"a:".$_[2].":".$_[1]} },
- child_code => sub { return "a:".$_[2].":".$_[1] },
- style => 'display: none',
+ have_children => sub { return $l->{"a:".$_[1].":".$_[0]} },
+ #style => 'display: none',
},{
# 3 u¾i pojam
- code_arr => sub { @{$l->{$_[0]}} },
+ code_arr => sub { @{$_[0]} },
filter_code => sub { shift },
- lookup_v900 => sub { shift },
- lookup_term => sub { shift @{$l->{"d:".$_[1]}} },
- lookup_mfn => sub { shift @{$l->{"900_mfn:".$_[1]}} },
+ 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 },
- child_code => sub { 0 },
+ have_children_at_level => sub { defined($l->{"a:".$_[1].":".$_[0]}) && return (9,$l->{"a:".$_[1].":".$_[0]}) },
+ },{
});
-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) || $log->logdie("can't lookup_v900 '$code'");
- $log->debug("# $level lookup_v900($code) = $v900");
- my $term = $tree[$level]->{'lookup_term'}->($code,$v900) || $log->logdie("can't lookup_term '$v900'");
- $log->debug("# $level lookup_term($code,$v900) = $term");
- my $mfn = $tree[$level]->{'lookup_mfn'}->($code,$v900) || $log->logdie("can't lookup_mfn '$v900'");
- $log->debug("# $level lookup_mfn($code,$v900) = $mfn");
-
- $log->debug("$code -> $v900 : $term [$mfn]");
-
-
- my $have_children = $tree[$level]->{'have_children'}->($code,$v900,$start_code);
-
- my ($link_start,$link_end) = ('','');
-
- ($link_start,$link_end) = (qq{},qq{}) if ($have_children);
-
- 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");
- }
-
- unless ($have_children) {
- $log->warn("doesn't have_children($code,$v900,$start_code)");
- next;
- }
-
- print HTML " " x $level .
- qq{\n {'style'} ? ' style="'.$tree[$level]->{'style'}.'"' : '').
- qq{>\n};
-
- unroll($level+1, $tree[$level]->{'child_code'}->($code,$v900,$start_code));
-
- print HTML " " x $level . qq{
\n};
+my $tree = new WebPAC::Tree(
+ tree => \@tree,
+);
- }
- }
-}
+$tree->output(
+ dir => './out',
+ html => 'browse.html',
+ template => './output_template/tree.tt',
+ js => 'tree-ids.js',
+);
-print HTML qq{
-
-
-
-};
+$tree->output(
+ dir => './eurovoc',
+ html => 'hijerarhija.html',
+ template => './output_template/hijerarhija.tt',
+ js => 'tree-ids.js',
+);
-close(HTML);
+$log->info("closing index");
$index->close;
$log->info("elapsed time: ",$webpac->fmt_time(time() - $webpac->{'start_t'}));