| 1 |
#!/usr/bin/perl -w |
| 2 |
|
| 3 |
# indexer, Dobrica Pavlinusic <dpavlin@rot13.org> 2002-06-19 |
| 4 |
# options: -q quiet |
| 5 |
# -d debug |
| 6 |
# -v verbose |
| 7 |
# -l limit regex |
| 8 |
|
| 9 |
# This indexer output xml data which is used to index content with |
| 10 |
# swish-e 2.2, http://www.swish-e.org/ |
| 11 |
# |
| 12 |
# xml is output is on STDOUT and informational oputput (for humas) is |
| 13 |
# on STDERR |
| 14 |
# |
| 15 |
|
| 16 |
use strict; |
| 17 |
use Getopt::Std; |
| 18 |
use Lingua::Spelling::Alternative; |
| 19 |
require Unicode::Map8; |
| 20 |
|
| 21 |
my $sadrzaj=0; |
| 22 |
my $nr=0; |
| 23 |
my $naslov=""; |
| 24 |
|
| 25 |
my $br; ## broj NN |
| 26 |
my $god; ## godina NN |
| 27 |
my $aname; ## ancor name na originalnim stranicama |
| 28 |
|
| 29 |
my $nn_dir="../"; # dir u kojem su wget-ani fileovi |
| 30 |
my $url="http://www.nn.hr/CijeliBrojS.asp?god=%d&br=%s&mid=%s#%d"; |
| 31 |
|
| 32 |
my %opts; |
| 33 |
getopts("vqdl:", \%opts); |
| 34 |
|
| 35 |
my $brojeva=0; |
| 36 |
my $zakona=0; |
| 37 |
|
| 38 |
my $hr = new Lingua::Spelling::Alternative( DEBUG => $opts{d} ); |
| 39 |
#$hr->load_affix("$nn_dir/search/croatian.aff"); |
| 40 |
$hr->load_findaffix("$nn_dir/prvih_50.txt"); |
| 41 |
|
| 42 |
my $l2_map = Unicode::Map8->new("ISO-8859-2") || die; |
| 43 |
|
| 44 |
#-------------------------------------------------------------------- |
| 45 |
sub dump_to_swish { |
| 46 |
my $xml = shift @_; |
| 47 |
my ($god,$br,$nr,$aname) = @_; |
| 48 |
|
| 49 |
use utf8; |
| 50 |
|
| 51 |
print "Path-Name: ".sprintf($url,$god,$br,$nr,$aname)."\n". |
| 52 |
"Content-Length: ".length($xml)."\n". |
| 53 |
"Document-Type: XML\n". |
| 54 |
"\n$xml"; |
| 55 |
} |
| 56 |
|
| 57 |
#-------------------------------------------------------------------- |
| 58 |
|
| 59 |
|
| 60 |
opendir(DIR,$nn_dir) || warn "opendir: $!"; |
| 61 |
my @files; |
| 62 |
if ($opts{l}) { |
| 63 |
# add limit regex |
| 64 |
@files = grep { /^CijeliBrojS/ && /$opts{l}/ && -f "$nn_dir/$_" } readdir(DIR); |
| 65 |
print STDERR "Using limit regex which is '$opts{l}'\n"; |
| 66 |
} else { |
| 67 |
@files = grep { /^CijeliBrojS/ && -f "$nn_dir/$_" } readdir(DIR); |
| 68 |
} |
| 69 |
closedir(DIR); |
| 70 |
|
| 71 |
foreach my $file (@files) { |
| 72 |
open(IN,"$nn_dir/$file") || die "can't open '$nn_dir/$file': $!"; |
| 73 |
|
| 74 |
if ($file=~m/god=(\d+)\&br=(\d+)/) { |
| 75 |
($br,$god) = ($2,$1); |
| 76 |
print STDERR "$file -- $2 -- $1\n" if (! $opts{q}); |
| 77 |
$brojeva++; |
| 78 |
} |
| 79 |
|
| 80 |
while(<IN>) { |
| 81 |
chomp; |
| 82 |
s/\015//g; # kill cr |
| 83 |
tr/ðèæÐÈÆ/¹ð¾èæ©Ð®ÈÆ/; # 1250 -> iso8859-2 |
| 84 |
|
| 85 |
if (m,<div class=sadrzaj>,) { |
| 86 |
$sadrzaj++; |
| 87 |
next; |
| 88 |
} |
| 89 |
|
| 90 |
if ($sadrzaj && m,</div>,) { |
| 91 |
$sadrzaj--; |
| 92 |
$naslov=~s/\s+/ /g; |
| 93 |
$naslov=~s/<[^>]+>//g; |
| 94 |
$naslov=~s/^\s+//g; |
| 95 |
$naslov=~s/\s+$//g; |
| 96 |
print STDERR "$god $br $nr: $naslov\n" if ($opts{v}); |
| 97 |
my $naslov_czs = lc($naslov); |
| 98 |
$naslov_czs =~ tr/¹©ðÐèÈæƾ®/sSdDcCcCzZ/; |
| 99 |
$naslov_czs =~ tr/a-zA-Z/ /cs; # non a-z -> space |
| 100 |
$naslov_czs = join(" ",$hr->alternatives(split(/ /,$naslov_czs))); |
| 101 |
# $naslov_czs = $hr->minimal(split(/ /,$naslov_czs)); |
| 102 |
my $xml="<nn>\n<br>$br</br>\n<god>$god</god>\n<nr>$nr</nr>\n<aname>$aname</aname>\n"; |
| 103 |
my $naslov_utf=$l2_map->tou($naslov)->utf8; |
| 104 |
|
| 105 |
# Escape <, >, & and ", and to produce valid XML |
| 106 |
my %escape = ('<'=>'<', '>'=>'>', '&'=>'&', '"'=>'"'); |
| 107 |
my $escape_re = join '|' => keys %escape; |
| 108 |
$naslov_utf =~ s/($escape_re)/$escape{$1}/g; |
| 109 |
|
| 110 |
$xml.="<naslov>$naslov_utf</naslov>\n"; |
| 111 |
$xml.="<naslov_czs>$naslov_czs</naslov_czs>\n</nn>\n\n"; |
| 112 |
dump_to_swish($xml,$god,$br,$nr,$aname); |
| 113 |
|
| 114 |
$naslov=""; |
| 115 |
$nr=0; |
| 116 |
$zakona++; |
| 117 |
} |
| 118 |
|
| 119 |
if ($sadrzaj) { |
| 120 |
if (s/<a href="#([^"]+)">\s*(\S+)\.\s*<[^>]+>//i) { |
| 121 |
($aname,$nr) = ($1,$2); |
| 122 |
} elsif (s/<a href="Javascript:Mojdok\((\d+),(\d+),'*(\w+)'*,(\d+)\)[^>]*>//i) { |
| 123 |
($nr,$aname) = ($3,$4); |
| 124 |
die "conflict in godina: $1 != $god" if ($god != $1); |
| 125 |
die "conflict in broj: $2 != $br" if ($br != $2); |
| 126 |
} else { |
| 127 |
die "can't find nr in line: $_"; |
| 128 |
} |
| 129 |
$naslov.=$_; |
| 130 |
$naslov=~s/^\s*$nr\.*\s*//g; |
| 131 |
} |
| 132 |
|
| 133 |
} |
| 134 |
|
| 135 |
close(IN); |
| 136 |
} |
| 137 |
|
| 138 |
print STDERR "Ukupno $brojeva brojeva NN, sa $zakona zakona...\n" if (! $opts{q}); |
| 139 |
|
Parent Directory
| 
Revision Log