/[swish]/trunk/spider/filter.pm
This is repository of my old source code which isn't updated any more. Go to git.rot13.org for current projects!
ViewVC logotype

Contents of /trunk/spider/filter.pm

Parent Directory Parent Directory | Revision Log Revision Log


Revision 85 - (show annotations)
Mon Aug 30 11:14:24 2004 UTC (19 years, 6 months ago) by dpavlin
File size: 5241 byte(s)
extract metadata for LJ

1 sub filter {
2 my $contents = shift || return;
3
4 my $verbose = 0;
5
6 # if you don't want content to be indexed, include it in
7 # <noindex> foobar </noindex> tags or surround it with comments
8 # <!-- noindex --> foobar <!-- /noindex -->
9 # <!-- noindex --> foobar <!-- index --> (also supported by swish)
10 $contents =~ s,<noindex>.+?</noindex>,,isg;
11 $contents =~ s,<!--\s*noindex\s*-->.+?<!--\s*/noindex\s*-->,,isg;
12 $contents =~ s,<!--\s*noindex\s*-->.+?<!--\s*index\s*-->,,isg;
13 # this will remove all script from indexing content
14 $contents =~ s,<script[^>]*>.+?</script>,,isg;
15 # remap Windows charset to ISO-8859-2
16 $contents =~ tr/šðžèæŠÐŽÈÆ/¹ð¾èæ©Ð®ÈÆ/; # 1250 -> iso8859-2
17 # this will fix badly formatted html in form:
18 # <head><title>some text</title
19 # ></head>
20 # which will confuse indexer (or libxml2?)
21 $contents =~ s/[\n\r]^(>)/$1\n/msg;
22 # remove comments between <html> and <head> texi2html inserts them
23 # there and swish can't find document title then (libxml or swish bug?)
24 while ($contents =~ s/(<html>.*)<!--.*?-->(.*<head>)/$1$2/msi) { };
25
26 # remove empty lines before/after <html>
27 $contents =~ s/^\s+(<html>)/$1/is;
28 $contents =~ s/(<\/html>)\s+$/$1/is;
29 # remove cr
30 $contents =~ s/\r//gs;
31
32 # remove SQL Magazine header and footer
33 $contents =~ s/<!-- begin topnav area -->.+?<!-- end topnav area -->/<\/table>/is;
34 $contents =~ s/<!--Begin Footer-->.+?<\/table>/<\/table>/is;
35
36 # remote TPJ left column
37 if ($contents =~ s,<!-- BEGIN LEFT SIDE BAR CELL -->.+?<!-- END LEFT SIDE BAR CELL -->,,isg) {
38 my $title;
39 # extract title and add to title
40 if ($contents =~ m,<!-- the article goes here -->\s*<h2[^>]*>(.+?)</h2>,si) {
41 $title = $1;
42 } elsif ($contents =~ m,<h1[^>]*>(.+?)</h1>,is) {
43 $title = $1;
44 } elsif ($contents =~ m,<h2[^>]*>(.+?)</h2>,is) {
45 $title = $1;
46 } else {
47 $title = "no detail title";
48 }
49 $contents =~ s,(<title>)([^<]+)(</title>),$1$2: $title$3,gsi if ($title);
50
51 } elsif ($contents =~ m,<!--\s+METADATA(.*?)METADATA\s+-->,is) {
52 # LJ metadata
53 my $metadata = $1;
54 my $title = "LJ";
55
56 $title .= "$1: " if ($metadata =~ m,<issue>(.+?)</issue>,is);
57 $title .= "$1 - " if ($metadata =~ m,<category>(.+?)</category>,is);
58 $title .= $1 if ($metadata =~ m,<title>(.+?)</title>,is);
59 $title =~ s/\s+/ /gsi;
60 $contents =~ s,<title>[^<]+</title>,<title>$title</title>,is;
61 }
62
63 # is second argument collection?
64 my $collection = shift || return $contents;
65
66 # construct new title (from various parts of DocBook if available)
67 my $new_title;
68
69 if ($contents =~ m,<!--SafTocEntry="([^"]+)"-->,is) {
70 $new_title = $1;
71 print STDERR "using title '$new_title' from <!--SafTocEntry-->\n" if ($verbose);
72 } elsif ($contents =~ m,<(h\d)\s+class="docPartTitle"[^>]*>(.+?)</\1>,is) {
73 $new_title = $2;
74 print STDERR "using title '$new_title' from docPartTitle\n" if ($verbose);
75 } elsif ($contents =~ m,<(h\d)\s+class="docChapterTitle"[^>]*>(.+?)</\1>,is) {
76 $new_title = $2;
77 print STDERR "using title '$new_title' from docChapterTitle\n" if ($verbose);
78 } elsif ($contents =~ m,<(h\d)\s+class="docSection1Title"[^>]*>(.+?)</\1>,is) {
79 $new_title = $2;
80 print STDERR "using title '$new_title' from docSection1Title\n" if ($verbose);
81 } elsif ($contents =~ m,<(h\d)\s+class="doc[^"]*Title"[^>]*>(.+?)</\1>,is) {
82 $new_title = $2;
83 print STDERR "using title '$new_title' from doc.+Title\n" if ($verbose);
84 } elsif ($contents =~ m,<(h\d)\s+class="chapter"[^>]*>(.+?)</\1>,is) {
85 $new_title = $2;
86 print STDERR "using title '$new_title' from chapter\n" if ($verbose);
87 } elsif ($contents =~ m,<(h\d)\s+class="sect1"[^>]*>(.+?)</\1>,is) {
88 $new_title = $2;
89 print STDERR "using title '$new_title' from sect1\n" if ($verbose);
90 } else {
91 if ($contents =~ m,<title>([^<]+)</title>,is) {
92 $new_title = $1;
93 print STDERR "using title '$new_title' from <title>\n" if ($verbose);
94 } elsif ($contents =~ m,<h\d[^>]*>([^<]+)</h\d>,is) {
95 $new_title = $1;
96 print STDERR "using title '$new_title' from <h_>\n" if ($verbose);
97 }
98 }
99
100 if ($new_title) {
101 # nuke html in title
102 $new_title =~ s/<br>\s+/: /gs;
103 $new_title =~ s/<\/*[^>]+>//gs;
104
105 # check if new title is same as collection name
106 my ($a,$b) = ($new_title,$collection);
107 $a =~ s/([^a-zA-Z])+/ /gs;
108 $b =~ s/([^a-zA-Z])+/ /gs;
109 if ($a =~ m/$b/i) {
110 $new_title = $collection;
111 print STDERR "new_title and collection are same! [$new_title]\n" if ($verbose);
112 } else {
113 $new_title = "$collection :: $new_title";
114 }
115 } else {
116 # fall-back to collection title
117 $new_title = $collection;
118 }
119
120 $new_title =~ s/\s\s+/ /g;
121
122 if ($contents =~ s,<title>(.*)</title>,<title>$new_title</title>,is) {
123 print STDERR "replace title '$1' with '$new_title'\n" if ($verbose);
124 } elsif ($contents =~ s,(<head>),$1<title>$new_title</title>,is) {
125 print STDERR "adding title '$new_title' after <head>\n" if ($verbose);
126
127 } elsif ($contents =~ s,(<html>),$1<head><title>$new_title</title></head>,is) {
128 print STDERR "adding title '$new_title' after <html>\n" if ($verbose);
129
130 } elsif ($contents =~ s,^,<title>$new_title</title>,) {
131 print STDERR "adding new title '$new_title'\n" if ($verbose);
132 } else {
133 print STDERR "WARNING: filter couldn't add new title '$new_title' anywhere!";
134 }
135
136 return $contents;
137 }
138
139 1;

Properties

Name Value
cvs2svn:cvs-rev 1.5

  ViewVC Help
Powered by ViewVC 1.1.26