1 |
sub filter { |
sub filter { |
2 |
my $contents = shift || return; |
my $contents = shift || return; |
3 |
|
|
4 |
|
my $verbose = 0; |
5 |
|
|
6 |
# if you don't want content to be indexed, include it in |
# if you don't want content to be indexed, include it in |
7 |
# <noindex> foobar </noindex> tags or surround it with comments |
# <noindex> foobar </noindex> tags or surround it with comments |
8 |
# <!-- noindex --> foobar <!-- /noindex --> |
# <!-- noindex --> foobar <!-- /noindex --> |
11 |
$contents =~ s,<!--\s*noindex\s*-->.+?<!--\s*/noindex\s*-->,,isg; |
$contents =~ s,<!--\s*noindex\s*-->.+?<!--\s*/noindex\s*-->,,isg; |
12 |
$contents =~ s,<!--\s*noindex\s*-->.+?<!--\s*index\s*-->,,isg; |
$contents =~ s,<!--\s*noindex\s*-->.+?<!--\s*index\s*-->,,isg; |
13 |
# this will remove all script from indexing content |
# this will remove all script from indexing content |
14 |
$contents =~ s,<script>.+?</script>,,isg; |
$contents =~ s,<script[^>]*>.+?</script>,,isg; |
15 |
# remap Windows charset to ISO-8859-2 |
# remap Windows charset to ISO-8859-2 |
16 |
$contents =~ tr/šðžèæŠÐŽÈÆ/¹ð¾èæ©Ð®ÈÆ/; # 1250 -> iso8859-2 |
$contents =~ tr/šðžèæŠÐŽÈÆ/¹ð¾èæ©Ð®ÈÆ/; # 1250 -> iso8859-2 |
17 |
# this will fix badly formatted html in form: |
# this will fix badly formatted html in form: |
22 |
# remove comments between <html> and <head> texi2html inserts them |
# remove comments between <html> and <head> texi2html inserts them |
23 |
# there and swish can't find document title then (libxml or swish bug?) |
# there and swish can't find document title then (libxml or swish bug?) |
24 |
while ($contents =~ s/(<html>.*)<!--.*?-->(.*<head>)/$1$2/msi) { }; |
while ($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 |
|
$contents =~ s/<!--Begin Footer-->.+?<\/table>/<\/table>/is; |
35 |
|
|
36 |
# remote TPJ left column |
# remote TPJ left column |
37 |
if ($contents =~ s,<!-- BEGIN LEFT SIDE BAR CELL -->.+?<!-- END LEFT SIDE BAR CELL -->,,isg) { |
if ($contents =~ s,<!-- BEGIN LEFT SIDE BAR CELL -->.+?<!-- END LEFT SIDE BAR CELL -->,,isg) { |
49 |
$contents =~ s,(<title>)([^<]+)(</title>),$1$2: $title$3,gsi if ($title); |
$contents =~ s,(<title>)([^<]+)(</title>),$1$2: $title$3,gsi if ($title); |
50 |
|
|
51 |
} |
} |
52 |
|
|
53 |
|
# is second argument collection? |
54 |
|
my $collection = shift || return $contents; |
55 |
|
|
56 |
|
# construct new title (from various parts of DocBook if available) |
57 |
|
my $new_title; |
58 |
|
|
59 |
|
if ($contents =~ m,<!--SafTocEntry="([^"]+)"-->,is) { |
60 |
|
$new_title = $1; |
61 |
|
} elsif ($contents =~ m,<(h\d)\sclass="docPartTitle"[^>]*>(.+?)<\1>,is) { |
62 |
|
$new_title = $2; |
63 |
|
} elsif ($contents =~ m,<(h\d)\sclass="docChapterTitle"[^>]*>(.+?)<\1>,is) { |
64 |
|
$new_title = $2; |
65 |
|
} elsif ($contents =~ m,<(h\d)\sclass="docSection1Title"[^>]*>(.+?)<\1>,is) { |
66 |
|
$new_title = $2; |
67 |
|
} elsif ($contents =~ m,<(h\d)\sclass="chapter"[^>]*>(.+?)<\1>,is) { |
68 |
|
$new_title = $2; |
69 |
|
} elsif ($contents =~ m,<(h\d)\sclass="sect1"[^>]*>(.+?)<\1>,is) { |
70 |
|
$new_title = $2; |
71 |
|
} else { |
72 |
|
if ($contents =~ m,<title>([^<]+)</title>,is) { |
73 |
|
$new_title = $1; |
74 |
|
} elsif ($contents =~ m,<h\d[^>]*>([^<]+)</h\d>,is) { |
75 |
|
$new_title = $1; |
76 |
|
} |
77 |
|
} |
78 |
|
|
79 |
|
if ($new_title) { |
80 |
|
# nuke html in title |
81 |
|
$new_title =~ s/<br>\s+/: /gs; |
82 |
|
$new_title =~ s/<\/*[^>]+>//gs; |
83 |
|
|
84 |
|
# check if new title is same as collection name |
85 |
|
my ($a,$b) = ($new_title,$collection); |
86 |
|
$a =~ s/([^a-zA-Z])+/ /gs; |
87 |
|
$b =~ s/([^a-zA-Z])+/ /gs; |
88 |
|
if ($a =~ m/$b/i) { |
89 |
|
$new_title = $collection; |
90 |
|
} else { |
91 |
|
$new_title = "$collection :: $new_title"; |
92 |
|
} |
93 |
|
} else { |
94 |
|
# fall-back to collection title |
95 |
|
$new_title = $collection; |
96 |
|
} |
97 |
|
|
98 |
|
$new_title =~ s/\s\s+/ /g; |
99 |
|
|
100 |
|
if ($contents =~ s,<title>(.*)</title>,<title>$new_title</title>,is) { |
101 |
|
print STDERR "replace title '$1' with '$new_title'\n" if ($verbose); |
102 |
|
} elsif ($contents =~ s,(<head>),$1<title>$new_title</title>,is) { |
103 |
|
print STDERR "adding title '$new_title' after <head>\n" if ($verbose); |
104 |
|
|
105 |
|
} elsif ($contents =~ s,(<html>),$1<head><title>$new_title</title></head>,is) { |
106 |
|
print STDERR "adding title '$new_title' after <html>\n" if ($verbose); |
107 |
|
|
108 |
|
} elsif ($contents =~ s,^,<title>$new_title</title>,) { |
109 |
|
print STDERR "adding new title '$new_title'\n" if ($verbose); |
110 |
|
} else { |
111 |
|
print STDERR "WARNING: filter couldn't add new title '$new_title' anywhere!"; |
112 |
|
} |
113 |
|
|
114 |
return $contents; |
return $contents; |
115 |
} |
} |
116 |
|
|