| 1 |
dpavlin |
46 |
sub filter { |
| 2 |
|
|
my $contents = shift || return; |
| 3 |
dpavlin |
57 |
|
| 4 |
dpavlin |
61 |
my $verbose = 0; |
| 5 |
dpavlin |
57 |
|
| 6 |
dpavlin |
46 |
# 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 |
dpavlin |
51 |
$contents =~ s,<script[^>]*>.+?</script>,,isg; |
| 15 |
dpavlin |
46 |
# remap Windows charset to ISO-8859-2 |
| 16 |
|
|
$contents =~ tr/ðèæÐÈÆ/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 |
dpavlin |
71 |
|
| 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 |
dpavlin |
46 |
|
| 32 |
dpavlin |
71 |
# remove SQL Magazine header and footer |
| 33 |
|
|
$contents =~ s/<!-- begin topnav area -->.+?<!-- end topnav area -->/<\/table>/is; |
| 34 |
|
|
$contents =~ s/<!--Begin Footer-->.+?<\/table>/<\/table>/is; |
| 35 |
|
|
|
| 36 |
dpavlin |
46 |
# remote TPJ left column |
| 37 |
|
|
if ($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 |
dpavlin |
85 |
} 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 |
dpavlin |
46 |
} |
| 62 |
dpavlin |
51 |
|
| 63 |
dpavlin |
56 |
# 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 |
dpavlin |
103 |
sub create_title($$) { |
| 70 |
dpavlin |
100 |
my $contents = shift || return; |
| 71 |
dpavlin |
103 |
my $verbose = shift; |
| 72 |
dpavlin |
100 |
my $new_title = substr($contents, 0, 4096); |
| 73 |
|
|
$new_title =~ s/<[^>]+>//gis; |
| 74 |
|
|
$new_title =~ s/^\s+//s; |
| 75 |
|
|
$new_title =~ s/^(.{50}.*?)[\n\r].+$/$1/s; |
| 76 |
|
|
$new_title =~ s/\s\s+/ /gis; |
| 77 |
|
|
print STDERR "using title '$new_title' from first lines in document\n" if ($verbose); |
| 78 |
|
|
return $new_title; |
| 79 |
|
|
} |
| 80 |
|
|
|
| 81 |
dpavlin |
56 |
if ($contents =~ m,<!--SafTocEntry="([^"]+)"-->,is) { |
| 82 |
|
|
$new_title = $1; |
| 83 |
dpavlin |
74 |
print STDERR "using title '$new_title' from <!--SafTocEntry-->\n" if ($verbose); |
| 84 |
|
|
} elsif ($contents =~ m,<(h\d)\s+class="docPartTitle"[^>]*>(.+?)</\1>,is) { |
| 85 |
dpavlin |
65 |
$new_title = $2; |
| 86 |
dpavlin |
74 |
print STDERR "using title '$new_title' from docPartTitle\n" if ($verbose); |
| 87 |
|
|
} elsif ($contents =~ m,<(h\d)\s+class="docChapterTitle"[^>]*>(.+?)</\1>,is) { |
| 88 |
dpavlin |
65 |
$new_title = $2; |
| 89 |
dpavlin |
74 |
print STDERR "using title '$new_title' from docChapterTitle\n" if ($verbose); |
| 90 |
|
|
} elsif ($contents =~ m,<(h\d)\s+class="docSection1Title"[^>]*>(.+?)</\1>,is) { |
| 91 |
dpavlin |
65 |
$new_title = $2; |
| 92 |
dpavlin |
74 |
print STDERR "using title '$new_title' from docSection1Title\n" if ($verbose); |
| 93 |
|
|
} elsif ($contents =~ m,<(h\d)\s+class="doc[^"]*Title"[^>]*>(.+?)</\1>,is) { |
| 94 |
dpavlin |
65 |
$new_title = $2; |
| 95 |
dpavlin |
74 |
print STDERR "using title '$new_title' from doc.+Title\n" if ($verbose); |
| 96 |
|
|
} elsif ($contents =~ m,<(h\d)\s+class="chapter"[^>]*>(.+?)</\1>,is) { |
| 97 |
dpavlin |
65 |
$new_title = $2; |
| 98 |
dpavlin |
74 |
print STDERR "using title '$new_title' from chapter\n" if ($verbose); |
| 99 |
|
|
} elsif ($contents =~ m,<(h\d)\s+class="sect1"[^>]*>(.+?)</\1>,is) { |
| 100 |
|
|
$new_title = $2; |
| 101 |
|
|
print STDERR "using title '$new_title' from sect1\n" if ($verbose); |
| 102 |
dpavlin |
56 |
} else { |
| 103 |
|
|
if ($contents =~ m,<title>([^<]+)</title>,is) { |
| 104 |
|
|
$new_title = $1; |
| 105 |
dpavlin |
74 |
print STDERR "using title '$new_title' from <title>\n" if ($verbose); |
| 106 |
dpavlin |
56 |
} elsif ($contents =~ m,<h\d[^>]*>([^<]+)</h\d>,is) { |
| 107 |
|
|
$new_title = $1; |
| 108 |
dpavlin |
74 |
print STDERR "using title '$new_title' from <h_>\n" if ($verbose); |
| 109 |
dpavlin |
100 |
} else { |
| 110 |
dpavlin |
103 |
$new_title = create_title($contents, $verbose); |
| 111 |
dpavlin |
56 |
} |
| 112 |
|
|
} |
| 113 |
|
|
|
| 114 |
|
|
if ($new_title) { |
| 115 |
dpavlin |
61 |
# nuke html in title |
| 116 |
|
|
$new_title =~ s/<br>\s+/: /gs; |
| 117 |
|
|
$new_title =~ s/<\/*[^>]+>//gs; |
| 118 |
|
|
|
| 119 |
dpavlin |
56 |
# check if new title is same as collection name |
| 120 |
|
|
my ($a,$b) = ($new_title,$collection); |
| 121 |
|
|
$a =~ s/([^a-zA-Z])+/ /gs; |
| 122 |
|
|
$b =~ s/([^a-zA-Z])+/ /gs; |
| 123 |
|
|
if ($a =~ m/$b/i) { |
| 124 |
|
|
$new_title = $collection; |
| 125 |
dpavlin |
74 |
print STDERR "new_title and collection are same! [$new_title]\n" if ($verbose); |
| 126 |
dpavlin |
56 |
} else { |
| 127 |
|
|
$new_title = "$collection :: $new_title"; |
| 128 |
|
|
} |
| 129 |
|
|
} else { |
| 130 |
|
|
# fall-back to collection title |
| 131 |
|
|
$new_title = $collection; |
| 132 |
dpavlin |
103 |
my $tmp = create_title($ontents, $verbose); |
| 133 |
dpavlin |
100 |
$new_title .= " :: $tmp" if ($tmp); |
| 134 |
dpavlin |
56 |
} |
| 135 |
|
|
|
| 136 |
|
|
$new_title =~ s/\s\s+/ /g; |
| 137 |
|
|
|
| 138 |
dpavlin |
57 |
if ($contents =~ s,<title>(.*)</title>,<title>$new_title</title>,is) { |
| 139 |
|
|
print STDERR "replace title '$1' with '$new_title'\n" if ($verbose); |
| 140 |
dpavlin |
69 |
} elsif ($contents =~ s,(<head>),$1<title>$new_title</title>,is) { |
| 141 |
|
|
print STDERR "adding title '$new_title' after <head>\n" if ($verbose); |
| 142 |
|
|
|
| 143 |
|
|
} elsif ($contents =~ s,(<html>),$1<head><title>$new_title</title></head>,is) { |
| 144 |
|
|
print STDERR "adding title '$new_title' after <html>\n" if ($verbose); |
| 145 |
|
|
|
| 146 |
|
|
} elsif ($contents =~ s,^,<title>$new_title</title>,) { |
| 147 |
dpavlin |
57 |
print STDERR "adding new title '$new_title'\n" if ($verbose); |
| 148 |
dpavlin |
69 |
} else { |
| 149 |
|
|
print STDERR "WARNING: filter couldn't add new title '$new_title' anywhere!"; |
| 150 |
dpavlin |
56 |
} |
| 151 |
|
|
|
| 152 |
dpavlin |
46 |
return $contents; |
| 153 |
|
|
} |
| 154 |
|
|
|
| 155 |
|
|
1; |
Parent Directory
| 
Revision Log