| 1 |
#!/usr/bin/perl -w |
| 2 |
|
| 3 |
use strict; |
| 4 |
|
| 5 |
use Getopt::Long; |
| 6 |
use Digest::MD5 qw(md5_hex); |
| 7 |
use Data::Dumper; |
| 8 |
|
| 9 |
require WAIT::Database; |
| 10 |
require WAIT::Config; |
| 11 |
require WAIT::Parse::HTML; |
| 12 |
require WAIT::Document::Find; |
| 13 |
|
| 14 |
use utf8; |
| 15 |
|
| 16 |
use lib '/data/wait/lib'; |
| 17 |
|
| 18 |
my %OPT = (database => 'DB', |
| 19 |
dir => $WAIT::Config->{WAIT_home} || '/tmp', |
| 20 |
table => 'html', |
| 21 |
clean => 0, |
| 22 |
remove => 0, |
| 23 |
force => 0, |
| 24 |
); |
| 25 |
|
| 26 |
GetOptions(\%OPT, |
| 27 |
'database=s', |
| 28 |
'dir=s', |
| 29 |
'table=s', |
| 30 |
'clean!', |
| 31 |
'remove!', |
| 32 |
'force!', |
| 33 |
); |
| 34 |
|
| 35 |
my $path = shift @ARGV || |
| 36 |
die "Usage: $0 [-database=$OPT{database}] [-dir=$OPT{dir}] [-table=$OPT{table}]\n\t[-clean] [-remove] [-force] directory_with_htmls\n"; |
| 37 |
|
| 38 |
if ($OPT{clean}) { |
| 39 |
my $db = WAIT::Database->open( |
| 40 |
name => $OPT{database}, |
| 41 |
'directory' => $OPT{dir}, |
| 42 |
) |
| 43 |
or die "Could not open database '$OPT{dir}/$OPT{database}': $@"; |
| 44 |
$db->drop_table(name => $OPT{table}) or |
| 45 |
die "Could not drop table '$OPT{tabel}': $@"; |
| 46 |
|
| 47 |
$db->close; |
| 48 |
} |
| 49 |
|
| 50 |
my $db = WAIT::Database->open( |
| 51 |
name => $OPT{database}, |
| 52 |
'directory' => $OPT{dir}, |
| 53 |
) |
| 54 |
|| WAIT::Database->create( |
| 55 |
name => $OPT{database}, |
| 56 |
'directory' => $OPT{dir}, |
| 57 |
) |
| 58 |
or die "Could not open/create database '$OPT{dir}/$OPT{database}': $@"; |
| 59 |
|
| 60 |
my $layout= new WAIT::Parse::HTML; |
| 61 |
my $stem = [{ |
| 62 |
'prefix' => ['isotr', 'isolc'], |
| 63 |
'intervall' => ['isotr', 'isolc'], |
| 64 |
},'decode_entities', 'isotr', 'isolc', 'split2', 'stop', 'Stem']; |
| 65 |
my $text = [{ |
| 66 |
'prefix' => ['isotr', 'isolc'], |
| 67 |
'intervall' => ['isotr', 'isolc'], |
| 68 |
}, |
| 69 |
'decode_entities', 'isotr', 'isolc', 'split2', 'stop']; |
| 70 |
my $sound = ['decode_entities', 'isotr', 'isolc', 'split2', 'Soundex']; |
| 71 |
|
| 72 |
my %D; |
| 73 |
|
| 74 |
my $access = tie (%D, 'WAIT::Document::Find', sub { $_[0] =~ /\.htm/; }, |
| 75 |
$path); |
| 76 |
die $@ unless defined $access; |
| 77 |
|
| 78 |
|
| 79 |
my $tb = $db->table(name => $OPT{table}) || |
| 80 |
$db->create_table |
| 81 |
(name => $OPT{table}, |
| 82 |
attr => ['docid', 'headline', 'size', 'md5'], |
| 83 |
keyset => [['docid', 'md5']], |
| 84 |
layout => $layout, |
| 85 |
access => $access, |
| 86 |
invindex => |
| 87 |
[ |
| 88 |
'text' => $stem, |
| 89 |
'title' => $stem, |
| 90 |
'title' => $text, |
| 91 |
] |
| 92 |
); |
| 93 |
die unless $tb; |
| 94 |
|
| 95 |
while (my ($path, $content) = each %D) { |
| 96 |
&index($path, $content); |
| 97 |
} |
| 98 |
$db->close(); |
| 99 |
exit; |
| 100 |
|
| 101 |
my $NO; |
| 102 |
sub index { |
| 103 |
my ($did, $value) = @_; |
| 104 |
if ($tb->have('docid' => $did)) { |
| 105 |
if (!$OPT{remove} && !$OPT{force}) { |
| 106 |
print "duplicate\n"; |
| 107 |
return; |
| 108 |
} |
| 109 |
} elsif ($OPT{remove}) { |
| 110 |
print "missing\n"; |
| 111 |
return; |
| 112 |
} |
| 113 |
|
| 114 |
if (-s $did < 100) { |
| 115 |
print "too small\n"; |
| 116 |
return; |
| 117 |
} |
| 118 |
|
| 119 |
unless (defined $value) { |
| 120 |
print "unavailable\n"; |
| 121 |
return; |
| 122 |
} |
| 123 |
printf STDERR "ok [%d]\n", ++$NO; |
| 124 |
|
| 125 |
my $record = $layout->split($value); |
| 126 |
$record->{size} = length($value); |
| 127 |
my $headline = $record->{title} || $did; |
| 128 |
$headline =~ s/\s+/ /g; $headline =~ s/^\s+//; |
| 129 |
printf "%s\n", substr($headline,0,80); |
| 130 |
my $t = \$record->{text}; |
| 131 |
if ($$t) { |
| 132 |
my $md5; |
| 133 |
if (utf8::is_utf8($$t)) { |
| 134 |
$md5 = Dumper($$t) |
| 135 |
} else { |
| 136 |
$md5 = md5_hex($$t); |
| 137 |
} |
| 138 |
if ($tb->have('md5' => $md5)) { |
| 139 |
print "duplicate md5\n"; |
| 140 |
return; |
| 141 |
} |
| 142 |
$record->{md5} = $md5; |
| 143 |
print "$md5\n"; |
| 144 |
} else { |
| 145 |
print "no page content! skipping...\n"; |
| 146 |
return; |
| 147 |
} |
| 148 |
|
| 149 |
if ($OPT{remove}) { |
| 150 |
$tb->delete('docid' => $did, headline => $headline, %{$record}); |
| 151 |
} else { |
| 152 |
$tb->insert('docid' => $did, headline => $headline, %{$record}); |
| 153 |
} |
| 154 |
} |
| 155 |
|
| 156 |
$WAIT::Config = $WAIT::Config; |
| 157 |
|
| 158 |
__END__ |
| 159 |
## ################################################################### |
| 160 |
## pod |
| 161 |
## ################################################################### |
| 162 |
|
| 163 |
=head1 NAME |
| 164 |
|
| 165 |
index_html - generate a html database for sman |
| 166 |
|
| 167 |
=head1 SYNOPSIS |
| 168 |
|
| 169 |
B<index_html> |
| 170 |
[B<-database> I<database name>] |
| 171 |
[B<-dir> I<database directory>] |
| 172 |
[B<-table> I<name>] |
| 173 |
[B<-remove>] |
| 174 |
[I<htmldir> ...] |
| 175 |
|
| 176 |
=head1 DESCRIPTION |
| 177 |
|
| 178 |
B<Index_html> generates/updates databases for B<sman>(1). If |
| 179 |
I<htmldir>s are specified, these are used. Otherwise the script dies. |
| 180 |
|
| 181 |
=head2 OPTIONS |
| 182 |
|
| 183 |
=over 10 |
| 184 |
|
| 185 |
=item B<-database> I<database name> |
| 186 |
|
| 187 |
Change the default database name to I<database name>. |
| 188 |
|
| 189 |
=item B<-dir> I<database directory> |
| 190 |
|
| 191 |
Change the default database directory to I<database directory>. |
| 192 |
|
| 193 |
=item B<-table> I<name> |
| 194 |
|
| 195 |
Use I<name> instead of C<man> as table name. |
| 196 |
|
| 197 |
=item B<-clean> |
| 198 |
|
| 199 |
Clean B<database> before indexing. |
| 200 |
|
| 201 |
=item B<-remove> |
| 202 |
|
| 203 |
Remove the selected directories from the database instead of |
| 204 |
adding/updating. This works only for the manuals which are unchanged |
| 205 |
since the indexing. |
| 206 |
|
| 207 |
=back |
| 208 |
|
| 209 |
=head1 SEE ALSO |
| 210 |
|
| 211 |
L<sman>. |
| 212 |
|
| 213 |
=head1 AUTHOR |
| 214 |
|
| 215 |
Ulrich Pfeifer E<lt>F<pfeifer@ls6.informatik.uni-dortmund.de>E<gt> |
Parent Directory
| 
Revision Log