1 |
#!/usr/bin/perl -w |
2 |
# -*- Mode: Perl -*- |
3 |
# $Basename$ |
4 |
# $Revision: 1.13 $ |
5 |
# Author : Ulrich Pfeifer |
6 |
# Created On : Mon Dec 31 13:57:11 2001 |
7 |
# Last Modified By: Ulrich Pfeifer |
8 |
# Last Modified On: Fri Jan 4 15:59:20 2002 |
9 |
# Language : CPerl |
10 |
# |
11 |
# (C) Copyright 2001, Ulrich Pfeifer |
12 |
# |
13 |
|
14 |
use 5.007; |
15 |
|
16 |
use strict; |
17 |
use Devel::Peek qw(Dump); |
18 |
|
19 |
use File::Path; |
20 |
use DB_File; |
21 |
use Getopt::Long; |
22 |
use Cwd; |
23 |
|
24 |
BEGIN {require WAIT::Config;} |
25 |
use WAIT::Database; |
26 |
use WAIT::Parse::Ora; |
27 |
use WAIT::Document::Ora; |
28 |
use WAIT::InvertedIndex; |
29 |
use Data::Dumper; |
30 |
|
31 |
|
32 |
$DB_BTREE->{'cachesize'} = 200_000 ; |
33 |
|
34 |
use lib "/usr/local/apache/lib"; |
35 |
use lib "/online/www/sites/ora/catalogsearch/run/lib"; |
36 |
use oreilly_de_catalog::config; |
37 |
use oreilly_de_catalog::wait_filter; |
38 |
|
39 |
my %OPT = ( |
40 |
database => 'oreilly_de_catalog', |
41 |
dir => oreilly_de_catalog::config::WAITDIR, |
42 |
table => 'ora', |
43 |
); |
44 |
|
45 |
my $droot = oreilly_de_catalog::config::CATALOG; |
46 |
|
47 |
GetOptions(\%OPT, |
48 |
'database=s', |
49 |
'dir=s', |
50 |
'table=s', |
51 |
'verbose!', |
52 |
'debug!', |
53 |
) || die "Usage: ...\n"; |
54 |
|
55 |
my @localtime = localtime; |
56 |
$localtime[5] += 1900; |
57 |
$localtime[4]++; |
58 |
my $jobid = sprintf "%04s-%02s-%02s_%02s:%02s_%d", @localtime[5,4,3,2,1], $$; |
59 |
my $db = WAIT::Database->create(name => "$OPT{database}-$jobid", |
60 |
directory => $OPT{dir}) |
61 |
or die "Could not create database $OPT{database}: $@\n"; |
62 |
|
63 |
my $layout = new WAIT::Parse::Ora; |
64 |
|
65 |
# my $stem = ['OR_tr_20020124', 'OR_lc_20020125', 'split2', 'stop', 'Stem']; |
66 |
# my $text = ['OR_tr_20020124', 'split2', 'OR_minus_20020311', 'OR_lc_20020125', 'split2', 'stop']; |
67 |
my $text = ['OR_tr_20020124', 'split', 'OR_minus_20020311', 'OR_lc_20020125']; |
68 |
my $wplus = ['OR_split_20020401', 'OR_lc_20020125', 'OR_mixedonly_20020221']; |
69 |
# my $sound = ['OR_tr_20020124', 'OR_lc_20020125', 'split2', 'Soundex']; |
70 |
my $trigr = ['OR_lc_20020125', 'OR_trigrams_20020125']; |
71 |
|
72 |
# split6 is better than split13 or split10: it allows them to enter |
73 |
# shorter sequences when searching. |
74 |
my $isbn = ['split6', 'OR_isbn_20020127']; |
75 |
|
76 |
my $cwd = cwd; |
77 |
|
78 |
my %D; |
79 |
my $access = tie %D, 'WAIT::Document::Ora', $droot, |
80 |
or die "Couldn't tie to dir $droot: $!\n"; |
81 |
|
82 |
my $tb = $db->create_table(name => $OPT{table}, |
83 |
attr => ['author', 'isbn', 'title', |
84 |
'headline', 'docid'], |
85 |
layout => $layout, |
86 |
access => $access, |
87 |
invindex => |
88 |
[ |
89 |
'aboutauthor' => $text, |
90 |
'aboutauthor' => $wplus, |
91 |
'abouttranslator' => $text, |
92 |
'abouttranslator' => $wplus, |
93 |
'abstract' => $text, |
94 |
'abstract' => $wplus, |
95 |
'author' => $text, |
96 |
'chapter' => $text, |
97 |
'chapter' => $wplus, |
98 |
'colophon' => $text, |
99 |
'colophon' => $wplus, |
100 |
'desc' => $text, |
101 |
'desc' => $wplus, |
102 |
'inx' => $text, |
103 |
'inx' => $wplus, |
104 |
'isbn' => $isbn, |
105 |
'subtitle' => $text, |
106 |
'subtitle' => $wplus, |
107 |
'title' => $text, |
108 |
'title' => $wplus, |
109 |
'title_orig' => $text, |
110 |
'title_orig' => $wplus, |
111 |
'toc' => $text, |
112 |
'toc' => $wplus, |
113 |
'translator' => $text, |
114 |
'translator' => $wplus, |
115 |
] |
116 |
); |
117 |
die "Couldn't create table $OPT{table}: $@\n" unless $tb; |
118 |
|
119 |
my ($did, $value); |
120 |
binmode STDOUT, ":utf8"; |
121 |
my $ALL; |
122 |
my $traceALL = 0; |
123 |
$traceALL = 1 if $OPT{debug}; # expensive |
124 |
my $done = 0; |
125 |
my $todo = keys %D; |
126 |
my $lasttimeround = my $starttime = time; |
127 |
|
128 |
DOC: while (($did, $value) = each %D) { |
129 |
my $superdebug = 0; |
130 |
if ($superdebug && $OPT{debug}) { |
131 |
# next unless $did =~ /perl/; |
132 |
} |
133 |
printf "%15s...(%d)\n", $did, $done if $OPT{verbose}; |
134 |
my $record = $layout->split($value); |
135 |
my $headline = $record->{title}; |
136 |
$headline =~ s/\s+/ /sg; |
137 |
# printf "%15s %s\n", $record->{isbn}, substr($headline,0,60); |
138 |
printf "%15s %s...\n", $did, substr($headline,0,60) if $OPT{verbose}; |
139 |
if ($superdebug && $OPT{debug}) { |
140 |
# $record = { chapter => $record->{chapter}}; |
141 |
} |
142 |
$done++; |
143 |
next if oreilly_de_catalog::config::DONT_INDEX_JUST_PARSE() ; |
144 |
$tb->insert('docid' => $did, |
145 |
headline => $headline, |
146 |
%{$record}); |
147 |
my $spenttime = time - $starttime; |
148 |
my $averagetime = $spenttime/$done; |
149 |
my $left = $todo-$done; |
150 |
printf("%2d secs, %3d done, %3d left, %4d s done, %5.1f s avg, %4d s left\n", |
151 |
time - $lasttimeround, |
152 |
$done, |
153 |
$left, |
154 |
$spenttime, |
155 |
$averagetime, |
156 |
$left*$averagetime, |
157 |
) if $OPT{verbose}; |
158 |
$lasttimeround = time; |
159 |
if ($traceALL) { # costs a lot when reaching the 100th file or so |
160 |
$ALL->{$did} = $record; |
161 |
open F, ">:utf8", "$OPT{dir}/$OPT{database}-$jobid/debug.dump" or die; |
162 |
print F Data::Dumper::Dumper($ALL); |
163 |
close F or die "Couldn't close debug.dump: $!";; |
164 |
} |
165 |
} |
166 |
undef $ALL; |
167 |
$tb->set(top=>1); |
168 |
|
169 |
my $tritb = $db->create_table( |
170 |
name => "$OPT{table}_fallback", |
171 |
attr => [qw(docid headline)], # name |
172 |
# "headline" |
173 |
# only for |
174 |
# sman |
175 |
invindex => [ headline => $trigr ], |
176 |
); |
177 |
my %dict; |
178 |
for my $f ($tb->fields) { |
179 |
my(@idx) = @{$tb->table->{inverted}{$f} || []}; |
180 |
for my $idx (@idx) { |
181 |
my $name = $idx->name; |
182 |
next if $name =~ /(_|\b)(mixedonly|Stem|Soundex)(\b|_)/; |
183 |
# irrelevant for alternatives |
184 |
my @keys = $idx->keys; |
185 |
@dict{@keys} = (); |
186 |
} |
187 |
} |
188 |
my @dictkeys = grep s/^p//, keys %dict; |
189 |
my $maxdebug = 5; |
190 |
for my $headline (@dictkeys) { |
191 |
if ($maxdebug && $headline =~ /[^\040-\177]/) { |
192 |
Dump $headline; |
193 |
$maxdebug--; |
194 |
} |
195 |
# printf "%s\n", substr($headline,0,60); |
196 |
$tritb->insert(docid => $headline, headline => $headline); |
197 |
} |
198 |
$tritb->set(top=>1); |
199 |
$tritb->close or die "Couldn't close table: $!"; |
200 |
$tb->close() or die "Couldn't close table: $!"; |
201 |
$db->close() or die "Couldn't close database: $!"; |
202 |
|
203 |
# Atomically relinking symlink: now we have a new database with a very |
204 |
# long name "$OPT{database}-$jobid" (e.g. |
205 |
# oreilly_de_catalog-2002-01-28_16:12_16467) and we want that database |
206 |
# to be accessible with the oreilly_de_catalog name. |
207 |
|
208 |
use File::Spec; |
209 |
chdir $OPT{dir} or die; |
210 |
my $dir = "$OPT{database}-$jobid"; |
211 |
my $slwant = File::Spec->catdir($OPT{dir}, $OPT{database}); |
212 |
my $sltmp = File::Spec->catdir($OPT{dir}, "$OPT{database}-$$"); |
213 |
unlink $sltmp; # may fail |
214 |
symlink $dir, $sltmp or die "Couldn't symlink $dir, $sltmp: $!"; |
215 |
rename $sltmp, $slwant or die "Couldn't rename $sltmp, $slwant: $!"; |
216 |
print "$slwant now points to $dir\n" if $OPT{verbose}; |
217 |
system("chmod 777 $slwant/*/read")==0 or die; |
218 |
|
219 |
opendir DIR, "." or die "Could not opendir .: $!"; |
220 |
for my $dirent (readdir DIR) { |
221 |
next if $dirent =~ /^\./; |
222 |
next unless $dirent =~ /^$OPT{database}(.*)/; |
223 |
my $ext = $1 or next; |
224 |
next unless -M $dirent > 4; |
225 |
warn "removing old index $dirent"; |
226 |
File::Path::rmtree($dirent); |
227 |
} |
228 |
closedir DIR; |
229 |
|
230 |
$WAIT::Config = $WAIT::Config; # make perl -w happy |
231 |
|
232 |
|
233 |
__END__ |
234 |
## ################################################################### |
235 |
## pod |
236 |
## ################################################################### |
237 |
|
238 |
=head1 NAME |
239 |
|
240 |
index_ora - generate an WAIT index for O'Reilly catalog |
241 |
|
242 |
=head1 SYNOPSIS |
243 |
|
244 |
B<index_ora> |
245 |
[B<-database> I<dbname>] |
246 |
[B<-dir> I<directory>] |
247 |
[B<-table> I<table name>] |
248 |
I<directory> |
249 |
|
250 |
=head1 DESCRIPTION |
251 |
|
252 |
=head1 OPTIONS |
253 |
|
254 |
=over 5 |
255 |
|
256 |
=item B<-database> I<dbname> |
257 |
|
258 |
Specify database name. Default is F<DB>. |
259 |
|
260 |
=item B<-dir> I<directory> |
261 |
|
262 |
Alternate directory where databases are located. Default is the |
263 |
directory specified during configuration of WAIT. |
264 |
|
265 |
=item B<-table> I<table name> |
266 |
|
267 |
Specify an alternate table name. Default is C<ora>. |
268 |
|
269 |
=head1 AUTHOR |
270 |
|
271 |
Ulrich Pfeifer E<lt>F<pfeifer@wait.de>E<gt> |
272 |
|