/[webpac2]/trunk/run.pl
This is repository of my old source code which isn't updated any more. Go to git.rot13.org for current projects!
ViewVC logotype

Annotation of /trunk/run.pl

Parent Directory Parent Directory | Revision Log Revision Log


Revision 509 - (hide annotations)
Mon May 15 17:23:38 2006 UTC (17 years, 11 months ago) by dpavlin
File MIME type: text/plain
File size: 7337 byte(s)
disable indexing if --stats option is used, added delimiter line to begin of
output to log

1 dpavlin 74 #!/usr/bin/perl -w
2    
3     use strict;
4    
5     use Cwd qw/abs_path/;
6     use File::Temp qw/tempdir/;
7     use Data::Dumper;
8     use lib './lib';
9    
10 dpavlin 255 use WebPAC::Common 0.02;
11 dpavlin 74 use WebPAC::Lookup;
12 dpavlin 287 use WebPAC::Input 0.03;
13 dpavlin 209 use WebPAC::Store 0.03;
14 dpavlin 74 use WebPAC::Normalize::XML;
15 dpavlin 492 use WebPAC::Normalize::Set;
16 dpavlin 74 use WebPAC::Output::TT;
17 dpavlin 141 use YAML qw/LoadFile/;
18 dpavlin 301 use Getopt::Long;
19     use File::Path;
20 dpavlin 389 use Time::HiRes qw/time/;
21 dpavlin 492 use File::Slurp;
22 dpavlin 74
23 dpavlin 301 =head1 NAME
24 dpavlin 76
25 dpavlin 301 run.pl - start WebPAC indexing
26 dpavlin 141
27 dpavlin 301 B<this command will probably go away. Don't get used to it!>
28 dpavlin 141
29 dpavlin 301 Options:
30    
31     =over 4
32    
33     =item --offset 42
34    
35     start loading (all) databases at offset 42
36    
37     =item --limit 100
38    
39     limit loading to 100 records
40    
41     =item --clean
42    
43     remove database and Hyper Estraier index before indexing
44    
45 dpavlin 423 =item --only=database_name
46 dpavlin 335
47 dpavlin 423 reindex just single database (legacy name is --one)
48 dpavlin 335
49 dpavlin 301 =item --config conf/config.yml
50    
51     path to YAML configuration file
52    
53 dpavlin 492 =item --force-set
54    
55 dpavlin 508 force conversion C<< normalize->path >> in C<config.yml> from
56 dpavlin 492 C<.xml> to C<.pl>
57    
58 dpavlin 507 =item --stats
59    
60 dpavlin 509 disable indexing and dump statistics about field and subfield
61     usage for each input
62 dpavlin 507
63 dpavlin 301 =back
64    
65     =cut
66    
67     my $offset;
68     my $limit;
69    
70     my $clean = 0;
71     my $config = 'conf/config.yml';
72     my $debug = 0;
73 dpavlin 423 my $only_db_name;
74 dpavlin 492 my $force_set = 0;
75 dpavlin 507 my $stats = 0;
76 dpavlin 301
77     GetOptions(
78     "limit=i" => \$limit,
79     "offset=i" => \$offset,
80     "clean" => \$clean,
81 dpavlin 423 "one=s" => \$only_db_name,
82     "only=s" => \$only_db_name,
83 dpavlin 301 "config" => \$config,
84     "debug" => \$debug,
85 dpavlin 492 "force-set" => \$force_set,
86 dpavlin 507 "stats" => \$stats,
87 dpavlin 301 );
88    
89     $config = LoadFile($config);
90    
91     print "config = ",Dumper($config) if ($debug);
92    
93 dpavlin 210 die "no databases in config file!\n" unless ($config->{databases});
94 dpavlin 431
95     my $log = _new WebPAC::Common()->_get_logger();
96 dpavlin 509 $log->info( "-" x 79 );
97 dpavlin 431
98 dpavlin 430 my $use_indexer = $config->{use_indexer} || 'hyperestraier';
99 dpavlin 509 if ($stats) {
100     $log->debug("option --stats disables update of indexing engine...");
101     $use_indexer = undef;
102     } else {
103     $log->info("using $use_indexer indexing engine...");
104     }
105 dpavlin 141
106 dpavlin 213 my $total_rows = 0;
107 dpavlin 389 my $start_t = time();
108 dpavlin 213
109 dpavlin 210 while (my ($database, $db_config) = each %{ $config->{databases} }) {
110 dpavlin 141
111 dpavlin 423 next if ($only_db_name && $database !~ m/$only_db_name/i);
112 dpavlin 335
113 dpavlin 430 my $indexer;
114 dpavlin 255
115 dpavlin 509 if ($use_indexer) {
116     my $indexer_config = $config->{$use_indexer} || $log->logdie("can't find '$use_indexer' part in confguration");
117     $indexer_config->{database} = $database;
118     $indexer_config->{clean} = $clean;
119     $indexer_config->{label} = $db_config->{name};
120 dpavlin 431
121 dpavlin 509 if ($use_indexer eq 'hyperestraier') {
122 dpavlin 255
123 dpavlin 509 # open Hyper Estraier database
124     use WebPAC::Output::Estraier '0.10';
125     $indexer = new WebPAC::Output::Estraier( %{ $indexer_config } );
126    
127     } elsif ($use_indexer eq 'kinosearch') {
128 dpavlin 430
129 dpavlin 509 # open KinoSearch
130     use WebPAC::Output::KinoSearch;
131     $indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path});
132     $indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } );
133 dpavlin 431
134 dpavlin 509 } else {
135     $log->logdie("unknown use_indexer: $use_indexer");
136     }
137    
138     $log->logide("can't continue without valid indexer") unless ($indexer);
139 dpavlin 430 }
140    
141    
142 dpavlin 255 #
143     # now WebPAC::Store
144     #
145 dpavlin 210 my $abs_path = abs_path($0);
146     $abs_path =~ s#/[^/]*$#/#;
147 dpavlin 141
148 dpavlin 210 my $db_path = $config->{webpac}->{db_path} . '/' . $database;
149 dpavlin 74
150 dpavlin 301 if ($clean) {
151     $log->info("creating new database $database in $db_path");
152     rmtree( $db_path ) || $log->warn("can't remove $db_path: $!");
153     } else {
154 dpavlin 391 $log->debug("working on $database in $db_path");
155 dpavlin 301 }
156 dpavlin 255
157 dpavlin 210 my $db = new WebPAC::Store(
158     path => $db_path,
159 dpavlin 217 database => $database,
160 dpavlin 301 debug => $debug,
161 dpavlin 210 );
162 dpavlin 74
163 dpavlin 233
164 dpavlin 213 #
165     # now, iterate through input formats
166     #
167 dpavlin 74
168 dpavlin 213 my @inputs;
169     if (ref($db_config->{input}) eq 'ARRAY') {
170     @inputs = @{ $db_config->{input} };
171 dpavlin 255 } elsif ($db_config->{input}) {
172     push @inputs, $db_config->{input};
173 dpavlin 213 } else {
174 dpavlin 255 $log->info("database $database doesn't have inputs defined");
175 dpavlin 213 }
176 dpavlin 74
177 dpavlin 286 my @supported_inputs = keys %{ $config->{webpac}->{inputs} };
178    
179 dpavlin 213 foreach my $input (@inputs) {
180 dpavlin 233
181     my $type = lc($input->{type});
182    
183 dpavlin 286 die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs));
184 dpavlin 233
185     my $lookup = new WebPAC::Lookup(
186     lookup_file => $input->{lookup},
187     );
188    
189 dpavlin 286 my $input_module = $config->{webpac}->{inputs}->{$type};
190 dpavlin 74
191 dpavlin 391 $log->info("working on input '$input->{path}' [$input->{type}] using $input_module lookup '$input->{lookup}'");
192 dpavlin 286
193 dpavlin 287 my $input_db = new WebPAC::Input(
194     module => $input_module,
195 dpavlin 213 code_page => $config->{webpac}->{webpac_encoding},
196 dpavlin 301 limit => $limit || $input->{limit},
197     offset => $offset,
198 dpavlin 251 lookup => $lookup,
199 dpavlin 416 recode => $input->{recode},
200 dpavlin 507 stats => $stats,
201 dpavlin 287 );
202 dpavlin 286 $log->logdie("can't create input using $input_module") unless ($input);
203 dpavlin 113
204 dpavlin 287 my $maxmfn = $input_db->open(
205 dpavlin 285 path => $input->{path},
206 dpavlin 213 code_page => $input->{encoding}, # database encoding
207     );
208 dpavlin 113
209 dpavlin 213 my $n = new WebPAC::Normalize::XML(
210     # filter => { 'foo' => sub { shift } },
211     db => $db,
212     lookup_regex => $lookup->regex,
213     lookup => $lookup,
214 dpavlin 221 prefix => $input->{name},
215 dpavlin 213 );
216 dpavlin 113
217 dpavlin 492 my $rules;
218 dpavlin 269 my $normalize_path = $input->{normalize}->{path};
219 dpavlin 210
220 dpavlin 492 if ($force_set) {
221     my $new_norm_path = $normalize_path;
222     $new_norm_path =~ s/\.xml$/.pl/;
223     if (-e $new_norm_path) {
224 dpavlin 493 $log->debug("--force-set replaced $normalize_path with $new_norm_path");
225 dpavlin 492 $normalize_path = $new_norm_path;
226     } else {
227 dpavlin 493 $log->debug("--force-set failed on $new_norm_path, fallback to $normalize_path");
228 dpavlin 492 }
229     }
230    
231 dpavlin 269 if ($normalize_path =~ m/\.xml$/i) {
232     $n->open(
233     tag => $input->{normalize}->{tag},
234 dpavlin 492 xml_file => $normalize_path,
235 dpavlin 269 );
236     } elsif ($normalize_path =~ m/\.(?:yml|yaml)$/i) {
237     $n->open_yaml(
238     path => $normalize_path,
239     tag => $input->{normalize}->{tag},
240     );
241 dpavlin 492 } elsif ($normalize_path =~ m/\.(?:pl)$/i) {
242     $n = undef;
243     $log->info("using WebPAC::Normalize::Set to process $normalize_path");
244     $rules = read_file( $normalize_path ) or die "can't open $normalize_path: $!";
245 dpavlin 269 }
246    
247 dpavlin 290 foreach my $pos ( 0 ... $input_db->size ) {
248 dpavlin 210
249 dpavlin 287 my $row = $input_db->fetch || next;
250 dpavlin 213
251 dpavlin 291 my $mfn = $row->{'000'}->[0];
252 dpavlin 213
253 dpavlin 291 if (! $mfn || $mfn !~ m#^\d+$#) {
254 dpavlin 290 $log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos");
255     $mfn = $pos;
256 dpavlin 291 push @{ $row->{'000'} }, $pos;
257 dpavlin 290 }
258    
259 dpavlin 508
260     my $ds;
261     if ($n) {
262     $ds = $n->data_structure($row);
263     } else {
264     $ds = WebPAC::Normalize::Set::data_structure(
265 dpavlin 492 row => $row,
266     rules => $rules,
267     lookup => $lookup->lookup_hash,
268     );
269 dpavlin 213
270 dpavlin 508 $db->save_ds(
271     id => $mfn,
272     ds => $ds,
273     prefix => $input->{name},
274 dpavlin 509 ) if ($ds && !$stats);
275 dpavlin 508 }
276    
277 dpavlin 430 $indexer->add(
278 dpavlin 291 id => $input->{name} . "/" . $mfn,
279 dpavlin 213 ds => $ds,
280 dpavlin 430 type => $config->{$use_indexer}->{type},
281 dpavlin 509 ) if ($indexer);
282 dpavlin 213
283     $total_rows++;
284     }
285    
286 dpavlin 507 $log->info("statistics of fields usage:\n", $input_db->stats) if ($stats);
287    
288 dpavlin 210 };
289    
290 dpavlin 509 eval { $indexer->finish } if ($indexer && $indexer->can('finish'));
291 dpavlin 434
292 dpavlin 389 my $dt = time() - $start_t;
293     $log->info("$total_rows records indexed in " .
294     sprintf("%.2f sec [%.2f rec/sec]",
295     $dt, ($total_rows / $dt)
296     )
297     );
298 dpavlin 255
299 dpavlin 431 #
300     # add Hyper Estraier links to other databases
301     #
302     if (ref($db_config->{links}) eq 'ARRAY') {
303     foreach my $link (@{ $db_config->{links} }) {
304     if ($use_indexer eq 'hyperestraier') {
305 dpavlin 430 $log->info("adding link $database -> $link->{to} [$link->{credit}]");
306     $indexer->add_link(
307     from => $database,
308     to => $link->{to},
309     credit => $link->{credit},
310     );
311 dpavlin 431 } else {
312     $log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]");
313 dpavlin 430 }
314 dpavlin 255 }
315     }
316    
317 dpavlin 210 }
318    

Properties

Name Value
svn:executable *

  ViewVC Help
Powered by ViewVC 1.1.26