9 |
|
|
10 |
use WebPAC::Common 0.02; |
use WebPAC::Common 0.02; |
11 |
use WebPAC::Lookup; |
use WebPAC::Lookup; |
12 |
use WebPAC::Input::ISIS 0.02; |
use WebPAC::Input 0.03; |
13 |
use WebPAC::Store 0.03; |
use WebPAC::Store 0.03; |
14 |
use WebPAC::Normalize::XML; |
use WebPAC::Normalize::XML; |
15 |
use WebPAC::Output::TT; |
use WebPAC::Output::TT; |
|
use WebPAC::Output::Estraier 0.05; |
|
16 |
use YAML qw/LoadFile/; |
use YAML qw/LoadFile/; |
17 |
use LWP::Simple; |
use Getopt::Long; |
18 |
|
use File::Path; |
19 |
|
use Time::HiRes qw/time/; |
20 |
|
|
21 |
my $limit = shift @ARGV; |
=head1 NAME |
22 |
|
|
23 |
my $config = LoadFile('conf/config.yml'); |
run.pl - start WebPAC indexing |
24 |
|
|
25 |
print "config = ",Dumper($config); |
B<this command will probably go away. Don't get used to it!> |
26 |
|
|
27 |
|
Options: |
28 |
|
|
29 |
|
=over 4 |
30 |
|
|
31 |
|
=item --offset 42 |
32 |
|
|
33 |
|
start loading (all) databases at offset 42 |
34 |
|
|
35 |
|
=item --limit 100 |
36 |
|
|
37 |
|
limit loading to 100 records |
38 |
|
|
39 |
|
=item --clean |
40 |
|
|
41 |
|
remove database and Hyper Estraier index before indexing |
42 |
|
|
43 |
|
=item --only=database_name |
44 |
|
|
45 |
|
reindex just single database (legacy name is --one) |
46 |
|
|
47 |
|
=item --config conf/config.yml |
48 |
|
|
49 |
|
path to YAML configuration file |
50 |
|
|
51 |
|
=back |
52 |
|
|
53 |
|
=cut |
54 |
|
|
55 |
|
my $offset; |
56 |
|
my $limit; |
57 |
|
|
58 |
|
my $clean = 0; |
59 |
|
my $config = 'conf/config.yml'; |
60 |
|
my $debug = 0; |
61 |
|
my $only_db_name; |
62 |
|
|
63 |
|
GetOptions( |
64 |
|
"limit=i" => \$limit, |
65 |
|
"offset=i" => \$offset, |
66 |
|
"clean" => \$clean, |
67 |
|
"one=s" => \$only_db_name, |
68 |
|
"only=s" => \$only_db_name, |
69 |
|
"config" => \$config, |
70 |
|
"debug" => \$debug, |
71 |
|
); |
72 |
|
|
73 |
|
$config = LoadFile($config); |
74 |
|
|
75 |
|
print "config = ",Dumper($config) if ($debug); |
76 |
|
|
77 |
die "no databases in config file!\n" unless ($config->{databases}); |
die "no databases in config file!\n" unless ($config->{databases}); |
78 |
|
|
79 |
|
my $log = _new WebPAC::Common()->_get_logger(); |
80 |
|
|
81 |
|
my $use_indexer = $config->{use_indexer} || 'hyperestraier'; |
82 |
|
$log->info("using $use_indexer indexing engine..."); |
83 |
|
|
84 |
my $total_rows = 0; |
my $total_rows = 0; |
85 |
|
my $start_t = time(); |
86 |
|
|
87 |
while (my ($database, $db_config) = each %{ $config->{databases} }) { |
while (my ($database, $db_config) = each %{ $config->{databases} }) { |
88 |
|
|
89 |
my $log = _new WebPAC::Common()->_get_logger(); |
next if ($only_db_name && $database !~ m/$only_db_name/i); |
90 |
|
|
91 |
# |
my $indexer; |
|
# open Hyper Estraier database |
|
|
# |
|
92 |
|
|
93 |
my $est_config = $config->{hyperestraier} || $log->logdie("can't find 'hyperestraier' part in confguration"); |
my $indexer_config = $config->{$use_indexer} || $log->logdie("can't find '$use_indexer' part in confguration"); |
94 |
$est_config->{database} = $database; |
$indexer_config->{database} = $database; |
95 |
|
$indexer_config->{clean} = $clean; |
96 |
|
$indexer_config->{label} = $db_config->{name}; |
97 |
|
|
98 |
|
# important: clean database just once! |
99 |
|
$clean = 0; |
100 |
|
|
101 |
|
if ($use_indexer eq 'hyperestraier') { |
102 |
|
|
103 |
|
# open Hyper Estraier database |
104 |
|
use WebPAC::Output::Estraier '0.10'; |
105 |
|
$indexer = new WebPAC::Output::Estraier( %{ $indexer_config } ); |
106 |
|
|
107 |
|
} elsif ($use_indexer eq 'kinosearch') { |
108 |
|
|
109 |
|
# open KinoSearch |
110 |
|
use WebPAC::Output::KinoSearch; |
111 |
|
$indexer_config->{clean} = 1 unless (-e $indexer_config->{index_path}); |
112 |
|
$indexer = new WebPAC::Output::KinoSearch( %{ $indexer_config } ); |
113 |
|
|
114 |
$log->info("using HyperEstraier URL $est_config->{masterurl}"); |
} else { |
115 |
|
$log->logdie("unknown use_indexer: $use_indexer"); |
116 |
|
} |
117 |
|
|
118 |
my $est = new WebPAC::Output::Estraier( |
$log->logide("can't continue without valid indexer") unless ($indexer); |
|
%{ $est_config }, |
|
|
); |
|
119 |
|
|
120 |
# |
# |
121 |
# now WebPAC::Store |
# now WebPAC::Store |
125 |
|
|
126 |
my $db_path = $config->{webpac}->{db_path} . '/' . $database; |
my $db_path = $config->{webpac}->{db_path} . '/' . $database; |
127 |
|
|
128 |
$log->info("working on $database in $db_path"); |
if ($clean) { |
129 |
|
$log->info("creating new database $database in $db_path"); |
130 |
|
rmtree( $db_path ) || $log->warn("can't remove $db_path: $!"); |
131 |
|
} else { |
132 |
|
$log->debug("working on $database in $db_path"); |
133 |
|
} |
134 |
|
|
135 |
my $db = new WebPAC::Store( |
my $db = new WebPAC::Store( |
136 |
path => $db_path, |
path => $db_path, |
137 |
database => $database, |
database => $database, |
138 |
debug => 1, |
debug => $debug, |
139 |
); |
); |
140 |
|
|
141 |
|
|
152 |
$log->info("database $database doesn't have inputs defined"); |
$log->info("database $database doesn't have inputs defined"); |
153 |
} |
} |
154 |
|
|
155 |
|
my @supported_inputs = keys %{ $config->{webpac}->{inputs} }; |
156 |
|
|
157 |
foreach my $input (@inputs) { |
foreach my $input (@inputs) { |
158 |
|
|
159 |
my $type = lc($input->{type}); |
my $type = lc($input->{type}); |
160 |
|
|
161 |
die "I know only how to handle input type isis, not '$type'!\n" unless ($type eq 'isis'); |
die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs)); |
162 |
|
|
163 |
my $lookup = new WebPAC::Lookup( |
my $lookup = new WebPAC::Lookup( |
164 |
lookup_file => $input->{lookup}, |
lookup_file => $input->{lookup}, |
165 |
); |
); |
166 |
|
|
167 |
$log->info("working on input $input->{path} [$input->{type}]"); |
my $input_module = $config->{webpac}->{inputs}->{$type}; |
168 |
|
|
169 |
|
$log->info("working on input '$input->{path}' [$input->{type}] using $input_module lookup '$input->{lookup}'"); |
170 |
|
|
171 |
my $isis = new WebPAC::Input::ISIS( |
my $input_db = new WebPAC::Input( |
172 |
|
module => $input_module, |
173 |
code_page => $config->{webpac}->{webpac_encoding}, |
code_page => $config->{webpac}->{webpac_encoding}, |
174 |
limit_mfn => $input->{limit}, |
limit => $limit || $input->{limit}, |
175 |
|
offset => $offset, |
176 |
lookup => $lookup, |
lookup => $lookup, |
177 |
|
recode => $input->{recode}, |
178 |
); |
); |
179 |
|
$log->logdie("can't create input using $input_module") unless ($input); |
180 |
|
|
181 |
my $maxmfn = $isis->open( |
my $maxmfn = $input_db->open( |
182 |
path => $input->{path}, |
path => $input->{path}, |
183 |
code_page => $input->{encoding}, # database encoding |
code_page => $input->{encoding}, # database encoding |
184 |
); |
); |
205 |
); |
); |
206 |
} |
} |
207 |
|
|
208 |
for ( 0 ... $isis->size ) { |
foreach my $pos ( 0 ... $input_db->size ) { |
209 |
|
|
210 |
my $row = $isis->fetch || next; |
my $row = $input_db->fetch || next; |
211 |
|
|
212 |
my $mfn = $row->{'000'}->[0] || die "can't find MFN"; |
my $mfn = $row->{'000'}->[0]; |
213 |
|
|
214 |
|
if (! $mfn || $mfn !~ m#^\d+$#) { |
215 |
|
$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos"); |
216 |
|
$mfn = $pos; |
217 |
|
push @{ $row->{'000'} }, $pos; |
218 |
|
} |
219 |
|
|
220 |
my $ds = $n->data_structure($row); |
my $ds = $n->data_structure($row); |
221 |
|
|
222 |
$est->add( |
$indexer->add( |
223 |
id => $input->{name} . "#" . $mfn, |
id => $input->{name} . "/" . $mfn, |
224 |
ds => $ds, |
ds => $ds, |
225 |
type => $config->{hyperestraier}->{type}, |
type => $config->{$use_indexer}->{type}, |
226 |
); |
); |
227 |
|
|
228 |
$total_rows++; |
$total_rows++; |
230 |
|
|
231 |
}; |
}; |
232 |
|
|
233 |
$log->info("$total_rows records indexed"); |
eval { $indexer->finish } if ($indexer->can('finish')); |
234 |
|
|
235 |
|
my $dt = time() - $start_t; |
236 |
|
$log->info("$total_rows records indexed in " . |
237 |
|
sprintf("%.2f sec [%.2f rec/sec]", |
238 |
|
$dt, ($total_rows / $dt) |
239 |
|
) |
240 |
|
); |
241 |
|
|
242 |
# |
# |
243 |
# add Hyper Estraier links to other databases |
# add Hyper Estraier links to other databases |
244 |
# |
# |
245 |
if (ref($db_config->{links}) eq 'ARRAY') { |
if (ref($db_config->{links}) eq 'ARRAY') { |
246 |
foreach my $link (@{ $db_config->{links} }) { |
foreach my $link (@{ $db_config->{links} }) { |
247 |
$log->info("adding link $database -> $link->{to} [$link->{credit}]"); |
if ($use_indexer eq 'hyperestraier') { |
248 |
$est->add_link( |
$log->info("adding link $database -> $link->{to} [$link->{credit}]"); |
249 |
from => $database, |
$indexer->add_link( |
250 |
to => $link->{to}, |
from => $database, |
251 |
credit => $link->{credit}, |
to => $link->{to}, |
252 |
); |
credit => $link->{credit}, |
253 |
|
); |
254 |
|
} else { |
255 |
|
$log->warn("NOT IMPLEMENTED WITH $use_indexer: adding link $database -> $link->{to} [$link->{credit}]"); |
256 |
|
} |
257 |
} |
} |
258 |
} |
} |
259 |
|
|