1 |
#!/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 |
use WebPAC::Common 0.02; |
11 |
use WebPAC::Lookup; |
12 |
use WebPAC::Input 0.03; |
13 |
use WebPAC::Store 0.03; |
14 |
use WebPAC::Normalize::XML; |
15 |
use WebPAC::Output::TT; |
16 |
use WebPAC::Output::Estraier 0.08; |
17 |
use YAML qw/LoadFile/; |
18 |
use Getopt::Long; |
19 |
use File::Path; |
20 |
|
21 |
=head1 NAME |
22 |
|
23 |
run.pl - start WebPAC indexing |
24 |
|
25 |
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 --one=database_name |
44 |
|
45 |
reindex just single database |
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 $one_db_name; |
62 |
|
63 |
GetOptions( |
64 |
"limit=i" => \$limit, |
65 |
"offset=i" => \$offset, |
66 |
"clean" => \$clean, |
67 |
"one=s" => \$one_db_name, |
68 |
"config" => \$config, |
69 |
"debug" => \$debug, |
70 |
); |
71 |
|
72 |
$config = LoadFile($config); |
73 |
|
74 |
print "config = ",Dumper($config) if ($debug); |
75 |
|
76 |
die "no databases in config file!\n" unless ($config->{databases}); |
77 |
|
78 |
my $total_rows = 0; |
79 |
|
80 |
while (my ($database, $db_config) = each %{ $config->{databases} }) { |
81 |
|
82 |
next if ($one_db_name && $database !~ m/$one_db_name/i); |
83 |
|
84 |
my $log = _new WebPAC::Common()->_get_logger(); |
85 |
|
86 |
# |
87 |
# open Hyper Estraier database |
88 |
# |
89 |
|
90 |
my $est_config = $config->{hyperestraier} || $log->logdie("can't find 'hyperestraier' part in confguration"); |
91 |
$est_config->{database} = $database; |
92 |
$est_config->{clean} = $clean; |
93 |
|
94 |
my $est = new WebPAC::Output::Estraier( %{ $est_config } ); |
95 |
|
96 |
# |
97 |
# now WebPAC::Store |
98 |
# |
99 |
my $abs_path = abs_path($0); |
100 |
$abs_path =~ s#/[^/]*$#/#; |
101 |
|
102 |
my $db_path = $config->{webpac}->{db_path} . '/' . $database; |
103 |
|
104 |
if ($clean) { |
105 |
$log->info("creating new database $database in $db_path"); |
106 |
rmtree( $db_path ) || $log->warn("can't remove $db_path: $!"); |
107 |
} else { |
108 |
$log->info("working on $database in $db_path"); |
109 |
} |
110 |
|
111 |
my $db = new WebPAC::Store( |
112 |
path => $db_path, |
113 |
database => $database, |
114 |
debug => $debug, |
115 |
); |
116 |
|
117 |
|
118 |
# |
119 |
# now, iterate through input formats |
120 |
# |
121 |
|
122 |
my @inputs; |
123 |
if (ref($db_config->{input}) eq 'ARRAY') { |
124 |
@inputs = @{ $db_config->{input} }; |
125 |
} elsif ($db_config->{input}) { |
126 |
push @inputs, $db_config->{input}; |
127 |
} else { |
128 |
$log->info("database $database doesn't have inputs defined"); |
129 |
} |
130 |
|
131 |
my @supported_inputs = keys %{ $config->{webpac}->{inputs} }; |
132 |
|
133 |
foreach my $input (@inputs) { |
134 |
|
135 |
my $type = lc($input->{type}); |
136 |
|
137 |
die "I know only how to handle input types ", join(",", @supported_inputs), " not '$type'!\n" unless (grep(/$type/, @supported_inputs)); |
138 |
|
139 |
my $lookup = new WebPAC::Lookup( |
140 |
lookup_file => $input->{lookup}, |
141 |
); |
142 |
|
143 |
my $input_module = $config->{webpac}->{inputs}->{$type}; |
144 |
|
145 |
$log->info("working on input $input->{path} [$input->{type}] using $input_module"); |
146 |
|
147 |
my $input_db = new WebPAC::Input( |
148 |
module => $input_module, |
149 |
code_page => $config->{webpac}->{webpac_encoding}, |
150 |
limit => $limit || $input->{limit}, |
151 |
offset => $offset, |
152 |
lookup => $lookup, |
153 |
); |
154 |
$log->logdie("can't create input using $input_module") unless ($input); |
155 |
|
156 |
my $maxmfn = $input_db->open( |
157 |
path => $input->{path}, |
158 |
code_page => $input->{encoding}, # database encoding |
159 |
); |
160 |
|
161 |
my $n = new WebPAC::Normalize::XML( |
162 |
# filter => { 'foo' => sub { shift } }, |
163 |
db => $db, |
164 |
lookup_regex => $lookup->regex, |
165 |
lookup => $lookup, |
166 |
prefix => $input->{name}, |
167 |
); |
168 |
|
169 |
my $normalize_path = $input->{normalize}->{path}; |
170 |
|
171 |
if ($normalize_path =~ m/\.xml$/i) { |
172 |
$n->open( |
173 |
tag => $input->{normalize}->{tag}, |
174 |
xml_file => $input->{normalize}->{path}, |
175 |
); |
176 |
} elsif ($normalize_path =~ m/\.(?:yml|yaml)$/i) { |
177 |
$n->open_yaml( |
178 |
path => $normalize_path, |
179 |
tag => $input->{normalize}->{tag}, |
180 |
); |
181 |
} |
182 |
|
183 |
foreach my $pos ( 0 ... $input_db->size ) { |
184 |
|
185 |
my $row = $input_db->fetch || next; |
186 |
|
187 |
my $mfn = $row->{'000'}->[0]; |
188 |
|
189 |
if (! $mfn || $mfn !~ m#^\d+$#) { |
190 |
$log->warn("record $pos doesn't have valid MFN but '$mfn', using $pos"); |
191 |
$mfn = $pos; |
192 |
push @{ $row->{'000'} }, $pos; |
193 |
} |
194 |
|
195 |
my $ds = $n->data_structure($row); |
196 |
|
197 |
$est->add( |
198 |
id => $input->{name} . "/" . $mfn, |
199 |
ds => $ds, |
200 |
type => $config->{hyperestraier}->{type}, |
201 |
); |
202 |
|
203 |
$total_rows++; |
204 |
} |
205 |
|
206 |
}; |
207 |
|
208 |
$log->info("$total_rows records indexed"); |
209 |
|
210 |
# |
211 |
# add Hyper Estraier links to other databases |
212 |
# |
213 |
if (ref($db_config->{links}) eq 'ARRAY') { |
214 |
foreach my $link (@{ $db_config->{links} }) { |
215 |
$log->info("adding link $database -> $link->{to} [$link->{credit}]"); |
216 |
$est->add_link( |
217 |
from => $database, |
218 |
to => $link->{to}, |
219 |
credit => $link->{credit}, |
220 |
); |
221 |
} |
222 |
} |
223 |
|
224 |
} |
225 |
|