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::Lookup; |
11 |
use WebPAC::Input::ISIS; |
12 |
use WebPAC::Store 0.03; |
13 |
use WebPAC::Normalize::XML; |
14 |
use WebPAC::Output::TT; |
15 |
use WebPAC::Output::Estraier 0.02; |
16 |
use YAML qw/LoadFile/; |
17 |
use LWP::Simple; |
18 |
|
19 |
my $limit = shift @ARGV; |
20 |
|
21 |
my $config = LoadFile('conf/config.yml'); |
22 |
|
23 |
print "config = ",Dumper($config); |
24 |
|
25 |
die "no databases in config file!\n" unless ($config->{databases}); |
26 |
|
27 |
my $total_rows = 0; |
28 |
|
29 |
while (my ($database, $db_config) = each %{ $config->{databases} }) { |
30 |
|
31 |
my $type = lc($db_config->{input}->{type}); |
32 |
|
33 |
die "I know only how to handle input type isis, not '$type'!\n" unless ($type eq 'isis'); |
34 |
|
35 |
my $abs_path = abs_path($0); |
36 |
$abs_path =~ s#/[^/]*$#/#; |
37 |
|
38 |
my $lookup = new WebPAC::Lookup( |
39 |
lookup_file => $db_config->{input}->{lookup}, |
40 |
); |
41 |
|
42 |
my $db_path = $config->{webpac}->{db_path} . '/' . $database; |
43 |
|
44 |
|
45 |
my $log = $lookup->_get_logger; |
46 |
$log->info("working on $database in $db_path"); |
47 |
|
48 |
my $db = new WebPAC::Store( |
49 |
path => $db_path, |
50 |
database => $database, |
51 |
); |
52 |
|
53 |
my $est_config = $config->{hyperestraier} || $log->logdie("can't find 'hyperestraier' part in confguration"); |
54 |
$est_config->{database} = $database; |
55 |
|
56 |
$log->info("using HyperEstraier URL $est_config->{masterurl}"); |
57 |
|
58 |
my $est = new WebPAC::Output::Estraier( |
59 |
%{ $est_config }, |
60 |
); |
61 |
|
62 |
# |
63 |
# now, iterate through input formats |
64 |
# |
65 |
|
66 |
my @inputs; |
67 |
if (ref($db_config->{input}) eq 'ARRAY') { |
68 |
@inputs = @{ $db_config->{input} }; |
69 |
} else { |
70 |
push @inputs, $db_config->{input}; |
71 |
} |
72 |
|
73 |
foreach my $input (@inputs) { |
74 |
$log->info("working on input $input->{path} [$input->{type}]"); |
75 |
|
76 |
my $isis = new WebPAC::Input::ISIS( |
77 |
code_page => $config->{webpac}->{webpac_encoding}, |
78 |
limit_mfn => $input->{limit}, |
79 |
); |
80 |
|
81 |
my $maxmfn = $isis->open( |
82 |
filename => $input->{path}, |
83 |
code_page => $input->{encoding}, # database encoding |
84 |
); |
85 |
|
86 |
my $n = new WebPAC::Normalize::XML( |
87 |
# filter => { 'foo' => sub { shift } }, |
88 |
db => $db, |
89 |
lookup_regex => $lookup->regex, |
90 |
lookup => $lookup, |
91 |
); |
92 |
|
93 |
$n->open( |
94 |
tag => $input->{normalize}->{tag}, |
95 |
xml_file => $input->{normalize}->{path}, |
96 |
); |
97 |
|
98 |
for ( 0 ... $isis->size ) { |
99 |
|
100 |
my $row = $isis->fetch || next; |
101 |
|
102 |
my $mfn = $row->{'000'}->[0] || die "can't find MFN"; |
103 |
|
104 |
my $ds = $n->data_structure($row); |
105 |
|
106 |
$est->add( |
107 |
id => $input->{name} . "#" . $mfn, |
108 |
ds => $ds, |
109 |
type => $config->{hyperestraier}->{type}, |
110 |
); |
111 |
|
112 |
$total_rows++; |
113 |
} |
114 |
|
115 |
}; |
116 |
|
117 |
$log->info("$total_rows records indexed"); |
118 |
} |
119 |
|