1 |
dpavlin |
1280 |
#!/usr/bin/perl |
2 |
|
|
|
3 |
|
|
use warnings; |
4 |
|
|
use strict; |
5 |
|
|
|
6 |
dpavlin |
1291 |
our $q = 'AD=Croatia'; |
7 |
dpavlin |
1289 |
my $range_size = 500; |
8 |
dpavlin |
1280 |
|
9 |
dpavlin |
1282 |
my $dump = @ARGV ? 1 : 0; |
10 |
|
|
|
11 |
dpavlin |
1289 |
$q = 'TS=psychology AND AD=Croatia'; |
12 |
dpavlin |
1282 |
|
13 |
dpavlin |
1280 |
use WWW::Mechanize; |
14 |
dpavlin |
1291 |
use Data::Dump qw(dump); |
15 |
|
|
use File::Path; |
16 |
dpavlin |
1280 |
|
17 |
dpavlin |
1289 |
our $mech = WWW::Mechanize->new( |
18 |
dpavlin |
1280 |
autocheck => 1, |
19 |
|
|
cookie_jar => undef, |
20 |
|
|
); |
21 |
|
|
|
22 |
dpavlin |
1281 |
our $step = 0; |
23 |
dpavlin |
1280 |
|
24 |
dpavlin |
1291 |
my $dir = '/tmp/isi/'; |
25 |
|
|
rmtree $dir if -e $dir; |
26 |
|
|
mkdir $dir; |
27 |
|
|
|
28 |
dpavlin |
1280 |
sub save_mech { |
29 |
dpavlin |
1281 |
my ( $mech, $path ) = @_; |
30 |
|
|
$step++; |
31 |
dpavlin |
1291 |
my $base_path = sprintf('%s/%04d', $dir,$step); |
32 |
dpavlin |
1290 |
$path ||= $base_path . ( $mech->{ct} =~ m{html}i ? '.html' : '.txt' ); |
33 |
dpavlin |
1281 |
$mech->save_content( $path ); |
34 |
dpavlin |
1290 |
warn "# [$step] $path ", -s $path, " ", $mech->ct, "\n"; |
35 |
|
|
open(my $dump, '>', "$base_path.dump.txt"); |
36 |
|
|
$mech->dump_all($dump); |
37 |
dpavlin |
1280 |
} |
38 |
|
|
|
39 |
|
|
warn "# get session"; |
40 |
|
|
$mech->get( 'http://isiknowledge.com/?DestApp=WOS' ); |
41 |
|
|
save_mech $mech; |
42 |
|
|
|
43 |
dpavlin |
1290 |
sub search { |
44 |
|
|
warn "# advanced serach"; |
45 |
|
|
$mech->follow_link( url_regex => qr/AdvancedSearch/ ); |
46 |
|
|
save_mech $mech; |
47 |
dpavlin |
1280 |
|
48 |
dpavlin |
1290 |
warn "# cookie_jar ", dump $mech->cookie_jar; |
49 |
dpavlin |
1280 |
|
50 |
dpavlin |
1290 |
$mech->submit_form( |
51 |
|
|
fields => { |
52 |
|
|
'value(input1)' => $q, |
53 |
|
|
} |
54 |
|
|
); |
55 |
|
|
save_mech $mech; |
56 |
dpavlin |
1280 |
|
57 |
dpavlin |
1290 |
warn "# summary"; |
58 |
|
|
$mech->follow_link( url_regex => qr/summary/ ); |
59 |
|
|
save_mech $mech; |
60 |
|
|
} |
61 |
|
|
|
62 |
dpavlin |
1289 |
sub get_results { |
63 |
|
|
my $from = 1; |
64 |
dpavlin |
1280 |
|
65 |
dpavlin |
1289 |
while ( 1 ) { |
66 |
dpavlin |
1280 |
|
67 |
dpavlin |
1289 |
my $to = $from + $range_size; |
68 |
dpavlin |
1280 |
|
69 |
dpavlin |
1289 |
$mech->submit_form( |
70 |
|
|
form_name => 'summary_output_form', |
71 |
|
|
fields => { |
72 |
|
|
record_select_type => 'range', |
73 |
|
|
mark_from => $from, |
74 |
|
|
mark_to => $to, |
75 |
|
|
mark_id => 'WOS', |
76 |
dpavlin |
1280 |
|
77 |
dpavlin |
1289 |
qo_fields => 'fullrecord', |
78 |
|
|
citedref => 'citedref', |
79 |
dpavlin |
1280 |
|
80 |
dpavlin |
1289 |
save_options => 'plain_text', |
81 |
dpavlin |
1280 |
|
82 |
dpavlin |
1289 |
fields => 'Full', |
83 |
|
|
format => 'save', |
84 |
|
|
}, |
85 |
|
|
button => 'save', |
86 |
|
|
); |
87 |
|
|
save_mech $mech; |
88 |
dpavlin |
1281 |
|
89 |
dpavlin |
1289 |
if ( $mech->content =~ m{invalid API call} ) { |
90 |
|
|
$mech->back; |
91 |
dpavlin |
1290 |
last; |
92 |
dpavlin |
1289 |
} |
93 |
dpavlin |
1281 |
|
94 |
dpavlin |
1290 |
warn "range $from - $to [$q]\n"; |
95 |
dpavlin |
1289 |
$mech->follow_link( url_regex => qr/save_file/ ); |
96 |
|
|
save_mech $mech => "/tmp/isi.$q.$from-$to.txt"; |
97 |
dpavlin |
1281 |
|
98 |
dpavlin |
1289 |
$from += $range_size; |
99 |
dpavlin |
1281 |
|
100 |
dpavlin |
1289 |
$mech->back; |
101 |
|
|
$mech->back; |
102 |
|
|
#save_mech $mech; |
103 |
dpavlin |
1281 |
|
104 |
dpavlin |
1290 |
} |
105 |
dpavlin |
1289 |
|
106 |
dpavlin |
1281 |
} |
107 |
dpavlin |
1289 |
|
108 |
|
|
|
109 |
dpavlin |
1290 |
sub citations { |
110 |
|
|
save_mech $mech; |
111 |
|
|
warn "# citation report"; |
112 |
|
|
$mech->follow_link( url_regex => qr/search_mode=CitationReport/ ); |
113 |
|
|
save_mech $mech; |
114 |
dpavlin |
1289 |
|
115 |
dpavlin |
1290 |
warn "view citing articles"; |
116 |
|
|
$mech->follow_link( url_regex => qr/search_mode=TotalCitingArticles/ ); |
117 |
|
|
save_mech $mech; |
118 |
|
|
} |
119 |
dpavlin |
1289 |
|
120 |
dpavlin |
1290 |
sub years { |
121 |
dpavlin |
1291 |
my $years_url = $mech->find_link( text_regex => qr/more options/ ); |
122 |
|
|
if ( ! $years_url ) { |
123 |
|
|
warn "W: can't find years\n"; |
124 |
|
|
return; |
125 |
|
|
} |
126 |
|
|
$years_url = $years_url->url_abs; |
127 |
dpavlin |
1290 |
warn "## $years_url"; |
128 |
dpavlin |
1291 |
if ( $years_url !~ s{ra_name=\w+}{ra_name=PublicationYear} ) { |
129 |
|
|
warn "W: no ra_name\n"; |
130 |
|
|
return; |
131 |
|
|
} |
132 |
dpavlin |
1290 |
warn "# refine years (hidden by javascript)"; |
133 |
|
|
warn "http://apps.isiknowledge.com/RAMore.do?product=WOS&search_mode=TotalCitingArticles&SID=T1o6bChdN9PGP1LN1Nh&qid=3&ra_mode=more&ra_name=PublicationYear&db_id=WOS&viewType=raMore\n$years_url\n"; |
134 |
|
|
$mech->get( $years_url ); |
135 |
|
|
save_mech $mech; |
136 |
|
|
|
137 |
|
|
my $html = $mech->content; |
138 |
dpavlin |
1291 |
my $years; |
139 |
dpavlin |
1290 |
while ( $html =~ s{>(\d\d\d\d)\s\((\d+)\)</label.+?value="PublicationYear_}{} ) { |
140 |
dpavlin |
1291 |
$years->{$1} = $2; |
141 |
dpavlin |
1290 |
} |
142 |
dpavlin |
1291 |
warn "# years ",dump $years; |
143 |
dpavlin |
1290 |
$mech->back; |
144 |
dpavlin |
1291 |
return $years; |
145 |
dpavlin |
1290 |
} |
146 |
|
|
|
147 |
dpavlin |
1291 |
search; |
148 |
dpavlin |
1290 |
years; |
149 |
dpavlin |
1291 |
get_results; |
150 |
dpavlin |
1290 |
|
151 |
dpavlin |
1291 |
|
152 |
dpavlin |
1290 |
citations; |
153 |
|
|
years; |
154 |
dpavlin |
1291 |
$q .= '.citing'; |
155 |
|
|
get_results; |
156 |
dpavlin |
1289 |
|