1 |
dpavlin |
1280 |
#!/usr/bin/perl |
2 |
|
|
|
3 |
|
|
use warnings; |
4 |
|
|
use strict; |
5 |
|
|
|
6 |
dpavlin |
1289 |
my $q = 'AD=Croatia'; |
7 |
|
|
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 |
|
|
use Data::Dump qw/dump/; |
15 |
|
|
|
16 |
dpavlin |
1289 |
our $mech = WWW::Mechanize->new( |
17 |
dpavlin |
1280 |
autocheck => 1, |
18 |
|
|
cookie_jar => undef, |
19 |
|
|
); |
20 |
|
|
|
21 |
dpavlin |
1281 |
our $step = 0; |
22 |
dpavlin |
1280 |
|
23 |
|
|
sub save_mech { |
24 |
dpavlin |
1281 |
my ( $mech, $path ) = @_; |
25 |
|
|
$step++; |
26 |
dpavlin |
1290 |
mkdir '/tmp/isi/' unless -e '/tmp/isi'; |
27 |
|
|
my $base_path = sprintf('/tmp/isi/%04d', $step); |
28 |
|
|
$path ||= $base_path . ( $mech->{ct} =~ m{html}i ? '.html' : '.txt' ); |
29 |
dpavlin |
1281 |
$mech->save_content( $path ); |
30 |
dpavlin |
1290 |
warn "# [$step] $path ", -s $path, " ", $mech->ct, "\n"; |
31 |
|
|
open(my $dump, '>', "$base_path.dump.txt"); |
32 |
|
|
$mech->dump_all($dump); |
33 |
dpavlin |
1280 |
} |
34 |
|
|
|
35 |
|
|
warn "# get session"; |
36 |
|
|
$mech->get( 'http://isiknowledge.com/?DestApp=WOS' ); |
37 |
|
|
save_mech $mech; |
38 |
|
|
|
39 |
dpavlin |
1290 |
sub search { |
40 |
|
|
my $q = shift; |
41 |
dpavlin |
1280 |
|
42 |
dpavlin |
1290 |
warn "# advanced serach"; |
43 |
|
|
$mech->follow_link( url_regex => qr/AdvancedSearch/ ); |
44 |
|
|
save_mech $mech; |
45 |
dpavlin |
1280 |
|
46 |
dpavlin |
1290 |
warn "# cookie_jar ", dump $mech->cookie_jar; |
47 |
dpavlin |
1280 |
|
48 |
dpavlin |
1290 |
$mech->submit_form( |
49 |
|
|
fields => { |
50 |
|
|
'value(input1)' => $q, |
51 |
|
|
} |
52 |
|
|
); |
53 |
|
|
save_mech $mech; |
54 |
dpavlin |
1280 |
|
55 |
dpavlin |
1290 |
warn "# summary"; |
56 |
|
|
$mech->follow_link( url_regex => qr/summary/ ); |
57 |
|
|
save_mech $mech; |
58 |
|
|
} |
59 |
|
|
|
60 |
dpavlin |
1289 |
sub get_results { |
61 |
|
|
my $q = shift; |
62 |
|
|
my $from = 1; |
63 |
dpavlin |
1280 |
|
64 |
dpavlin |
1289 |
while ( 1 ) { |
65 |
dpavlin |
1280 |
|
66 |
dpavlin |
1289 |
my $to = $from + $range_size; |
67 |
dpavlin |
1280 |
|
68 |
dpavlin |
1289 |
$mech->submit_form( |
69 |
|
|
form_name => 'summary_output_form', |
70 |
|
|
fields => { |
71 |
|
|
record_select_type => 'range', |
72 |
|
|
mark_from => $from, |
73 |
|
|
mark_to => $to, |
74 |
|
|
mark_id => 'WOS', |
75 |
dpavlin |
1280 |
|
76 |
dpavlin |
1289 |
qo_fields => 'fullrecord', |
77 |
|
|
citedref => 'citedref', |
78 |
dpavlin |
1280 |
|
79 |
dpavlin |
1289 |
save_options => 'plain_text', |
80 |
dpavlin |
1280 |
|
81 |
dpavlin |
1289 |
fields => 'Full', |
82 |
|
|
format => 'save', |
83 |
|
|
}, |
84 |
|
|
button => 'save', |
85 |
|
|
); |
86 |
|
|
save_mech $mech; |
87 |
dpavlin |
1281 |
|
88 |
dpavlin |
1289 |
if ( $mech->content =~ m{invalid API call} ) { |
89 |
|
|
$mech->back; |
90 |
dpavlin |
1290 |
last; |
91 |
dpavlin |
1289 |
} |
92 |
dpavlin |
1281 |
|
93 |
dpavlin |
1290 |
warn "range $from - $to [$q]\n"; |
94 |
dpavlin |
1289 |
$mech->follow_link( url_regex => qr/save_file/ ); |
95 |
|
|
save_mech $mech => "/tmp/isi.$q.$from-$to.txt"; |
96 |
dpavlin |
1281 |
|
97 |
dpavlin |
1289 |
$from += $range_size; |
98 |
dpavlin |
1281 |
|
99 |
dpavlin |
1289 |
$mech->back; |
100 |
|
|
$mech->back; |
101 |
|
|
#save_mech $mech; |
102 |
dpavlin |
1281 |
|
103 |
dpavlin |
1290 |
} |
104 |
dpavlin |
1289 |
|
105 |
dpavlin |
1281 |
} |
106 |
dpavlin |
1289 |
|
107 |
|
|
|
108 |
dpavlin |
1290 |
sub citations { |
109 |
|
|
save_mech $mech; |
110 |
|
|
warn "# citation report"; |
111 |
|
|
$mech->follow_link( url_regex => qr/search_mode=CitationReport/ ); |
112 |
|
|
save_mech $mech; |
113 |
dpavlin |
1289 |
|
114 |
dpavlin |
1290 |
warn "view citing articles"; |
115 |
|
|
$mech->follow_link( url_regex => qr/search_mode=TotalCitingArticles/ ); |
116 |
|
|
save_mech $mech; |
117 |
|
|
} |
118 |
dpavlin |
1289 |
|
119 |
dpavlin |
1290 |
sub years { |
120 |
|
|
my $years_url = $mech->find_link( text_regex => qr/more options/ )->url_abs; |
121 |
|
|
warn "## $years_url"; |
122 |
|
|
$years_url =~ s{ra_name=\w+}{ra_name=PublicationYear} || die "ra_name"; |
123 |
|
|
warn "# refine years (hidden by javascript)"; |
124 |
|
|
warn "http://apps.isiknowledge.com/RAMore.do?product=WOS&search_mode=TotalCitingArticles&SID=T1o6bChdN9PGP1LN1Nh&qid=3&ra_mode=more&ra_name=PublicationYear&db_id=WOS&viewType=raMore\n$years_url\n"; |
125 |
|
|
$mech->get( $years_url ); |
126 |
|
|
save_mech $mech; |
127 |
|
|
|
128 |
|
|
my $html = $mech->content; |
129 |
|
|
my @years; |
130 |
|
|
while ( $html =~ s{>(\d\d\d\d)\s\((\d+)\)</label.+?value="PublicationYear_}{} ) { |
131 |
|
|
push @years, [ $1 => $2 ]; |
132 |
|
|
} |
133 |
|
|
warn "# years ",dump @years; |
134 |
|
|
$mech->back; |
135 |
|
|
return @years; |
136 |
|
|
} |
137 |
|
|
|
138 |
|
|
search $q; |
139 |
|
|
years; |
140 |
|
|
get_results $q; |
141 |
|
|
|
142 |
|
|
citations; |
143 |
|
|
years; |
144 |
dpavlin |
1289 |
get_results $q . '.citing'; |
145 |
|
|
|