1 |
#!/usr/bin/perl |
2 |
|
3 |
use warnings; |
4 |
use strict; |
5 |
|
6 |
use WWW::Mechanize; |
7 |
use Data::Dump qw(dump); |
8 |
use File::Path; |
9 |
use Text::Unaccent; |
10 |
|
11 |
# Advanced search syntax: |
12 |
# http://images.isiknowledge.com/WOK46/help/WOS/h_advanced_examples.html |
13 |
|
14 |
our $q = 'AD=Croatia'; |
15 |
my $range_size = 500; |
16 |
my $overlap = 3; # between previous and this range |
17 |
my $skip_results = 1; |
18 |
my $cites_by_year = 0; |
19 |
|
20 |
my $max_cites = 5000; # ISI limit to get cites |
21 |
|
22 |
if ( 0 ) { |
23 |
$q = 'TS=psychology AND AD=Croatia'; |
24 |
$range_size = 50; |
25 |
$overlap = 0; |
26 |
$max_cites = 50; |
27 |
} |
28 |
|
29 |
$q = unac_string( 'utf-8', join(' ', @ARGV) ) if @ARGV; |
30 |
|
31 |
our $mech = WWW::Mechanize->new( |
32 |
autocheck => 0, # it dies in reference download with it! |
33 |
cookie_jar => undef, |
34 |
); |
35 |
|
36 |
our $step = 0; |
37 |
our @ranges; |
38 |
|
39 |
my $dir = '/tmp/isi/'; |
40 |
#rmtree $dir if -e $dir; |
41 |
mkdir $dir unless -d $dir; |
42 |
|
43 |
sub save_mech { |
44 |
my $path = shift; |
45 |
$step++; |
46 |
my $base_path = sprintf('%s/%04d', $dir,$step); |
47 |
$path ||= $base_path; |
48 |
$path .= $mech->{ct} =~ m{html}i ? '.html' : '.txt'; |
49 |
$mech->save_content( $path ); |
50 |
warn "# [$step] $path ", -s $path, " ", $mech->ct, "\n"; |
51 |
open(my $dump, '>', "$base_path.dump.txt"); |
52 |
$mech->dump_all($dump); |
53 |
} |
54 |
|
55 |
warn "# get session"; |
56 |
$mech->get( 'http://isiknowledge.com/?DestApp=WOS' ); |
57 |
save_mech; |
58 |
|
59 |
sub search { |
60 |
warn "# advanced serach"; |
61 |
$mech->follow_link( url_regex => qr/AdvancedSearch/ ); |
62 |
save_mech; |
63 |
|
64 |
warn "# cookie_jar ", dump $mech->cookie_jar; |
65 |
|
66 |
my $q_this = $q; |
67 |
|
68 |
if ( @ranges ) { |
69 |
$q_this .= ' AND (' . join(' OR ', map { "PY=$_" } @{ shift @ranges } ) . ')'; |
70 |
} |
71 |
|
72 |
warn "# submit_form search: $q_this\n"; |
73 |
$mech->submit_form( |
74 |
fields => { |
75 |
'value(input1)' => $q_this, |
76 |
}, |
77 |
); |
78 |
save_mech; |
79 |
|
80 |
warn "# summary"; |
81 |
$mech->follow_link( url_regex => qr/summary/ ); |
82 |
save_mech; |
83 |
} |
84 |
|
85 |
sub get_results { |
86 |
my $desc = shift; |
87 |
my $from = 1; |
88 |
|
89 |
while ( 1 ) { |
90 |
|
91 |
my $to = $from + $range_size; |
92 |
|
93 |
warn "# submit_form results $from - $to\n"; |
94 |
|
95 |
$mech->submit_form( |
96 |
form_name => 'summary_output_form', |
97 |
fields => { |
98 |
record_select_type => 'range', |
99 |
mark_from => $from, |
100 |
mark_to => $to, |
101 |
mark_id => 'WOS', |
102 |
|
103 |
qo_fields => 'fullrecord', |
104 |
citedref => 'citedref', |
105 |
|
106 |
save_options => 'plain_text', |
107 |
|
108 |
fields => 'Full', |
109 |
format => 'save', |
110 |
}, |
111 |
button => 'save', |
112 |
); |
113 |
save_mech; |
114 |
|
115 |
|
116 |
if ( $mech->content =~ m{invalid API call} ) { |
117 |
$mech->back; |
118 |
last; |
119 |
} |
120 |
|
121 |
|
122 |
my $path = "/tmp/isi.$q.$from-$to"; |
123 |
$path .= '.' . $desc if $desc; |
124 |
|
125 |
warn "save $from - $to into $path\n"; |
126 |
$mech->follow_link( url_regex => qr/save_file/ ); |
127 |
save_mech $path; |
128 |
|
129 |
$from += $range_size - $overlap; |
130 |
|
131 |
$mech->back; |
132 |
$mech->back; |
133 |
#save_mech; |
134 |
} |
135 |
} |
136 |
|
137 |
|
138 |
sub citations { |
139 |
warn "# citation report"; |
140 |
$mech->follow_link( url_regex => qr/search_mode=CitationReport/ ); |
141 |
save_mech; |
142 |
|
143 |
warn "view citing articles"; |
144 |
$mech->follow_link( url_regex => qr/search_mode=TotalCitingArticles/ ); |
145 |
save_mech; |
146 |
} |
147 |
|
148 |
sub years { |
149 |
my $years_url = $mech->find_link( url_regex => qr/ra_name=/ ); |
150 |
if ( ! $years_url ) { |
151 |
warn "W: can't find ra_name link\n"; |
152 |
return; |
153 |
} |
154 |
$years_url = $years_url->url_abs; |
155 |
warn "## $years_url"; |
156 |
if ( $years_url !~ s{ra_name=\w+}{ra_name=PublicationYear} ) { |
157 |
warn "W: no ra_name in $years_url\n"; |
158 |
return; |
159 |
} |
160 |
warn "# refine years (hidden by javascript)"; |
161 |
# warn "http://apps.isiknowledge.com/RAMore.do?product=WOS&search_mode=TotalCitingArticles&SID=T1o6bChdN9PGP1LN1Nh&qid=3&ra_mode=more&ra_name=PublicationYear&db_id=WOS&viewType=raMore\n$years_url\n"; |
162 |
$mech->get( $years_url ); |
163 |
save_mech; |
164 |
|
165 |
my $html = $mech->content; |
166 |
my $years; |
167 |
while ( $html =~ s{<label.+?PublicationYear.+?>(\d{4})\s\(([\d,]+)\)</label>}{} ) { |
168 |
my ( $year, $count ) = ( $1, $2 ); |
169 |
$count =~ s{,}{}g; |
170 |
$years->{$year} = $count; |
171 |
} |
172 |
warn "# years ",dump $years; |
173 |
$mech->back; |
174 |
|
175 |
my @y = sort keys %$years; |
176 |
|
177 |
@ranges = (); |
178 |
|
179 |
if ( $cites_by_year ) { |
180 |
push @ranges, [ $_ ] foreach @y; |
181 |
warn "# cites_by_year ranges ", dump @ranges; |
182 |
return; |
183 |
} |
184 |
|
185 |
my $y = shift @y; |
186 |
my $size = $years->{$y}; |
187 |
|
188 |
my $cites_range; |
189 |
$cites_range = [$y] if $y; |
190 |
|
191 |
foreach my $y ( @y ) { |
192 |
if ( $size + $years->{$y} > $max_cites ) { |
193 |
push @ranges, $cites_range; |
194 |
warn "# cites_range $size years ",dump( $cites_range ),$/; |
195 |
|
196 |
$cites_range = []; |
197 |
$size = 0; |
198 |
} |
199 |
$size += $years->{$y}; |
200 |
push @$cites_range, $y; |
201 |
} |
202 |
|
203 |
if ( $cites_range ) { |
204 |
push @ranges, $cites_range; |
205 |
warn "# cites_range $size years ",dump( $cites_range ), " FINAL\n" |
206 |
} |
207 |
|
208 |
warn '# ranges ', dump @ranges; |
209 |
@ranges = () if $#ranges == 1; # just take all |
210 |
|
211 |
return $years; |
212 |
} |
213 |
|
214 |
search; |
215 |
years; |
216 |
get_results unless $skip_results; |
217 |
|
218 |
|
219 |
citations; |
220 |
|
221 |
do { |
222 |
my $part; |
223 |
if ( @ranges ) { |
224 |
$part .= $ranges[0]->[0] . '.'; |
225 |
search; |
226 |
citations; |
227 |
} |
228 |
$part .= 'citing'; |
229 |
get_results $part; |
230 |
} while ( @ranges ); |
231 |
|