1 |
dpavlin |
1280 |
#!/usr/bin/perl |
2 |
|
|
|
3 |
|
|
use warnings; |
4 |
|
|
use strict; |
5 |
|
|
|
6 |
dpavlin |
1332 |
use WWW::Mechanize; |
7 |
|
|
use Data::Dump qw(dump); |
8 |
|
|
use File::Path; |
9 |
|
|
use Text::Unaccent; |
10 |
|
|
|
11 |
dpavlin |
1292 |
# Advanced search syntax: |
12 |
|
|
# http://images.isiknowledge.com/WOK46/help/WOS/h_advanced_examples.html |
13 |
|
|
|
14 |
dpavlin |
1291 |
our $q = 'AD=Croatia'; |
15 |
dpavlin |
1335 |
$q = 'CA=BRATKO, D'; |
16 |
|
|
|
17 |
dpavlin |
1289 |
my $range_size = 500; |
18 |
dpavlin |
1294 |
my $overlap = 3; # between previous and this range |
19 |
dpavlin |
1335 |
|
20 |
dpavlin |
1334 |
my $results = 0; |
21 |
|
|
my $citations = 0; |
22 |
dpavlin |
1336 |
my $cited_reference = 0; # html tables |
23 |
|
|
my $citing_articles = 1; # as many files as cited articles |
24 |
dpavlin |
1335 |
|
25 |
dpavlin |
1332 |
my $cites_by_year = 0; |
26 |
dpavlin |
1280 |
|
27 |
dpavlin |
1292 |
my $max_cites = 5000; # ISI limit to get cites |
28 |
dpavlin |
1282 |
|
29 |
dpavlin |
1332 |
$q = unac_string( 'utf-8', join(' ', @ARGV) ) if @ARGV; |
30 |
dpavlin |
1331 |
|
31 |
dpavlin |
1335 |
warn "QUERY: $q\n"; |
32 |
|
|
|
33 |
dpavlin |
1289 |
our $mech = WWW::Mechanize->new( |
34 |
dpavlin |
1333 |
autocheck => 0, # it dies in reference download with it! |
35 |
dpavlin |
1280 |
cookie_jar => undef, |
36 |
|
|
); |
37 |
|
|
|
38 |
dpavlin |
1281 |
our $step = 0; |
39 |
dpavlin |
1294 |
our @ranges; |
40 |
dpavlin |
1280 |
|
41 |
dpavlin |
1291 |
my $dir = '/tmp/isi/'; |
42 |
dpavlin |
1331 |
#rmtree $dir if -e $dir; |
43 |
|
|
mkdir $dir unless -d $dir; |
44 |
dpavlin |
1291 |
|
45 |
dpavlin |
1280 |
sub save_mech { |
46 |
dpavlin |
1292 |
my $path = shift; |
47 |
dpavlin |
1281 |
$step++; |
48 |
dpavlin |
1291 |
my $base_path = sprintf('%s/%04d', $dir,$step); |
49 |
dpavlin |
1292 |
$path ||= $base_path; |
50 |
|
|
$path .= $mech->{ct} =~ m{html}i ? '.html' : '.txt'; |
51 |
dpavlin |
1281 |
$mech->save_content( $path ); |
52 |
dpavlin |
1290 |
warn "# [$step] $path ", -s $path, " ", $mech->ct, "\n"; |
53 |
|
|
open(my $dump, '>', "$base_path.dump.txt"); |
54 |
|
|
$mech->dump_all($dump); |
55 |
dpavlin |
1280 |
} |
56 |
|
|
|
57 |
|
|
warn "# get session"; |
58 |
|
|
$mech->get( 'http://isiknowledge.com/?DestApp=WOS' ); |
59 |
dpavlin |
1292 |
save_mech; |
60 |
dpavlin |
1280 |
|
61 |
dpavlin |
1290 |
sub search { |
62 |
|
|
warn "# advanced serach"; |
63 |
|
|
$mech->follow_link( url_regex => qr/AdvancedSearch/ ); |
64 |
dpavlin |
1292 |
save_mech; |
65 |
dpavlin |
1280 |
|
66 |
dpavlin |
1290 |
warn "# cookie_jar ", dump $mech->cookie_jar; |
67 |
dpavlin |
1280 |
|
68 |
dpavlin |
1292 |
my $q_this = $q; |
69 |
|
|
|
70 |
|
|
if ( @ranges ) { |
71 |
|
|
$q_this .= ' AND (' . join(' OR ', map { "PY=$_" } @{ shift @ranges } ) . ')'; |
72 |
|
|
} |
73 |
|
|
|
74 |
|
|
warn "# submit_form search: $q_this\n"; |
75 |
dpavlin |
1290 |
$mech->submit_form( |
76 |
|
|
fields => { |
77 |
dpavlin |
1292 |
'value(input1)' => $q_this, |
78 |
|
|
}, |
79 |
dpavlin |
1290 |
); |
80 |
dpavlin |
1292 |
save_mech; |
81 |
dpavlin |
1280 |
|
82 |
dpavlin |
1290 |
warn "# summary"; |
83 |
|
|
$mech->follow_link( url_regex => qr/summary/ ); |
84 |
dpavlin |
1292 |
save_mech; |
85 |
dpavlin |
1290 |
} |
86 |
|
|
|
87 |
dpavlin |
1289 |
sub get_results { |
88 |
dpavlin |
1292 |
my $desc = shift; |
89 |
dpavlin |
1289 |
my $from = 1; |
90 |
dpavlin |
1280 |
|
91 |
dpavlin |
1289 |
while ( 1 ) { |
92 |
dpavlin |
1280 |
|
93 |
dpavlin |
1289 |
my $to = $from + $range_size; |
94 |
dpavlin |
1280 |
|
95 |
dpavlin |
1292 |
warn "# submit_form results $from - $to\n"; |
96 |
|
|
|
97 |
dpavlin |
1289 |
$mech->submit_form( |
98 |
|
|
form_name => 'summary_output_form', |
99 |
|
|
fields => { |
100 |
|
|
record_select_type => 'range', |
101 |
|
|
mark_from => $from, |
102 |
|
|
mark_to => $to, |
103 |
|
|
mark_id => 'WOS', |
104 |
dpavlin |
1280 |
|
105 |
dpavlin |
1289 |
qo_fields => 'fullrecord', |
106 |
|
|
citedref => 'citedref', |
107 |
dpavlin |
1280 |
|
108 |
dpavlin |
1289 |
save_options => 'plain_text', |
109 |
dpavlin |
1280 |
|
110 |
dpavlin |
1289 |
fields => 'Full', |
111 |
|
|
format => 'save', |
112 |
|
|
}, |
113 |
|
|
button => 'save', |
114 |
|
|
); |
115 |
dpavlin |
1292 |
save_mech; |
116 |
dpavlin |
1281 |
|
117 |
dpavlin |
1292 |
|
118 |
dpavlin |
1289 |
if ( $mech->content =~ m{invalid API call} ) { |
119 |
|
|
$mech->back; |
120 |
dpavlin |
1290 |
last; |
121 |
dpavlin |
1289 |
} |
122 |
dpavlin |
1281 |
|
123 |
dpavlin |
1336 |
if ( $mech->content =~ m{Please wait while your request is processed} ) { |
124 |
|
|
warn "WARNING: processing request"; |
125 |
|
|
} |
126 |
dpavlin |
1292 |
|
127 |
dpavlin |
1336 |
|
128 |
dpavlin |
1292 |
my $path = "/tmp/isi.$q.$from-$to"; |
129 |
|
|
$path .= '.' . $desc if $desc; |
130 |
|
|
|
131 |
|
|
warn "save $from - $to into $path\n"; |
132 |
dpavlin |
1289 |
$mech->follow_link( url_regex => qr/save_file/ ); |
133 |
dpavlin |
1292 |
save_mech $path; |
134 |
dpavlin |
1281 |
|
135 |
dpavlin |
1292 |
$from += $range_size - $overlap; |
136 |
dpavlin |
1281 |
|
137 |
dpavlin |
1289 |
$mech->back; |
138 |
|
|
$mech->back; |
139 |
dpavlin |
1292 |
#save_mech; |
140 |
dpavlin |
1290 |
} |
141 |
dpavlin |
1281 |
} |
142 |
dpavlin |
1289 |
|
143 |
|
|
|
144 |
dpavlin |
1290 |
sub citations { |
145 |
|
|
warn "# citation report"; |
146 |
|
|
$mech->follow_link( url_regex => qr/search_mode=CitationReport/ ); |
147 |
dpavlin |
1292 |
save_mech; |
148 |
dpavlin |
1289 |
|
149 |
dpavlin |
1290 |
warn "view citing articles"; |
150 |
|
|
$mech->follow_link( url_regex => qr/search_mode=TotalCitingArticles/ ); |
151 |
dpavlin |
1292 |
save_mech; |
152 |
dpavlin |
1290 |
} |
153 |
dpavlin |
1289 |
|
154 |
dpavlin |
1290 |
sub years { |
155 |
dpavlin |
1292 |
my $years_url = $mech->find_link( url_regex => qr/ra_name=/ ); |
156 |
dpavlin |
1291 |
if ( ! $years_url ) { |
157 |
dpavlin |
1292 |
warn "W: can't find ra_name link\n"; |
158 |
dpavlin |
1291 |
return; |
159 |
|
|
} |
160 |
|
|
$years_url = $years_url->url_abs; |
161 |
dpavlin |
1290 |
warn "## $years_url"; |
162 |
dpavlin |
1291 |
if ( $years_url !~ s{ra_name=\w+}{ra_name=PublicationYear} ) { |
163 |
dpavlin |
1292 |
warn "W: no ra_name in $years_url\n"; |
164 |
dpavlin |
1291 |
return; |
165 |
|
|
} |
166 |
dpavlin |
1290 |
warn "# refine years (hidden by javascript)"; |
167 |
dpavlin |
1292 |
# warn "http://apps.isiknowledge.com/RAMore.do?product=WOS&search_mode=TotalCitingArticles&SID=T1o6bChdN9PGP1LN1Nh&qid=3&ra_mode=more&ra_name=PublicationYear&db_id=WOS&viewType=raMore\n$years_url\n"; |
168 |
dpavlin |
1290 |
$mech->get( $years_url ); |
169 |
dpavlin |
1292 |
save_mech; |
170 |
dpavlin |
1290 |
|
171 |
|
|
my $html = $mech->content; |
172 |
dpavlin |
1291 |
my $years; |
173 |
dpavlin |
1292 |
while ( $html =~ s{<label.+?PublicationYear.+?>(\d{4})\s\(([\d,]+)\)</label>}{} ) { |
174 |
|
|
my ( $year, $count ) = ( $1, $2 ); |
175 |
|
|
$count =~ s{,}{}g; |
176 |
|
|
$years->{$year} = $count; |
177 |
dpavlin |
1290 |
} |
178 |
dpavlin |
1291 |
warn "# years ",dump $years; |
179 |
dpavlin |
1290 |
$mech->back; |
180 |
dpavlin |
1292 |
|
181 |
|
|
my @y = sort keys %$years; |
182 |
|
|
|
183 |
dpavlin |
1300 |
@ranges = (); |
184 |
|
|
|
185 |
|
|
if ( $cites_by_year ) { |
186 |
|
|
push @ranges, [ $_ ] foreach @y; |
187 |
|
|
warn "# cites_by_year ranges ", dump @ranges; |
188 |
|
|
return; |
189 |
|
|
} |
190 |
|
|
|
191 |
dpavlin |
1292 |
my $y = shift @y; |
192 |
|
|
my $size = $years->{$y}; |
193 |
|
|
|
194 |
|
|
my $cites_range; |
195 |
|
|
$cites_range = [$y] if $y; |
196 |
|
|
|
197 |
|
|
foreach my $y ( @y ) { |
198 |
|
|
if ( $size + $years->{$y} > $max_cites ) { |
199 |
|
|
push @ranges, $cites_range; |
200 |
|
|
warn "# cites_range $size years ",dump( $cites_range ),$/; |
201 |
|
|
|
202 |
|
|
$cites_range = []; |
203 |
|
|
$size = 0; |
204 |
|
|
} |
205 |
|
|
$size += $years->{$y}; |
206 |
|
|
push @$cites_range, $y; |
207 |
|
|
} |
208 |
|
|
|
209 |
|
|
if ( $cites_range ) { |
210 |
|
|
push @ranges, $cites_range; |
211 |
|
|
warn "# cites_range $size years ",dump( $cites_range ), " FINAL\n" |
212 |
|
|
} |
213 |
|
|
|
214 |
|
|
warn '# ranges ', dump @ranges; |
215 |
|
|
@ranges = () if $#ranges == 1; # just take all |
216 |
|
|
|
217 |
dpavlin |
1291 |
return $years; |
218 |
dpavlin |
1290 |
} |
219 |
|
|
|
220 |
dpavlin |
1337 |
|
221 |
|
|
our $page = 1; |
222 |
|
|
sub next_page { |
223 |
|
|
$page++; |
224 |
|
|
warn "next_page $page\n"; |
225 |
|
|
|
226 |
|
|
$mech->submit_form( |
227 |
|
|
form_name => 'summary_navigation', |
228 |
|
|
fields => { |
229 |
|
|
'page' => $page, |
230 |
|
|
}, |
231 |
|
|
); |
232 |
|
|
|
233 |
|
|
save_mech; |
234 |
|
|
|
235 |
|
|
$mech->form_name( 'summary_navigation' ); |
236 |
|
|
my $is_next_page = $mech->value('page') == $page; |
237 |
|
|
warn "no next_page" unless $is_next_page; |
238 |
|
|
return $is_next_page; |
239 |
|
|
} |
240 |
|
|
|
241 |
dpavlin |
1335 |
if ( $results ) { |
242 |
|
|
search; |
243 |
|
|
years; |
244 |
|
|
get_results; |
245 |
|
|
} |
246 |
dpavlin |
1290 |
|
247 |
dpavlin |
1334 |
if ( $citations ) { |
248 |
dpavlin |
1291 |
|
249 |
dpavlin |
1334 |
citations; |
250 |
dpavlin |
1335 |
years unless @ranges; |
251 |
dpavlin |
1289 |
|
252 |
dpavlin |
1334 |
do { |
253 |
|
|
my $part; |
254 |
|
|
if ( @ranges ) { |
255 |
|
|
$part .= $ranges[0]->[0] . '.'; |
256 |
|
|
search; |
257 |
|
|
citations; |
258 |
|
|
} |
259 |
|
|
$part .= 'citing'; |
260 |
|
|
get_results $part; |
261 |
|
|
} while ( @ranges ); |
262 |
dpavlin |
1292 |
|
263 |
dpavlin |
1334 |
} |
264 |
dpavlin |
1335 |
|
265 |
dpavlin |
1337 |
|
266 |
|
|
|
267 |
dpavlin |
1336 |
if ( $q =~ m{CA=(.+)} && $cited_reference ) { |
268 |
dpavlin |
1335 |
|
269 |
|
|
my $CA = $1; |
270 |
|
|
|
271 |
|
|
warn "# citated reference search"; |
272 |
|
|
$mech->follow_link( url_regex => qr/CitedReferenceSearch/ ); |
273 |
|
|
save_mech; |
274 |
|
|
|
275 |
|
|
|
276 |
|
|
$mech->submit_form( |
277 |
|
|
form_name => 'WOS_CitedReferenceSearch_input_form', |
278 |
|
|
fields => { |
279 |
|
|
'value(input1)' => $CA, |
280 |
|
|
}, |
281 |
|
|
); |
282 |
|
|
|
283 |
|
|
my $page = 1; |
284 |
|
|
my $records = $1 if $mech->content =~ m/(\d+)\s+records/; |
285 |
|
|
warn "# found $records records\n"; |
286 |
|
|
my $last_span = 'fake'; |
287 |
|
|
|
288 |
|
|
while (1) { |
289 |
|
|
save_mech "/tmp/isi.$q.citedref.$page"; |
290 |
|
|
|
291 |
dpavlin |
1337 |
last unless next_page(); |
292 |
dpavlin |
1335 |
|
293 |
|
|
if ( $mech->content =~ m/(\d+\s*-\s*(\d+))/ ) { |
294 |
|
|
warn "span: $1\n"; |
295 |
|
|
last if $2 == $records; |
296 |
|
|
last if $1 == $last_span; |
297 |
|
|
$last_span = $1; |
298 |
|
|
} elsif ( $page > 5 ) { |
299 |
|
|
warn "ARTIFICALLY LIMITED TO 5 PAGES WITHOUT VALID SPAN!"; |
300 |
|
|
last; |
301 |
|
|
} |
302 |
|
|
|
303 |
|
|
} |
304 |
|
|
|
305 |
|
|
} |
306 |
|
|
|
307 |
dpavlin |
1336 |
if ( $q =~ m{CA=(.+)} && $citing_articles ) { |
308 |
|
|
|
309 |
|
|
search; |
310 |
|
|
|
311 |
|
|
my $orig_q = $q; |
312 |
|
|
my $nr = 0; |
313 |
|
|
|
314 |
dpavlin |
1337 |
do { |
315 |
|
|
|
316 |
dpavlin |
1336 |
foreach my $link ( $mech->find_all_links( url_regex => qr/CitingArticles.do/ ) ) { |
317 |
|
|
$nr++; |
318 |
|
|
warn "link $nr\n"; |
319 |
|
|
$mech->get( $link->url ); |
320 |
|
|
save_mech; |
321 |
|
|
$q = $orig_q . '.citing_article.' . $nr; |
322 |
|
|
get_results; |
323 |
|
|
$mech->back; |
324 |
|
|
$mech->back; |
325 |
dpavlin |
1337 |
|
326 |
|
|
#last if $nr > 3; # FIXME only for development |
327 |
dpavlin |
1336 |
} |
328 |
|
|
|
329 |
dpavlin |
1337 |
} while next_page; |
330 |
|
|
|
331 |
dpavlin |
1336 |
$q = $orig_q; |
332 |
|
|
} |
333 |
|
|
|
334 |
dpavlin |
1335 |
warn "OVER\n"; |