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 |
$q = 'AU=BRATKO, D'; |
16 |
#$q = 'AD=(croat* OR hrvat*)'; |
17 |
|
18 |
my $range_size = 100; |
19 |
my $overlap = 3; # between previous and this range |
20 |
|
21 |
my $results = 1; |
22 |
my $citations = 0; |
23 |
my $cited_reference = 0; # html tables |
24 |
my $citing_articles = 0; # as many files as cited articles |
25 |
|
26 |
my $cites_by_year = 0; |
27 |
|
28 |
my $max_cites = 5000; # ISI limit to get cites |
29 |
|
30 |
$q = unac_string( 'utf-8', join(' ', @ARGV) ) if @ARGV; |
31 |
|
32 |
warn "QUERY: $q\n"; |
33 |
|
34 |
our $mech = WWW::Mechanize->new( |
35 |
autocheck => 0, # it dies in reference download with it! |
36 |
cookie_jar => undef, |
37 |
); |
38 |
|
39 |
our $step = 0; |
40 |
our @ranges; |
41 |
|
42 |
my $dir = '/tmp/isi/'; |
43 |
#rmtree $dir if -e $dir; |
44 |
mkdir $dir unless -d $dir; |
45 |
|
46 |
sub save_mech { |
47 |
my $path = shift; |
48 |
$step++; |
49 |
my $base_path = sprintf('%s/%04d', $dir,$step); |
50 |
$path ||= $base_path; |
51 |
$path .= $mech->{ct} =~ m{html}i ? '.html' : '.txt'; |
52 |
$mech->save_content( $path ); |
53 |
warn "# [$step] $path ", -s $path, " ", $mech->ct, "\n"; |
54 |
open(my $dump, '>', "$base_path.dump.txt"); |
55 |
$mech->dump_all($dump); |
56 |
} |
57 |
|
58 |
warn "# get session"; |
59 |
$mech->get( 'http://www.webofknowledge.com/?DestApp=WOS' ); |
60 |
save_mech; |
61 |
|
62 |
sub search { |
63 |
warn "# advanced serach"; |
64 |
$mech->follow_link( url_regex => qr/AdvancedSearch/ ); |
65 |
save_mech; |
66 |
|
67 |
warn "# cookie_jar ", dump $mech->cookie_jar; |
68 |
|
69 |
my $q_this = $q; |
70 |
|
71 |
if ( @ranges ) { |
72 |
$q_this .= ' AND (' . join(' OR ', map { "PY=$_" } @{ shift @ranges } ) . ')'; |
73 |
} |
74 |
|
75 |
warn "# submit_form search: $q_this\n"; |
76 |
$mech->submit_form( |
77 |
fields => { |
78 |
'value(input1)' => $q_this, |
79 |
}, |
80 |
); |
81 |
save_mech; |
82 |
|
83 |
warn "# summary"; |
84 |
$mech->follow_link( url_regex => qr/summary/ ); |
85 |
save_mech; |
86 |
} |
87 |
|
88 |
sub get_results { |
89 |
my $desc = shift; |
90 |
my $from = 1; |
91 |
|
92 |
while ( 1 ) { |
93 |
|
94 |
my $to = $from + $range_size; |
95 |
|
96 |
warn "# submit_form results $from - $to\n"; |
97 |
save_mech; |
98 |
|
99 |
$mech->submit_form( |
100 |
form_name => 'output_form', |
101 |
fields => { |
102 |
'value(record_select_type)' => 'range', |
103 |
markFrom => $from, |
104 |
markTo => $to, |
105 |
|
106 |
mark_from => $from, |
107 |
mark_to => $to, |
108 |
mark_id => 'WOS', |
109 |
|
110 |
fields_selection => 'ACCESSION_NUM FUNDING SUBJECT_CATEGORY JCR_CATEGORY LANG IDS PAGEC SABBR CITREFC ISSN PUBINFO KEYWORDS CITTIMES ADDRS CONFERENCE_SPONSORS DOCTYPE ABSTRACT CONFERENCE_INFO SOURCE TITLE AUTHORS', |
111 |
filters => 'ACCESSION_NUM FUNDING SUBJECT_CATEGORY JCR_CATEGORY LANG IDS PAGEC SABBR CITREFC ISSN PUBINFO KEYWORDS CITTIMES ADDRS CONFERENCE_SPONSORS DOCTYPE ABSTRACT CONFERENCE_INFO SOURCE TITLE AUTHORS CITREF', |
112 |
fullrec_fields_option => 'CITREF', |
113 |
|
114 |
save_options => 'fieldtagged', |
115 |
format => 'saveToFile', |
116 |
|
117 |
}, |
118 |
button => 'saveToFile', |
119 |
); |
120 |
save_mech; |
121 |
|
122 |
|
123 |
if ( $mech->content =~ m{invalid API call} ) { |
124 |
$mech->back; |
125 |
last; |
126 |
} |
127 |
|
128 |
if ( $mech->content !~ m{Please wait while your request is processed} ) { |
129 |
warn "WARNING: expecting processing request"; |
130 |
} |
131 |
|
132 |
|
133 |
my $path = "/tmp/isi.$q.$from-$to"; |
134 |
$path .= '.' . $desc if $desc; |
135 |
|
136 |
warn "save $from - $to into $path\n"; |
137 |
$mech->submit_form( |
138 |
form_name => 'etsForm', |
139 |
); |
140 |
save_mech $path; |
141 |
|
142 |
$from += $range_size - $overlap; |
143 |
|
144 |
$mech->back; |
145 |
$mech->back; |
146 |
save_mech; |
147 |
} |
148 |
} |
149 |
|
150 |
|
151 |
sub citations { |
152 |
warn "# citation report"; |
153 |
$mech->follow_link( url_regex => qr/search_mode=CitationReport/ ); |
154 |
save_mech; |
155 |
|
156 |
warn "view citing articles"; |
157 |
$mech->follow_link( url_regex => qr/search_mode=TotalCitingArticles/ ); |
158 |
save_mech; |
159 |
} |
160 |
|
161 |
sub years { |
162 |
my $years_url = $mech->find_link( url_regex => qr/ra_name=/ ); |
163 |
if ( ! $years_url ) { |
164 |
warn "W: can't find ra_name link\n"; |
165 |
return; |
166 |
} |
167 |
$years_url = $years_url->url_abs; |
168 |
warn "## $years_url"; |
169 |
if ( $years_url !~ s{ra_name=\w+}{ra_name=PublicationYear} ) { |
170 |
warn "W: no ra_name in $years_url\n"; |
171 |
return; |
172 |
} |
173 |
warn "# refine years (hidden by javascript)"; |
174 |
# warn "http://apps.isiknowledge.com/RAMore.do?product=WOS&search_mode=TotalCitingArticles&SID=T1o6bChdN9PGP1LN1Nh&qid=3&ra_mode=more&ra_name=PublicationYear&db_id=WOS&viewType=raMore\n$years_url\n"; |
175 |
$mech->get( $years_url ); |
176 |
save_mech; |
177 |
|
178 |
my $html = $mech->content; |
179 |
my $years; |
180 |
while ( $html =~ s{<label.+?PublicationYear.+?>(\d{4})\s\(([\d,]+)\)</label>}{} ) { |
181 |
my ( $year, $count ) = ( $1, $2 ); |
182 |
$count =~ s{,}{}g; |
183 |
$years->{$year} = $count; |
184 |
} |
185 |
warn "# years ",dump $years; |
186 |
$mech->back; |
187 |
|
188 |
my @y = sort keys %$years; |
189 |
|
190 |
@ranges = (); |
191 |
|
192 |
if ( $cites_by_year ) { |
193 |
push @ranges, [ $_ ] foreach @y; |
194 |
warn "# cites_by_year ranges ", dump @ranges; |
195 |
return; |
196 |
} |
197 |
|
198 |
my $y = shift @y; |
199 |
my $size = $years->{$y}; |
200 |
|
201 |
my $cites_range; |
202 |
$cites_range = [$y] if $y; |
203 |
|
204 |
foreach my $y ( @y ) { |
205 |
if ( $size + $years->{$y} > $max_cites ) { |
206 |
push @ranges, $cites_range; |
207 |
warn "# cites_range $size years ",dump( $cites_range ),$/; |
208 |
|
209 |
$cites_range = []; |
210 |
$size = 0; |
211 |
} |
212 |
$size += $years->{$y}; |
213 |
push @$cites_range, $y; |
214 |
} |
215 |
|
216 |
if ( $cites_range ) { |
217 |
push @ranges, $cites_range; |
218 |
warn "# cites_range $size years ",dump( $cites_range ), " FINAL\n" |
219 |
} |
220 |
|
221 |
warn '# ranges ', dump @ranges; |
222 |
@ranges = () if $#ranges == 1; # just take all |
223 |
|
224 |
return $years; |
225 |
} |
226 |
|
227 |
|
228 |
our $page = 1; |
229 |
sub next_page { |
230 |
$page++; |
231 |
warn "next_page $page\n"; |
232 |
|
233 |
$mech->submit_form( |
234 |
form_name => 'summary_navigation', |
235 |
fields => { |
236 |
'page' => $page, |
237 |
}, |
238 |
); |
239 |
|
240 |
save_mech; |
241 |
|
242 |
$mech->form_name( 'summary_navigation' ); |
243 |
my $is_next_page = $mech->value('page') == $page; |
244 |
warn "no next_page" unless $is_next_page; |
245 |
return $is_next_page; |
246 |
} |
247 |
|
248 |
if ( $results ) { |
249 |
search; |
250 |
years; |
251 |
get_results; |
252 |
} |
253 |
|
254 |
if ( $citations ) { |
255 |
|
256 |
citations; |
257 |
years unless @ranges; |
258 |
|
259 |
do { |
260 |
my $part; |
261 |
if ( @ranges ) { |
262 |
$part .= $ranges[0]->[0] . '.'; |
263 |
search; |
264 |
citations; |
265 |
} |
266 |
$part .= 'citing'; |
267 |
get_results $part; |
268 |
} while ( @ranges ); |
269 |
|
270 |
} |
271 |
|
272 |
|
273 |
|
274 |
if ( $q =~ m{CA=(.+)} && $cited_reference ) { |
275 |
|
276 |
my $CA = $1; |
277 |
|
278 |
warn "# citated reference search"; |
279 |
$mech->follow_link( url_regex => qr/CitedReferenceSearch/ ); |
280 |
save_mech; |
281 |
|
282 |
|
283 |
$mech->submit_form( |
284 |
form_name => 'WOS_CitedReferenceSearch_input_form', |
285 |
fields => { |
286 |
'value(input1)' => $CA, |
287 |
}, |
288 |
); |
289 |
|
290 |
my $page = 1; |
291 |
my $records = $1 if $mech->content =~ m/(\d+)\s+records/; |
292 |
warn "# found $records records\n"; |
293 |
my $last_span = 'fake'; |
294 |
|
295 |
while (1) { |
296 |
save_mech "/tmp/isi.$q.citedref.$page"; |
297 |
|
298 |
last unless next_page(); |
299 |
|
300 |
if ( $mech->content =~ m/(\d+\s*-\s*(\d+))/ ) { |
301 |
warn "span: $1\n"; |
302 |
last if $2 == $records; |
303 |
last if $1 == $last_span; |
304 |
$last_span = $1; |
305 |
} elsif ( $page > 5 ) { |
306 |
warn "ARTIFICALLY LIMITED TO 5 PAGES WITHOUT VALID SPAN!"; |
307 |
last; |
308 |
} |
309 |
|
310 |
} |
311 |
|
312 |
} |
313 |
|
314 |
if ( $q =~ m{CA=(.+)} && $citing_articles ) { |
315 |
|
316 |
search; |
317 |
|
318 |
my $orig_q = $q; |
319 |
my $nr = 0; |
320 |
|
321 |
do { |
322 |
|
323 |
foreach my $link ( $mech->find_all_links( url_regex => qr/CitingArticles.do/ ) ) { |
324 |
$nr++; |
325 |
warn "link $nr\n"; |
326 |
$mech->get( $link->url ); |
327 |
save_mech; |
328 |
$q = $orig_q . '.citing_article.' . $nr; |
329 |
get_results; |
330 |
$mech->back; |
331 |
$mech->back; |
332 |
|
333 |
#last if $nr > 3; # FIXME only for development |
334 |
} |
335 |
|
336 |
} while next_page; |
337 |
|
338 |
$q = $orig_q; |
339 |
} |
340 |
|
341 |
warn "OVER\n"; |