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 |
$q = 'CA=BRATKO, D'; |
17 |
#$q = 'AD=(croat* OR hrvat*)'; |
18 |
|
19 |
my $range_size = 100; |
20 |
my $overlap = 3; # between previous and this range |
21 |
|
22 |
my $results = 0; |
23 |
my $citations = 0; |
24 |
my $cited_reference = 1; # html tables |
25 |
my $citing_articles = 1; # as many files as cited articles |
26 |
|
27 |
my $cites_by_year = 0; |
28 |
|
29 |
my $max_cites = 5000; # ISI limit to get cites |
30 |
|
31 |
$q = unac_string( 'utf-8', join(' ', @ARGV) ) if @ARGV; |
32 |
|
33 |
warn "QUERY: $q\n"; |
34 |
|
35 |
our $mech = WWW::Mechanize->new( |
36 |
autocheck => 0, # it dies in reference download with it! |
37 |
cookie_jar => undef, |
38 |
); |
39 |
|
40 |
our $step = 0; |
41 |
our @ranges; |
42 |
|
43 |
my $dir = '/tmp/isi/'; |
44 |
#rmtree $dir if -e $dir; |
45 |
mkdir $dir unless -d $dir; |
46 |
|
47 |
sub save_mech { |
48 |
my $path = shift; |
49 |
$step++; |
50 |
my $base_path = sprintf('%s/%04d', $dir,$step); |
51 |
$path ||= $base_path; |
52 |
$path .= $mech->{ct} =~ m{html}i ? '.html' : '.txt'; |
53 |
$mech->save_content( $path ); |
54 |
warn "# [$step] $path ", -s $path, " ", $mech->ct, "\n"; |
55 |
open(my $dump, '>', "$base_path.dump.txt"); |
56 |
$mech->dump_all($dump); |
57 |
} |
58 |
|
59 |
warn "# get session"; |
60 |
$mech->get( 'http://www.webofknowledge.com/?DestApp=WOS' ); |
61 |
save_mech; |
62 |
|
63 |
sub search { |
64 |
warn "# advanced serach"; |
65 |
$mech->follow_link( url_regex => qr/AdvancedSearch/ ); |
66 |
save_mech; |
67 |
|
68 |
warn "# cookie_jar ", dump $mech->cookie_jar; |
69 |
|
70 |
my $q_this = $q; |
71 |
|
72 |
if ( @ranges ) { |
73 |
$q_this .= ' AND (' . join(' OR ', map { "PY=$_" } @{ shift @ranges } ) . ')'; |
74 |
} |
75 |
|
76 |
warn "# submit_form search: $q_this\n"; |
77 |
$mech->submit_form( |
78 |
fields => { |
79 |
'value(input1)' => $q_this, |
80 |
}, |
81 |
); |
82 |
save_mech; |
83 |
|
84 |
warn "# summary"; |
85 |
$mech->follow_link( url_regex => qr/summary/ ); |
86 |
save_mech; |
87 |
} |
88 |
|
89 |
sub get_results { |
90 |
my $desc = shift; |
91 |
my $from = 1; |
92 |
|
93 |
while ( 1 ) { |
94 |
|
95 |
my $to = $from + $range_size; |
96 |
|
97 |
warn "# submit_form results $from - $to\n"; |
98 |
save_mech; |
99 |
|
100 |
$mech->submit_form( |
101 |
form_name => 'output_form', |
102 |
fields => { |
103 |
'value(record_select_type)' => 'range', |
104 |
markFrom => $from, |
105 |
markTo => $to, |
106 |
|
107 |
mark_from => $from, |
108 |
mark_to => $to, |
109 |
mark_id => 'WOS', |
110 |
|
111 |
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', |
112 |
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', |
113 |
fullrec_fields_option => 'CITREF', |
114 |
|
115 |
save_options => 'fieldtagged', |
116 |
format => 'saveToFile', |
117 |
|
118 |
}, |
119 |
button => 'saveToFile', |
120 |
); |
121 |
save_mech; |
122 |
|
123 |
|
124 |
if ( $mech->content =~ m{invalid API call} ) { |
125 |
$mech->back; |
126 |
last; |
127 |
} |
128 |
|
129 |
if ( $mech->content !~ m{Please wait while your request is processed} ) { |
130 |
warn "WARNING: expecting processing request"; |
131 |
} |
132 |
|
133 |
|
134 |
my $path = "/tmp/isi.$q.$from-$to"; |
135 |
$path .= '.' . $desc if $desc; |
136 |
|
137 |
warn "save $from - $to into $path\n"; |
138 |
$mech->submit_form( |
139 |
form_name => 'etsForm', |
140 |
); |
141 |
save_mech $path; |
142 |
|
143 |
$from += $range_size - $overlap; |
144 |
|
145 |
$mech->back; |
146 |
$mech->back; |
147 |
save_mech; |
148 |
} |
149 |
} |
150 |
|
151 |
|
152 |
sub citations { |
153 |
warn "# citation report"; |
154 |
$mech->follow_link( url_regex => qr/search_mode=CitationReport/ ); |
155 |
save_mech; |
156 |
|
157 |
warn "view citing articles"; |
158 |
$mech->follow_link( url_regex => qr/search_mode=TotalCitingArticles/ ); |
159 |
save_mech; |
160 |
} |
161 |
|
162 |
sub years { |
163 |
my $years_url = $mech->find_link( url_regex => qr/ra_name=/ ); |
164 |
if ( ! $years_url ) { |
165 |
warn "W: can't find ra_name link\n"; |
166 |
return; |
167 |
} |
168 |
$years_url = $years_url->url_abs; |
169 |
warn "## $years_url"; |
170 |
if ( $years_url !~ s{ra_name=\w+}{ra_name=PublicationYear} ) { |
171 |
warn "W: no ra_name in $years_url\n"; |
172 |
return; |
173 |
} |
174 |
warn "# refine years (hidden by javascript)"; |
175 |
# warn "http://apps.isiknowledge.com/RAMore.do?product=WOS&search_mode=TotalCitingArticles&SID=T1o6bChdN9PGP1LN1Nh&qid=3&ra_mode=more&ra_name=PublicationYear&db_id=WOS&viewType=raMore\n$years_url\n"; |
176 |
$mech->get( $years_url ); |
177 |
save_mech; |
178 |
|
179 |
my $html = $mech->content; |
180 |
my $years; |
181 |
while ( $html =~ s{<label.+?PublicationYear.+?>(\d{4})\s\(([\d,]+)\)</label>}{} ) { |
182 |
my ( $year, $count ) = ( $1, $2 ); |
183 |
$count =~ s{,}{}g; |
184 |
$years->{$year} = $count; |
185 |
} |
186 |
warn "# years ",dump $years; |
187 |
$mech->back; |
188 |
|
189 |
my @y = sort keys %$years; |
190 |
|
191 |
@ranges = (); |
192 |
|
193 |
if ( $cites_by_year ) { |
194 |
push @ranges, [ $_ ] foreach @y; |
195 |
warn "# cites_by_year ranges ", dump @ranges; |
196 |
return; |
197 |
} |
198 |
|
199 |
my $y = shift @y; |
200 |
my $size = $years->{$y}; |
201 |
|
202 |
my $cites_range; |
203 |
$cites_range = [$y] if $y; |
204 |
|
205 |
foreach my $y ( @y ) { |
206 |
if ( $size + $years->{$y} > $max_cites ) { |
207 |
push @ranges, $cites_range; |
208 |
warn "# cites_range $size years ",dump( $cites_range ),$/; |
209 |
|
210 |
$cites_range = []; |
211 |
$size = 0; |
212 |
} |
213 |
$size += $years->{$y}; |
214 |
push @$cites_range, $y; |
215 |
} |
216 |
|
217 |
if ( $cites_range ) { |
218 |
push @ranges, $cites_range; |
219 |
warn "# cites_range $size years ",dump( $cites_range ), " FINAL\n" |
220 |
} |
221 |
|
222 |
warn '# ranges ', dump @ranges; |
223 |
@ranges = () if $#ranges == 1; # just take all |
224 |
|
225 |
return $years; |
226 |
} |
227 |
|
228 |
|
229 |
our $page = 1; |
230 |
sub next_page { |
231 |
$page++; |
232 |
warn "next_page $page\n"; |
233 |
|
234 |
$mech->submit_form( |
235 |
form_name => 'summary_navigation', |
236 |
fields => { |
237 |
'page' => $page, |
238 |
}, |
239 |
); |
240 |
|
241 |
save_mech; |
242 |
|
243 |
$mech->form_name( 'summary_navigation' ); |
244 |
my $is_next_page = $mech->value('page') == $page; |
245 |
warn "no next_page" unless $is_next_page; |
246 |
return $is_next_page; |
247 |
} |
248 |
|
249 |
if ( $results ) { |
250 |
search; |
251 |
years; |
252 |
get_results; |
253 |
} |
254 |
|
255 |
if ( $citations ) { |
256 |
|
257 |
citations; |
258 |
years unless @ranges; |
259 |
|
260 |
do { |
261 |
my $part; |
262 |
if ( @ranges ) { |
263 |
$part .= $ranges[0]->[0] . '.'; |
264 |
search; |
265 |
citations; |
266 |
} |
267 |
$part .= 'citing'; |
268 |
get_results $part; |
269 |
} while ( @ranges ); |
270 |
|
271 |
} |
272 |
|
273 |
|
274 |
|
275 |
if ( $q =~ m{CA=(.+)} && $cited_reference ) { |
276 |
|
277 |
my $CA = $1; |
278 |
|
279 |
warn "# citated reference search"; |
280 |
$mech->follow_link( url_regex => qr/CitedReferenceSearch/ ); |
281 |
save_mech; |
282 |
|
283 |
|
284 |
$mech->submit_form( |
285 |
form_name => 'WOS_CitedReferenceSearch_input_form', |
286 |
fields => { |
287 |
'value(input1)' => $CA, |
288 |
}, |
289 |
); |
290 |
|
291 |
my $page = 1; |
292 |
my $records = $1 if $mech->content =~ m/(\d+)\s+records/; |
293 |
warn "# found $records records\n"; |
294 |
my $last_span = 'fake'; |
295 |
|
296 |
while (1) { |
297 |
save_mech "/tmp/isi.$q.citedref.$page"; |
298 |
|
299 |
last unless next_page(); |
300 |
$page++; |
301 |
|
302 |
if ( $mech->content =~ m/(\d+\s*-\s*(\d+))/ ) { |
303 |
warn "span: $1\n"; |
304 |
last if $2 == $records; |
305 |
last if $1 == $last_span; |
306 |
$last_span = $1; |
307 |
} elsif ( $page > 5 ) { |
308 |
warn "ARTIFICALLY LIMITED TO 5 PAGES WITHOUT VALID SPAN!"; |
309 |
last; |
310 |
} |
311 |
|
312 |
} |
313 |
|
314 |
} |
315 |
|
316 |
if ( $q =~ m{CA=(.+)} && $citing_articles ) { |
317 |
|
318 |
search; |
319 |
|
320 |
my $orig_q = $q; |
321 |
my $nr = 0; |
322 |
|
323 |
do { |
324 |
|
325 |
foreach my $link ( $mech->find_all_links( url_regex => qr/CitingArticles.do/ ) ) { |
326 |
$nr++; |
327 |
warn "link $nr\n"; |
328 |
$mech->get( $link->url ); |
329 |
save_mech; |
330 |
$q = $orig_q . '.citing_article.' . $nr; |
331 |
get_results; |
332 |
$mech->back; |
333 |
$mech->back; |
334 |
|
335 |
#last if $nr > 3; # FIXME only for development |
336 |
} |
337 |
|
338 |
} while next_page; |
339 |
|
340 |
$q = $orig_q; |
341 |
} |
342 |
|
343 |
warn "OVER\n"; |