$skip_results to get just citations
[webpac2] / bin / isi-download-results.pl
1 #!/usr/bin/perl
2
3 use warnings;
4 use strict;
5
6 # Advanced search syntax:
7 # http://images.isiknowledge.com/WOK46/help/WOS/h_advanced_examples.html
8
9 our $q = 'AD=Croatia';
10 my $range_size = 500;
11 my $overlap    = 3; # between previous and this range
12 my $skip_results = 1;
13
14 my $max_cites = 5000; # ISI limit to get cites
15
16 if ( 0 ) {
17         $q = 'TS=psychology AND AD=Croatia';
18         $range_size = 50;
19         $overlap    = 0;
20         $max_cites  = 50;
21 }
22
23 use WWW::Mechanize;
24 use Data::Dump qw(dump);
25 use File::Path;
26
27 our $mech = WWW::Mechanize->new(
28         autocheck => 1,
29         cookie_jar => undef,
30 );
31
32 our $step = 0;
33 our @ranges;
34
35 my $dir = '/tmp/isi/';
36 rmtree $dir if -e $dir;
37 mkdir $dir;
38
39 sub save_mech {
40         my $path = shift;
41         $step++;
42         my $base_path = sprintf('%s/%04d', $dir,$step);
43         $path ||= $base_path;
44         $path .= $mech->{ct} =~ m{html}i ? '.html' : '.txt';
45         $mech->save_content( $path );
46         warn "# [$step] $path ", -s $path, " ", $mech->ct, "\n";
47         open(my $dump, '>', "$base_path.dump.txt");
48         $mech->dump_all($dump);
49 }
50
51 warn "# get session";
52 $mech->get( 'http://isiknowledge.com/?DestApp=WOS' );
53 save_mech;
54
55 sub search {
56         warn "# advanced serach";
57         $mech->follow_link( url_regex => qr/AdvancedSearch/ );
58         save_mech;
59
60         warn "# cookie_jar ", dump $mech->cookie_jar;
61
62         my $q_this = $q;
63
64         if ( @ranges ) {
65                 $q_this .= ' AND (' . join(' OR ', map { "PY=$_" } @{ shift @ranges } ) . ')';
66         }
67
68         warn "# submit_form search: $q_this\n";
69         $mech->submit_form(
70                 fields => {
71                         'value(input1)' => $q_this,
72                 },
73         );
74         save_mech;
75
76         warn "# summary";
77         $mech->follow_link( url_regex => qr/summary/ );
78         save_mech;
79 }
80
81 sub get_results {
82         my $desc = shift;
83         my $from = 1;
84
85         while ( 1 ) {
86
87                 my $to = $from + $range_size;
88
89                 warn "# submit_form results $from - $to\n";
90
91                 $mech->submit_form(
92                         form_name => 'summary_output_form',
93                         fields => {
94                                 record_select_type => 'range',
95                                 mark_from => $from,
96                                 mark_to => $to,
97                                 mark_id => 'WOS',
98
99                                 qo_fields => 'fullrecord',
100                                 citedref => 'citedref',
101
102                                 save_options => 'plain_text',
103
104                                 fields => 'Full',
105                                 format => 'save',
106                         },
107                         button => 'save',
108                 );
109                 save_mech;
110
111
112                 if ( $mech->content =~ m{invalid API call} ) {
113                         $mech->back;
114                         last;
115                 }
116
117
118                 my $path = "/tmp/isi.$q.$from-$to";
119                 $path .= '.' . $desc if $desc;
120
121                 warn "save $from - $to into $path\n";
122                 $mech->follow_link( url_regex => qr/save_file/ );
123                 save_mech $path;
124
125                 $from += $range_size - $overlap;
126
127                 $mech->back;
128                 $mech->back;
129                 #save_mech;
130         }
131 }
132
133
134 sub citations {
135         warn "# citation report";
136         $mech->follow_link( url_regex => qr/search_mode=CitationReport/ );
137         save_mech;
138
139         warn "view citing articles";
140         $mech->follow_link( url_regex => qr/search_mode=TotalCitingArticles/ );
141         save_mech;
142 }
143
144 sub years {
145         my $years_url = $mech->find_link( url_regex => qr/ra_name=/ );
146         if ( ! $years_url ) {
147                 warn "W: can't find ra_name link\n";
148                 return;
149         }
150         $years_url = $years_url->url_abs;
151         warn "## $years_url";
152         if ( $years_url !~ s{ra_name=\w+}{ra_name=PublicationYear} ) {
153                 warn "W: no ra_name in $years_url\n";
154                 return;
155         }
156         warn "# refine years (hidden by javascript)";
157 #       warn "http://apps.isiknowledge.com/RAMore.do?product=WOS&search_mode=TotalCitingArticles&SID=T1o6bChdN9PGP1LN1Nh&qid=3&ra_mode=more&ra_name=PublicationYear&db_id=WOS&viewType=raMore\n$years_url\n";
158         $mech->get( $years_url );
159         save_mech;
160
161         my $html = $mech->content;
162         my $years;
163         while ( $html =~ s{<label.+?PublicationYear.+?>(\d{4})\s\(([\d,]+)\)</label>}{} ) {
164                 my ( $year, $count ) = ( $1, $2 );
165                 $count =~ s{,}{}g;
166                 $years->{$year} = $count;
167         }
168         warn "# years ",dump $years;
169         $mech->back;
170
171         my @y = sort keys %$years;
172
173         my $y = shift @y;
174         my $size = $years->{$y};
175
176         @ranges = ();
177         my $cites_range;
178         $cites_range = [$y] if $y;
179
180         foreach my $y ( @y ) {
181                 if ( $size + $years->{$y} > $max_cites ) {
182                         push @ranges, $cites_range;
183                         warn "# cites_range $size years ",dump( $cites_range ),$/;
184
185                         $cites_range = [];
186                         $size = 0;
187                 }
188                 $size += $years->{$y};
189                 push @$cites_range, $y;
190         }
191
192         if ( $cites_range ) {
193                 push @ranges, $cites_range;
194                 warn "# cites_range $size years ",dump( $cites_range ), " FINAL\n"
195         }
196
197         warn '# ranges ', dump @ranges;
198         @ranges = () if $#ranges == 1; # just take all
199
200         return $years;
201 }
202
203 search;
204 years;
205 get_results unless $skip_results;
206
207
208 citations;
209
210 do {
211         my $part;
212         if ( @ranges ) {
213                 $part .= $ranges[0]->[0] . '.';
214                 search;
215                 citations;
216         }
217         $part .= 'citing';
218         get_results $part;
219 } while ( @ranges );
220