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