git.rot13.org Git - webpac2/blob - 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 my $cites_by_year = 1;
  14
  15 my $max_cites = 5000; # ISI limit to get cites
  16
  17 if ( 0 ) {
  18         $q = 'TS=psychology AND AD=Croatia';
  19         $range_size = 50;
  20         $overlap    = 0;
  21         $max_cites  = 50;
  22 }
  23
  24 use WWW::Mechanize;
  25 use Data::Dump qw(dump);
  26 use File::Path;
  27
  28 our $mech = WWW::Mechanize->new(
  29         autocheck => 1,
  30         cookie_jar => undef,
  31 );
  32
  33 our $step = 0;
  34 our @ranges;
  35
  36 my $dir = '/tmp/isi/';
  37 rmtree $dir if -e $dir;
  38 mkdir $dir;
  39
  40 sub save_mech {
  41         my $path = shift;
  42         $step++;
  43         my $base_path = sprintf('%s/%04d', $dir,$step);
  44         $path ||= $base_path;
  45         $path .= $mech->{ct} =~ m{html}i ? '.html' : '.txt';
  46         $mech->save_content( $path );
  47         warn "# [$step] $path ", -s $path, " ", $mech->ct, "\n";
  48         open(my $dump, '>', "$base_path.dump.txt");
  49         $mech->dump_all($dump);
  50 }
  51
  52 warn "# get session";
  53 $mech->get( 'http://isiknowledge.com/?DestApp=WOS' );
  54 save_mech;
  55
  56 sub search {
  57         warn "# advanced serach";
  58         $mech->follow_link( url_regex => qr/AdvancedSearch/ );
  59         save_mech;
  60
  61         warn "# cookie_jar ", dump $mech->cookie_jar;
  62
  63         my $q_this = $q;
  64
  65         if ( @ranges ) {
  66                 $q_this .= ' AND (' . join(' OR ', map { "PY=$_" } @{ shift @ranges } ) . ')';
  67         }
  68
  69         warn "# submit_form search: $q_this\n";
  70         $mech->submit_form(
  71                 fields => {
  72                         'value(input1)' => $q_this,
  73                 },
  74         );
  75         save_mech;
  76
  77         warn "# summary";
  78         $mech->follow_link( url_regex => qr/summary/ );
  79         save_mech;
  80 }
  81
  82 sub get_results {
  83         my $desc = shift;
  84         my $from = 1;
  85
  86         while ( 1 ) {
  87
  88                 my $to = $from + $range_size;
  89
  90                 warn "# submit_form results $from - $to\n";
  91
  92                 $mech->submit_form(
  93                         form_name => 'summary_output_form',
  94                         fields => {
  95                                 record_select_type => 'range',
  96                                 mark_from => $from,
  97                                 mark_to => $to,
  98                                 mark_id => 'WOS',
  99
 100                                 qo_fields => 'fullrecord',
 101                                 citedref => 'citedref',
 102
 103                                 save_options => 'plain_text',
 104
 105                                 fields => 'Full',
 106                                 format => 'save',
 107                         },
 108                         button => 'save',
 109                 );
 110                 save_mech;
 111
 112
 113                 if ( $mech->content =~ m{invalid API call} ) {
 114                         $mech->back;
 115                         last;
 116                 }
 117
 118
 119                 my $path = "/tmp/isi.$q.$from-$to";
 120                 $path .= '.' . $desc if $desc;
 121
 122                 warn "save $from - $to into $path\n";
 123                 $mech->follow_link( url_regex => qr/save_file/ );
 124                 save_mech $path;
 125
 126                 $from += $range_size - $overlap;
 127
 128                 $mech->back;
 129                 $mech->back;
 130                 #save_mech;
 131         }
 132 }
 133
 134
 135 sub citations {
 136         warn "# citation report";
 137         $mech->follow_link( url_regex => qr/search_mode=CitationReport/ );
 138         save_mech;
 139
 140         warn "view citing articles";
 141         $mech->follow_link( url_regex => qr/search_mode=TotalCitingArticles/ );
 142         save_mech;
 143 }
 144
 145 sub years {
 146         my $years_url = $mech->find_link( url_regex => qr/ra_name=/ );
 147         if ( ! $years_url ) {
 148                 warn "W: can't find ra_name link\n";
 149                 return;
 150         }
 151         $years_url = $years_url->url_abs;
 152         warn "## $years_url";
 153         if ( $years_url !~ s{ra_name=\w+}{ra_name=PublicationYear} ) {
 154                 warn "W: no ra_name in $years_url\n";
 155                 return;
 156         }
 157         warn "# refine years (hidden by javascript)";
 158 #       warn "http://apps.isiknowledge.com/RAMore.do?product=WOS&search_mode=TotalCitingArticles&SID=T1o6bChdN9PGP1LN1Nh&qid=3&ra_mode=more&ra_name=PublicationYear&db_id=WOS&viewType=raMore\n$years_url\n";
 159         $mech->get( $years_url );
 160         save_mech;
 161
 162         my $html = $mech->content;
 163         my $years;
 164         while ( $html =~ s{<label.+?PublicationYear.+?>(\d{4})\s\(([\d,]+)\)</label>}{} ) {
 165                 my ( $year, $count ) = ( $1, $2 );
 166                 $count =~ s{,}{}g;
 167                 $years->{$year} = $count;
 168         }
 169         warn "# years ",dump $years;
 170         $mech->back;
 171
 172         my @y = sort keys %$years;
 173
 174         @ranges = ();
 175
 176         if ( $cites_by_year ) {
 177                 push @ranges, [ $_ ] foreach @y;
 178                 warn "# cites_by_year ranges ", dump @ranges;
 179                 return;
 180         }
 181
 182         my $y = shift @y;
 183         my $size = $years->{$y};
 184
 185         my $cites_range;
 186         $cites_range = [$y] if $y;
 187
 188         foreach my $y ( @y ) {
 189                 if ( $size + $years->{$y} > $max_cites ) {
 190                         push @ranges, $cites_range;
 191                         warn "# cites_range $size years ",dump( $cites_range ),$/;
 192
 193                         $cites_range = [];
 194                         $size = 0;
 195                 }
 196                 $size += $years->{$y};
 197                 push @$cites_range, $y;
 198         }
 199
 200         if ( $cites_range ) {
 201                 push @ranges, $cites_range;
 202                 warn "# cites_range $size years ",dump( $cites_range ), " FINAL\n"
 203         }
 204
 205         warn '# ranges ', dump @ranges;
 206         @ranges = () if $#ranges == 1; # just take all
 207
 208         return $years;
 209 }
 210
 211 search;
 212 years;
 213 get_results unless $skip_results;
 214
 215
 216 citations;
 217
 218 do {
 219         my $part;
 220         if ( @ranges ) {
 221                 $part .= $ranges[0]->[0] . '.';
 222                 search;
 223                 citations;
 224         }
 225         $part .= 'citing';
 226         get_results $part;
 227 } while ( @ranges );
 228