git.rot13.org Git - webpac2/blob - bin/isi-download-results.pl

   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 my $range_size = 500;
  16 my $overlap    = 3; # between previous and this range
  17 my $results = 0;
  18 my $citations = 0;
  19 my $cites_by_year = 0;
  20
  21 my $max_cites = 5000; # ISI limit to get cites
  22
  23 if ( 0 ) {
  24         $q = 'TS=psychology AND AD=Croatia';
  25         $range_size = 50;
  26         $overlap    = 0;
  27         $max_cites  = 50;
  28 }
  29
  30 $q = unac_string( 'utf-8', join(' ', @ARGV) ) if @ARGV;
  31
  32 our $mech = WWW::Mechanize->new(
  33         autocheck => 0, # it dies in reference download with it!
  34         cookie_jar => undef,
  35 );
  36
  37 our $step = 0;
  38 our @ranges;
  39
  40 my $dir = '/tmp/isi/';
  41 #rmtree $dir if -e $dir;
  42 mkdir $dir unless -d $dir;
  43
  44 sub save_mech {
  45         my $path = shift;
  46         $step++;
  47         my $base_path = sprintf('%s/%04d', $dir,$step);
  48         $path ||= $base_path;
  49         $path .= $mech->{ct} =~ m{html}i ? '.html' : '.txt';
  50         $mech->save_content( $path );
  51         warn "# [$step] $path ", -s $path, " ", $mech->ct, "\n";
  52         open(my $dump, '>', "$base_path.dump.txt");
  53         $mech->dump_all($dump);
  54 }
  55
  56 warn "# get session";
  57 $mech->get( 'http://isiknowledge.com/?DestApp=WOS' );
  58 save_mech;
  59
  60 sub search {
  61         warn "# advanced serach";
  62         $mech->follow_link( url_regex => qr/AdvancedSearch/ );
  63         save_mech;
  64
  65         warn "# cookie_jar ", dump $mech->cookie_jar;
  66
  67         my $q_this = $q;
  68
  69         if ( @ranges ) {
  70                 $q_this .= ' AND (' . join(' OR ', map { "PY=$_" } @{ shift @ranges } ) . ')';
  71         }
  72
  73         warn "# submit_form search: $q_this\n";
  74         $mech->submit_form(
  75                 fields => {
  76                         'value(input1)' => $q_this,
  77                 },
  78         );
  79         save_mech;
  80
  81         warn "# summary";
  82         $mech->follow_link( url_regex => qr/summary/ );
  83         save_mech;
  84 }
  85
  86 sub get_results {
  87         my $desc = shift;
  88         my $from = 1;
  89
  90         while ( 1 ) {
  91
  92                 my $to = $from + $range_size;
  93
  94                 warn "# submit_form results $from - $to\n";
  95
  96                 $mech->submit_form(
  97                         form_name => 'summary_output_form',
  98                         fields => {
  99                                 record_select_type => 'range',
 100                                 mark_from => $from,
 101                                 mark_to => $to,
 102                                 mark_id => 'WOS',
 103
 104                                 qo_fields => 'fullrecord',
 105                                 citedref => 'citedref',
 106
 107                                 save_options => 'plain_text',
 108
 109                                 fields => 'Full',
 110                                 format => 'save',
 111                         },
 112                         button => 'save',
 113                 );
 114                 save_mech;
 115
 116
 117                 if ( $mech->content =~ m{invalid API call} ) {
 118                         $mech->back;
 119                         last;
 120                 }
 121
 122
 123                 my $path = "/tmp/isi.$q.$from-$to";
 124                 $path .= '.' . $desc if $desc;
 125
 126                 warn "save $from - $to into $path\n";
 127                 $mech->follow_link( url_regex => qr/save_file/ );
 128                 save_mech $path;
 129
 130                 $from += $range_size - $overlap;
 131
 132                 $mech->back;
 133                 $mech->back;
 134                 #save_mech;
 135         }
 136 }
 137
 138
 139 sub citations {
 140         warn "# citation report";
 141         $mech->follow_link( url_regex => qr/search_mode=CitationReport/ );
 142         save_mech;
 143
 144         warn "view citing articles";
 145         $mech->follow_link( url_regex => qr/search_mode=TotalCitingArticles/ );
 146         save_mech;
 147 }
 148
 149 sub years {
 150         my $years_url = $mech->find_link( url_regex => qr/ra_name=/ );
 151         if ( ! $years_url ) {
 152                 warn "W: can't find ra_name link\n";
 153                 return;
 154         }
 155         $years_url = $years_url->url_abs;
 156         warn "## $years_url";
 157         if ( $years_url !~ s{ra_name=\w+}{ra_name=PublicationYear} ) {
 158                 warn "W: no ra_name in $years_url\n";
 159                 return;
 160         }
 161         warn "# refine years (hidden by javascript)";
 162 #       warn "http://apps.isiknowledge.com/RAMore.do?product=WOS&search_mode=TotalCitingArticles&SID=T1o6bChdN9PGP1LN1Nh&qid=3&ra_mode=more&ra_name=PublicationYear&db_id=WOS&viewType=raMore\n$years_url\n";
 163         $mech->get( $years_url );
 164         save_mech;
 165
 166         my $html = $mech->content;
 167         my $years;
 168         while ( $html =~ s{<label.+?PublicationYear.+?>(\d{4})\s\(([\d,]+)\)</label>}{} ) {
 169                 my ( $year, $count ) = ( $1, $2 );
 170                 $count =~ s{,}{}g;
 171                 $years->{$year} = $count;
 172         }
 173         warn "# years ",dump $years;
 174         $mech->back;
 175
 176         my @y = sort keys %$years;
 177
 178         @ranges = ();
 179
 180         if ( $cites_by_year ) {
 181                 push @ranges, [ $_ ] foreach @y;
 182                 warn "# cites_by_year ranges ", dump @ranges;
 183                 return;
 184         }
 185
 186         my $y = shift @y;
 187         my $size = $years->{$y};
 188
 189         my $cites_range;
 190         $cites_range = [$y] if $y;
 191
 192         foreach my $y ( @y ) {
 193                 if ( $size + $years->{$y} > $max_cites ) {
 194                         push @ranges, $cites_range;
 195                         warn "# cites_range $size years ",dump( $cites_range ),$/;
 196
 197                         $cites_range = [];
 198                         $size = 0;
 199                 }
 200                 $size += $years->{$y};
 201                 push @$cites_range, $y;
 202         }
 203
 204         if ( $cites_range ) {
 205                 push @ranges, $cites_range;
 206                 warn "# cites_range $size years ",dump( $cites_range ), " FINAL\n"
 207         }
 208
 209         warn '# ranges ', dump @ranges;
 210         @ranges = () if $#ranges == 1; # just take all
 211
 212         return $years;
 213 }
 214
 215 search;
 216 years;
 217 get_results 'results' if $results;
 218
 219 if ( $citations ) {
 220
 221         citations;
 222
 223         do {
 224                 my $part;
 225                 if ( @ranges ) {
 226                         $part .= $ranges[0]->[0] . '.';
 227                         search;
 228                         citations;
 229                 }
 230                 $part .= 'citing';
 231                 get_results $part;
 232         } while ( @ranges );
 233
 234 }