processing request page is expected

[webpac2] / bin / isi-download-results.pl

diff --git a/bin/isi-download-results.pl b/bin/isi-download-results.pl
index 4d497f3..5bd18be 100755 (executable)
--- a/bin/isi-download-results.pl
+++ b/bin/isi-download-results.pl
@@ -12,24 +12,26 @@ use Text::Unaccent;
 # http://images.isiknowledge.com/WOK46/help/WOS/h_advanced_examples.html
 
 our $q = 'AD=Croatia';
+$q = 'CA=BRATKO, D';
+
 my $range_size = 500;
 my $overlap    = 3; # between previous and this range
-my $skip_results = 0;
+
+my $results = 0;
+my $citations = 0;
+my $cited_reference = 0; # html tables
+my $citing_articles = 1; # as many files as cited articles
+
 my $cites_by_year = 0;
 
 my $max_cites = 5000; # ISI limit to get cites
 
-if ( 0 ) {
-       $q = 'TS=psychology AND AD=Croatia';
-       $range_size = 50;
-       $overlap    = 0;
-       $max_cites  = 50;
-}
-
 $q = unac_string( 'utf-8', join(' ', @ARGV) ) if @ARGV;
 
+warn "QUERY: $q\n";
+
 our $mech = WWW::Mechanize->new(
-       autocheck => 1,
+       autocheck => 0, # it dies in reference download with it!
        cookie_jar => undef,
 );
 
@@ -118,6 +120,10 @@ sub get_results {
                        last;
                }
 
+               if ( $mech->content !~ m{Please wait while your request is processed} ) {
+                       warn "WARNING: expecting processing request";
+               }
+
 
                my $path = "/tmp/isi.$q.$from-$to";
                $path .= '.' . $desc if $desc;
@@ -211,21 +217,118 @@ sub years {
        return $years;
 }
 
-search;
-years;
-get_results unless $skip_results;
 
+our $page = 1;
+sub next_page {
+       $page++;
+       warn "next_page $page\n";
 
-citations;
+       $mech->submit_form(
+               form_name => 'summary_navigation',
+               fields => {
+                       'page' => $page,
+               },
+       );
+
+       save_mech;
+
+       $mech->form_name( 'summary_navigation' );
+       my $is_next_page = $mech->value('page') == $page;
+       warn "no next_page" unless $is_next_page;
+       return $is_next_page;
+}
+
+if ( $results ) {
+       search;
+       years;
+       get_results;
+}
+
+if ( $citations ) {
+
+       citations;
+       years unless @ranges;
+
+       do {
+               my $part;
+               if ( @ranges ) {
+                       $part .= $ranges[0]->[0] . '.';
+                       search;
+                       citations;
+               }
+               $part .= 'citing';
+               get_results $part;
+       } while ( @ranges );
+
+}
+
+
+
+if ( $q =~ m{CA=(.+)} && $cited_reference ) {
+
+       my $CA = $1;
+
+       warn "# citated reference search";
+       $mech->follow_link( url_regex => qr/CitedReferenceSearch/ );
+       save_mech;
+
+
+       $mech->submit_form(
+               form_name => 'WOS_CitedReferenceSearch_input_form',
+               fields => {
+                       'value(input1)' => $CA,
+               },
+       );
+
+       my $page = 1;
+       my $records = $1 if $mech->content =~ m/(\d+)\s+records/;
+       warn "# found $records records\n";
+       my $last_span = 'fake';
+
+       while (1) {
+               save_mech "/tmp/isi.$q.citedref.$page";
+
+               last unless next_page();
+
+               if ( $mech->content =~ m/(\d+\s*-\s*(\d+))/ ) {
+                       warn "span: $1\n";
+                       last if $2 == $records;
+                       last if $1 == $last_span;
+                       $last_span = $1;
+               } elsif ( $page > 5 ) {
+                       warn "ARTIFICALLY LIMITED TO 5 PAGES WITHOUT VALID SPAN!";
+                       last;
+               }
 
-do {
-       my $part;
-       if ( @ranges ) {
-               $part .= $ranges[0]->[0] . '.';
-               search;
-               citations;
        }
-       $part .= 'citing';
-       get_results $part;
-} while ( @ranges );
 
+}
+
+if ( $q =~ m{CA=(.+)} && $citing_articles ) {
+
+       search;
+
+       my $orig_q = $q;
+       my $nr = 0;
+
+       do {
+
+       foreach my $link ( $mech->find_all_links( url_regex => qr/CitingArticles.do/ ) ) {
+               $nr++;
+               warn "link $nr\n";
+               $mech->get( $link->url );
+               save_mech;
+               $q = $orig_q . '.citing_article.' . $nr;
+               get_results;
+               $mech->back;
+               $mech->back;
+
+               #last if $nr > 3; # FIXME only for development
+       }
+
+       } while next_page;
+
+       $q = $orig_q;
+}
+
+warn "OVER\n";