- get citing records grouped into max_cites sized chunks using PY as criteria

- implement 10 record $overlap when getting more than $range_size (500) records
- cleanup save_mech
- fix PublictionYear extraction from html

git-svn-id: svn+ssh://mjesec/home/dpavlin/svn/webpac2/trunk@1292 07558da8-63fa-0310-ba24-9fe276d99e06

diff --git a/bin/isi-download-results.pl b/bin/isi-download-results.pl
index 8e0f918..ac09112 100755 (executable)
--- a/bin/isi-download-results.pl
+++ b/bin/isi-download-results.pl
@@ -3,12 +3,17 @@
 use warnings;
 use strict;
 
+# Advanced search syntax:
+# http://images.isiknowledge.com/WOK46/help/WOS/h_advanced_examples.html
+
 our $q = 'AD=Croatia';
 my $range_size = 500;
+my $overlap    = 10; # between previous and this range
 
-my $dump = @ARGV ? 1 : 0;
+$q = 'TS=psychology AND AD=Croatia' if @ARGV; # FIXME debug
 
-$q = 'TS=psychology AND AD=Croatia';
+my $max_cites = 5000; # ISI limit to get cites
+our @ranges;
 
 use WWW::Mechanize;
 use Data::Dump qw(dump);
@@ -26,10 +31,11 @@ rmtree $dir if -e $dir;
 mkdir $dir;
 
 sub save_mech {
-       my ( $mech, $path ) = @_;
+       my $path = shift;
        $step++;
        my $base_path = sprintf('%s/%04d', $dir,$step);
-       $path ||= $base_path . ( $mech->{ct} =~ m{html}i ? '.html' : '.txt' );
+       $path ||= $base_path;
+       $path .= $mech->{ct} =~ m{html}i ? '.html' : '.txt';
        $mech->save_content( $path );
        warn "# [$step] $path ", -s $path, " ", $mech->ct, "\n";
        open(my $dump, '>', "$base_path.dump.txt");
@@ -38,34 +44,44 @@ sub save_mech {
 
 warn "# get session";
 $mech->get( 'http://isiknowledge.com/?DestApp=WOS' );
-save_mech $mech;
+save_mech;
 
 sub search {
        warn "# advanced serach";
        $mech->follow_link( url_regex => qr/AdvancedSearch/ );
-       save_mech $mech;
+       save_mech;
 
        warn "# cookie_jar ", dump $mech->cookie_jar;
 
+       my $q_this = $q;
+
+       if ( @ranges ) {
+               $q_this .= ' AND (' . join(' OR ', map { "PY=$_" } @{ shift @ranges } ) . ')';
+       }
+
+       warn "# submit_form search: $q_this\n";
        $mech->submit_form(
                fields => {
-                       'value(input1)' => $q,
-               }
+                       'value(input1)' => $q_this,
+               },
        );
-       save_mech $mech;
+       save_mech;
 
        warn "# summary";
        $mech->follow_link( url_regex => qr/summary/ );
-       save_mech $mech;
+       save_mech;
 }
 
 sub get_results {
+       my $desc = shift;
        my $from = 1;
 
        while ( 1 ) {
 
                my $to = $from + $range_size;
 
+               warn "# submit_form results $from - $to\n";
+
                $mech->submit_form(
                        form_name => 'summary_output_form',
                        fields => {
@@ -84,63 +100,97 @@ sub get_results {
                        },
                        button => 'save',
                );
-               save_mech $mech;
+               save_mech;
+
 
                if ( $mech->content =~ m{invalid API call} ) {
                        $mech->back;
                        last;
                }
 
-               warn "range $from - $to [$q]\n";
+
+               my $path = "/tmp/isi.$q.$from-$to";
+               $path .= '.' . $desc if $desc;
+
+               warn "save $from - $to into $path\n";
                $mech->follow_link( url_regex => qr/save_file/ );
-               save_mech $mech => "/tmp/isi.$q.$from-$to.txt";
+               save_mech $path;
 
-               $from += $range_size;
+               $from += $range_size - $overlap;
 
                $mech->back;
                $mech->back;
-               #save_mech $mech;
-
+               #save_mech;
        }
-
 }
 
 
 sub citations {
-       save_mech $mech;
        warn "# citation report";
        $mech->follow_link( url_regex => qr/search_mode=CitationReport/ );
-       save_mech $mech;
+       save_mech;
 
        warn "view citing articles";
        $mech->follow_link( url_regex => qr/search_mode=TotalCitingArticles/ );
-       save_mech $mech;
+       save_mech;
 }
 
 sub years {
-       my $years_url = $mech->find_link( text_regex => qr/more options/ );
+       my $years_url = $mech->find_link( url_regex => qr/ra_name=/ );
        if ( ! $years_url ) {
-               warn "W: can't find years\n";
+               warn "W: can't find ra_name link\n";
                return;
        }
        $years_url = $years_url->url_abs;
        warn "## $years_url";
        if ( $years_url !~ s{ra_name=\w+}{ra_name=PublicationYear} ) {
-               warn "W: no ra_name\n";
+               warn "W: no ra_name in $years_url\n";
                return;
        }
        warn "# refine years (hidden by javascript)";
-       warn "http://apps.isiknowledge.com/RAMore.do?product=WOS&search_mode=TotalCitingArticles&SID=T1o6bChdN9PGP1LN1Nh&qid=3&ra_mode=more&ra_name=PublicationYear&db_id=WOS&viewType=raMore\n$years_url\n";
+#      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";
        $mech->get( $years_url );
-       save_mech $mech;
+       save_mech;
 
        my $html = $mech->content;
        my $years;
-       while ( $html =~ s{>(\d\d\d\d)\s\((\d+)\)</label.+?value="PublicationYear_}{} ) {
-               $years->{$1} = $2;
+       while ( $html =~ s{<label.+?PublicationYear.+?>(\d{4})\s\(([\d,]+)\)</label>}{} ) {
+               my ( $year, $count ) = ( $1, $2 );
+               $count =~ s{,}{}g;
+               $years->{$year} = $count;
        }
        warn "# years ",dump $years;
        $mech->back;
+
+       my @y = sort keys %$years;
+
+       my $y = shift @y;
+       my $size = $years->{$y};
+
+       @ranges = ();
+       my $cites_range;
+       $cites_range = [$y] if $y;
+
+       foreach my $y ( @y ) {
+               if ( $size + $years->{$y} > $max_cites ) {
+                       push @ranges, $cites_range;
+                       warn "# cites_range $size years ",dump( $cites_range ),$/;
+
+                       $cites_range = [];
+                       $size = 0;
+               }
+               $size += $years->{$y};
+               push @$cites_range, $y;
+       }
+
+       if ( $cites_range ) {
+               push @ranges, $cites_range;
+               warn "# cites_range $size years ",dump( $cites_range ), " FINAL\n"
+       }
+
+       warn '# ranges ', dump @ranges;
+       @ranges = () if $#ranges == 1; # just take all
+
        return $years;
 }
 
@@ -150,7 +200,14 @@ get_results;
 
 
 citations;
-years;
-$q .= '.citing';
-get_results;
+
+do {
+       my $part;
+       if ( @ranges ) {
+               $part .= $ranges[0]->[0] . '.';
+               search;
+       }
+       $part .= 'citing';
+       get_results $part;
+} while ( @ranges );