use warnings;
use strict;
-my $q = 'AD=Croatia';
+use WWW::Mechanize;
+use Data::Dump qw(dump);
+use File::Path;
+use Text::Unaccent;
+
+# Advanced search syntax:
+# 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 $dump = @ARGV ? 1 : 0;
+my $results = 0;
+my $citations = 0;
+my $cited_reference = 0; # html tables
+my $citing_articles = 1; # as many files as cited articles
-$q = 'TS=psychology AND AD=Croatia';
+my $cites_by_year = 0;
-use WWW::Mechanize;
-use Data::Dump qw/dump/;
+my $max_cites = 5000; # ISI limit to get cites
+
+$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,
);
our $step = 0;
+our @ranges;
+
+my $dir = '/tmp/isi/';
+#rmtree $dir if -e $dir;
+mkdir $dir unless -d $dir;
sub save_mech {
- my ( $mech, $path ) = @_;
+ my $path = shift;
$step++;
- mkdir '/tmp/isi/' unless -e '/tmp/isi';
- my $base_path = sprintf('/tmp/isi/%04d', $step);
- $path ||= $base_path . ( $mech->{ct} =~ m{html}i ? '.html' : '.txt' );
+ my $base_path = sprintf('%s/%04d', $dir,$step);
+ $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");
warn "# get session";
$mech->get( 'http://isiknowledge.com/?DestApp=WOS' );
-save_mech $mech;
+save_mech;
sub search {
- my $q = shift;
-
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 $q = shift;
+ 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 => {
},
button => 'save',
);
- save_mech $mech;
+ save_mech;
+
if ( $mech->content =~ m{invalid API call} ) {
$mech->back;
last;
}
- warn "range $from - $to [$q]\n";
+ 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;
+
+ 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/ )->url_abs;
+ my $years_url = $mech->find_link( url_regex => qr/ra_name=/ );
+ if ( ! $years_url ) {
+ warn "W: can't find ra_name link\n";
+ return;
+ }
+ $years_url = $years_url->url_abs;
warn "## $years_url";
- $years_url =~ s{ra_name=\w+}{ra_name=PublicationYear} || die "ra_name";
+ if ( $years_url !~ s{ra_name=\w+}{ra_name=PublicationYear} ) {
+ 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_}{} ) {
- push @years, [ $1 => $2 ];
+ my $years;
+ 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;
+ warn "# years ",dump $years;
$mech->back;
- return @years;
+
+ my @y = sort keys %$years;
+
+ @ranges = ();
+
+ if ( $cites_by_year ) {
+ push @ranges, [ $_ ] foreach @y;
+ warn "# cites_by_year ranges ", dump @ranges;
+ return;
+ }
+
+ my $y = shift @y;
+ my $size = $years->{$y};
+
+ 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;
+}
+
+
+our $page = 1;
+sub next_page {
+ $page++;
+ warn "next_page $page\n";
+
+ $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;
+ }
+
+ }
+
}
-search $q;
-years;
-get_results $q;
+if ( $q =~ m{CA=(.+)} && $citing_articles ) {
-citations;
-years;
-get_results $q . '.citing';
+ 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";
projects / webpac2 / blobdiff