handle new genecards site

[function2gene.git] / bin / parse_harvester_results

diff --git a/bin/parse_harvester_results b/bin/parse_harvester_results
index 29d07194a8a4e7c830fe7af55bdaa4d9e873a15d..e52066127ceab16107b04bfc68c919ee150adb86 100755 (executable)
--- a/bin/parse_harvester_results
+++ b/bin/parse_harvester_results
@@ -77,9 +77,12 @@ my %options = (debug    => 0,
               man      => 0,
               dir      => '.',
               keyword  => undef,
+              keywords => 0,
              );
 
-GetOptions(\%options,'keyword|k=s','dir|D=s','debug|d+','help|h|?','man|m');
+GetOptions(\%options,'keyword|k=s','dir|D=s','debug|d+','help|h|?','man|m',
+          'keywords',
+         );
 
 
 pod2usage() if $options{help};
@@ -99,6 +102,15 @@ use constant {NAME        => 0,
              FILENAME    => 8,
             };
 
+if ($options{keywords}) {
+     if (@ARGV != 1) {
+         pod2usage("If the --keywords option is used, exactly one argument (the keyword) must be passed");
+     }
+     $options{dir} = "$ARGV[0]_results_harvester";
+}
+
+
+
 if (not -d $options{dir}) {
      die "$options{dir} does not exist or is not a directory";
 }
@@ -130,8 +142,16 @@ while ($_ = $dir->read) {
          ($results[NAME]) = $result =~ m&<TR>\s*<TD\s*BGCOLOR="\#FEFE99"\s*VALIGN="top"\s*NOWRAP>Entry\s*name</TD>\s*
                                          <TD\s*VALIGN="top"\s*COLSPAN="5">\s*<b>\s*([^<]+?)\s*</b></TD>\s*</TR>&xis;
      }
+     if (not defined $results[NAME]) {
+         ($results[NAME]) = $result =~ m{<TITLE>[^:]+:\s*[^\*]+\*[^\*]+\*\s*([^-]+)}xis;
+         $results[NAME] =~ s/\s*$// if defined $results[NAME];
+         $results[NAME] =~ s/^\s*$// if defined $results[NAME];
+         $results[NAME] =~ s/\d+\s*kDa\s*protein// if defined $results[NAME];
+         $results[NAME] =~ s/\s*similar to .+// if defined $results[NAME];
+     }
 
      $results[NAME] ||= 'NO NAME';
+     $results[NAME] =~ s/_HUMAN//;
 
      # Find REF SEQ number
      ($results[REFSEQ]) = $result =~ m&<a\s+href="http://www.ncbi.nlm.nih.gov/entrez/