[BACK]Return to crawler CVS log [TXT] [DIR] Up to [Coherent Logic Development] / pandia

Diff for /pandia/crawler between versions 1.2 and 1.7

version 1.2, 2025/06/25 19:38:48 version 1.7, 2025/07/02 15:03:05
Line 9 Line 9
 # Licensed AGPL-3.0  # Licensed AGPL-3.0
 #  #
 # $Log$  # $Log$
   # Revision 1.7  2025/07/02 15:03:05  snw
   # Add support for restricted mode
   #
   # Revision 1.6  2025/07/01 19:20:47  snw
   # Stop crawling and indexing URLs containing page fragments
   #
   # Revision 1.5  2025/06/28 00:33:32  snw
   # Update locking
   #
   # Revision 1.4  2025/06/27 16:20:30  snw
   # Add blacklist
   #
   # Revision 1.3  2025/06/27 02:14:47  snw
   # Initial operational capability
   #
 # Revision 1.2  2025/06/25 19:38:48  snw  # Revision 1.2  2025/06/25 19:38:48  snw
 # Add indexer  # Add indexer
 #  #
Line 29  use HTML::TreeBuilder; Line 44  use HTML::TreeBuilder;
 use URI;  use URI;
 use DBI;  use DBI;
 use WWW::RobotRules;  use WWW::RobotRules;
 my $rules = WWW::RobotRules->new('pandia-crawler/0.0.1');  use Fcntl qw(:flock);
 use LWP::Simple qw(get);  use LWP::Simple qw(get);
   use Config::IniFiles;
   
   my $rules = WWW::RobotRules->new('pandia-crawler/0.0.1');
 my $dbh = "";  my $dbh = "";
 my $dsn = "";  my $dsn = "";
 my $skips = 0;  my $skips = 0;
Line 41  my $depth = 0; Line 58  my $depth = 0;
 my $blacklist_matches = 0;  my $blacklist_matches = 0;
 my $robots_txt_denies = 0;  my $robots_txt_denies = 0;
 my $invalid_scheme_skips = 0;  my $invalid_scheme_skips = 0;
   my $mode;
   
 sub store_url {  sub store_url {
     my ($url, $parent) = @_;      my ($url, $parent) = @_;
   
     if($url ne "" && length($url) <= 255 && substr($url, 0, 6) ne "mailto") {      if (index($url, '#') != -1) {
           print "F";
           return;
       }
       
       if($url ne "" && length($url) <= 255 && substr($url, 0, 6) ne "mailto" && substr($url, 0, 4) eq "http") {       
           
         my $u = URI->new($url);          my $u = URI->new($url);
         my $domain = $u->host;          my $domain = $u->host;  
         my $scheme = $u->scheme;          my $scheme = $u->scheme;
           my @parts = split($domain, '.');
           my $tld = $parts[-1];
           
           if ($mode eq 'restricted') {
               my $tld_ok = 0;
               foreach (@allowed_tlds) {
                   my $allowed = $_;
                   
                   if($tld eq $allowed) {
                       $tld_ok = 1;
                       last;
                   }
               }
               if($tld_ok == 0) {
                   print "T";
                   return;
               }
           }
   
         my $sth = $dbh->prepare("INSERT INTO url_domains (url_domain) VALUES (?)");          my $sth = $dbh->prepare("INSERT INTO url_domains (url_domain) VALUES (?)");
         $sth->execute($domain);          $sth->execute($domain);
Line 114  sub crawl_url { Line 155  sub crawl_url {
     my $tree = HTML::TreeBuilder->new();      my $tree = HTML::TreeBuilder->new();
   
     my $response = $http->get($url);      my $response = $http->get($url);
   
   
     $tree->parse($response->{content});      $tree->parse($response->{content});
   
     my @links = $tree->find_by_tag_name('a');      my @links = $tree->find_by_tag_name('a');
Line 147  sub crawl_url { Line 190  sub crawl_url {
 }  }
   
 $| = 1;  $| = 1;
   
 print "pandia crawler v0.0.1\n";  print "pandia crawler v0.0.1\n";
 print " Copyright (C) 2025 Coherent Logic Development LLC\n\n";  print " Copyright (C) 2025 Coherent Logic Development LLC\n\n";
   
 GetOptions("dbhost=s" => \$dbhost,  my $profile;
            "dbname=s" => \$dbname,  
            "dbusername=s" => \$dbusername,  GetOptions("profile=s" => \$profile,
            "dbpw=s" => \$dbpw,             "seed=s" => \$seed,
            "seed=s" => \$seed,             "maxdepth=n" =>\$maxdepth)
            "maxdepth=n" =>\$maxdepth)  
     or die("error in command line arguments");      or die("error in command line arguments");
   
   my $cfg = Config::IniFiles->new(-file => "/etc/pandia.ini");
   
   $dbhost = $cfg->val($profile, 'dbhost');
   $dbname = $cfg->val($profile, 'dbname');
   $dbusername = $cfg->val($profile, 'dbuser');
   $dbpw = $cfg->val($profile, 'dbpass');
   $tmp = $cfg->val($profile, 'allowed_tlds');
   
   if($tmp ne '*') {
       $mode = 'restrictive';
       @allowed_tlds = split(',', $tmp);
       print "pandia:  crawler restricted to these TLDs:  ";
       foreach (@allowed_tlds) {
           print ".$_ ";
       }
       print "\n";
   }
   else {
       print "pandia:  crawler unrestricted\n";
       $mode = 'normal';
   }
   
 print "pandia:  connecting to $dbname database at $dbhost...";  print "pandia:  connecting to $dbname database at $dbhost...";
   
 $dsn = "DBI:mysql:database=$dbname;host=$dbhost;port=3306;mysql_connect_timeout=5;";  $dsn = "DBI:mysql:database=$dbname;host=$dbhost;port=3306;mysql_connect_timeout=5;";
Line 167  die "pandia:  failed to connect to MySQL Line 230  die "pandia:  failed to connect to MySQL
   
 print "[OK]\n";  print "[OK]\n";
   
 if($seed ne "") {  print "pandia:  each character represents the following status for a URL:\n";
     print "pandia:  crawling seed $seed to a maximum depth of $maxdepth";  print "  .    URL added to indexer queue\n";
   print "  l    crawl exceeded max depth\n";
   print "  x    URL too long or invalid scheme\n";
   print "  d    URL was a duplicate\n";
   print "  b    crawl was blocked by robots.txt\n";
   print "  F    URL contained a fragment\n";
   print "  T    URL was from a disallowed top-level domain\n\n";
   
   if($seed ne "") {        
       print "pandia:  crawling seed $seed to a maximum depth of $maxdepth...";
     sleep 1;      sleep 1;
     crawl_url($seed);      crawl_url($seed);
     print "[OK]\n";      print "[OK]\n";
 }  }
 else {  else {
       open my $file, ">", "/tmp/pandia_crawler.lock" or die $!; 
       flock $file, LOCK_EX|LOCK_NB or die "Unable to lock file $!";
   
     my $sth = $dbh->prepare("SELECT url FROM crawl_queue");      my $sth = $dbh->prepare("SELECT url FROM crawl_queue");
     $sth->execute();      $sth->execute();
     my $qlen = $sth->rows;      my $qlen = $sth->rows;
Removed from v.1.2  
changed lines
  Added in v.1.7

FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>