| 1 | #!/usr/bin/env perl |
| 2 | use strict; |
| 3 | use warnings; |
| 4 | use threads; |
| 5 | #use Thread::Queue; |
| 6 | #use threads::shared; |
| 7 | use LWP::Simple qw($ua get); $ua->timeout(3); |
| 8 | use BerkeleyDB; |
| 9 | |
| 10 | print STDERR "This is libwww-perl-$LWP::VERSION\n"; |
| 11 | |
| 12 | #my @queue :shared; |
| 13 | #my %crawled :shared; |
| 14 | tie my @queue, "BerkeleyDB::Recno", -Filename => 'queue.db', -Flags => DB_CREATE || die "Cannot open DB1!\n"; |
| 15 | tie my %crawled, "BerkeleyDB::Hash", -Filename => 'urls.db', -Flags => DB_CREATE || die "Cannot open DB2!\n"; |
| 16 | |
| 17 | push(@queue, 'http://root.cz/'); |
| 18 | push(@queue, 'http://blog.harvie.cz/'); |
| 19 | |
| 20 | untie @queue; |
| 21 | untie %crawled; |
| 22 | |
| 23 | |
| 24 | sub uniq(\@) { |
| 25 | my ($i) = @_; |
| 26 | my %h; |
| 27 | @$i = grep(!$h{$_}++, @$i); |
| 28 | return(@$i); |
| 29 | } |
| 30 | |
| 31 | my $threads = 3; |
| 32 | for(my $i=0; $i<$threads; $i++) { |
| 33 | my $thr = threads->create(sub { |
| 34 | print("[*] Worker #$i running!\n"); |
| 35 | |
| 36 | my $env = new BerkeleyDB::Env |
| 37 | -Home => "/tmp/" , |
| 38 | -Flags => DB_CREATE| DB_INIT_CDB | DB_INIT_MPOOL |
| 39 | or die "cannot open environment: $BerkeleyDB::Error\n"; |
| 40 | |
| 41 | |
| 42 | tie my @queue, "BerkeleyDB::Recno", -Filename => 'queue.db', -Flags => DB_CREATE, -Env => $env || die "Cannot open DB1!\n"; |
| 43 | tie my %crawled, "BerkeleyDB::Hash", -Filename => 'urls.db', -Flags => DB_CREATE, -Env => $env || die "Cannot open DB2!\n"; |
| 44 | |
| 45 | while (1) { |
| 46 | print "omg ".shift(@queue)."\n"; |
| 47 | if(length($_) < 3) { sleep(1); next; } |
| 48 | |
| 49 | print "#$i:\tGET $_\n"; |
| 50 | $_ = LWP::Simple::get($_); |
| 51 | |
| 52 | my @urls; |
| 53 | if($_) { |
| 54 | @urls = /(http:\/\/[_a-zA-Z0-9\.\-]+\.[a-zA-Z]{2,4}\/{1}[-_~&=\ ?\.a-z0-9\/]*)/g; #urls |
| 55 | #@urls = /(http:\/\/[^\/'" ]*)/g; #domains |
| 56 | if($i && @queue < 1000) { |
| 57 | #print("#$i:\tENQ: @urls\n"); |
| 58 | push(@queue, uniq(@urls)); |
| 59 | #while(uniq(@urls)) { push(@queue, $_); } |
| 60 | } |
| 61 | } |
| 62 | } |
| 63 | |
| 64 | print("[*] Worker #$i stopped!\n"); |
| 65 | }) -> detach(); |
| 66 | } |
| 67 | |
| 68 | while(1) { |
| 69 | print '###: '.@queue."\n"; |
| 70 | uniq(@queue); |
| 71 | print '###: '.@queue."\n"; |
| 72 | |
| 73 | sleep(5); |
| 74 | } |