Multithread Web Crawler

xuqy has asked for the wisdom of the Perl Monks concerning the following question:

I am interested in "focused crawling" (crawling web pages of some specific topic and ignoring all the others) and have written a "focused crawler" recently. Perl is a reasonable alternative to writing web crawler for its LWP module and CPAN. However, when I planed to implement multithread strategy in crawling, I was confused by perl 5.8.4's "threads" module, especially threads::shared. How is a object reference shared by multiple threads? I want to utilize "Cache::File::Heap" module to sort the urls in "crawling frontier" by heuristic prediction of its "harvest outcome". Below is the relevant code part:

#!/usr/bin/perl -w
use strict;
use threads;
use threads::shared;
use Cache::File::Heap;

my $heap = Cache::File::Heap->new('frontier');
my $heap_lock : shared = 0; 

...

sub go {#crawling thread's control flow
       ....
       #xtracted best promising url
       {
           lock $heap_lock;
           my($value, $url) = $heap->extract_minimum;
       }

       ...
       #after downloading and extract hyperlinks
       {
           lock $heap_lock;
           $heap->add($value, $url) for 
        }
        ...
}

my @threads;
for(1..10) {
    push @threads, threads->new(\&go);
}
for(@threads) {
    $_->join;
}
[download]

All is fine, just untill all the threads joined by main thread and main thread exists. Following error message appears: Scalar leaks : -1 Segmentation fault. My question is : How to share object reference (such as Cache::File::Heap) ?? Cache::File::Heap is the wrapper of BerkeleyDB's BTREE, is BerkeleyDB thread-safe?

Comment on Multithread Web Crawler Download Code

Replies are listed 'Best First'.
Re: Multithread Web Crawler by merlyn (Sage) on Sep 22, 2005 at 00:23 UTC
You might want real processes there, instead of "threads". See any of my "link checker" articles that involve fork. -- Randal L. Schwartz, Perl hacker Be sure to read my standard disclaimer if this is a reply.	[reply]
Re^2: Multithread Web Crawler by aufflick (Deacon) on Sep 22, 2005 at 07:40 UTC
Agreed here. A web crawling application is not going to see benefit from the light-weightedness of multiple threads since it is by it's nature fairly heavy. If you decide that threads don't really hold an advantage for your application you can save yourself a whole load of work by forking off processes. As pointed to in a recent node, Parallel::ForkManager might be of use to you. The module description includes: This module is intended for use in operations that can be done in parallel where the number of processes to be forked off should be limited. Typical use is a downloader which will be retrieving hundreds/thousands of files. Sounds right up your tree? Or is that down your tree? (I never did work out where the roots for a red-black tree would go).	[reply]
Re^3: Multithread Web Crawler by xuqy (Initiate) on Sep 23, 2005 at 13:40 UTC
Thank you so much. I do tried Parallel::ForkManager but come up with a puzzle: How to share data between processes? To avoid crawling the same page repeatly, a global tied hash has to be shared by all the crawling processes. I experimented and found that all the forked processes just ended up with the same crawling history. Can you do me a favor to suggest a patch to it?	[reply]