Parent Directory
Bloom-Faster-1.3.1.meta 16-Mar-2007 19:08:05 312
Bloom-Faster-1.3.1.readme 09-Mar-2007 23:16:51 2.5K
Bloom-Faster-1.3.1.tar.gz 17-Mar-2007 06:25:34 8.5M
Bloom-Faster-1.3.meta 23-Feb-2007 05:48:37 310
Bloom-Faster-1.3.readme 23-Feb-2007 05:45:36 2.5K
Bloom-Faster-1.3.tar.gz 23-Feb-2007 05:51:31 476.8K
Bloom-Faster-1.4.meta 17-Mar-2007 07:44:29 310
Bloom-Faster-1.4.readme 09-Mar-2007 23:16:51 2.5K
Bloom-Faster-1.4.tar.gz 17-Mar-2007 07:48:00 602.0K
CHECKSUMS 17-Mar-2008 14:26:14 2.1K
README 23-Feb-2007 17:54:51 2.5K
README
NAME Bloom::Faster - Perl extension for the c library libbloom. INSTALLATION see INSTALL SYNOPSIS use Bloom::Faster; # m = ideal vector size. # k = # of hash functions to use. my $bloom = new Bloom::Faster({m => 1000000,k => 5}); # this gives us very tight control of memory usage (a function of m) # and performance (a function of k). but in most applications, we won't # know the optimal values of either of these. for these cases, it is # much easier to supply: # # n = number of expected elements to check for duplicates, # e = acceptable error rate (probability of false positive) # # my $bloom = new Bloom::Faster({n => 1000000, e => 0.00001}); while (<>) { chomp; # Bloom::Faster->add() returns true when the value is a duplicate. if ($bloom->add($_)) { print "DUP: $_\n"; } } DESCRIPTION Bloom filters are a lightweight duplicate detection algorithm proposed by Burton Bloom (http://portal.acm.org/citation.cfm?id=362692&dl=ACM&coll=portal), with applications in stream data processing, among others. Bloom filters are a very cool thing. Where occasional false positives are acceptable, bloom filters give us the ability to detect duplicates in a fast and resource-friendly manner. The allocation of memory for the bit vector is handled in the c layer, but perl's oo capability handles the garbage collection. when a Bloom::Faster object goes out of scope, the vector pointed to by the c structure will be free()d. to manually do this, the DESTROY builtin method can be called. A bloom filter perl module is currently avaible on CPAN, but it is profoundly slow and cannot handle large vectors. This alternative uses a more efficient c library which can handle arbitrarily large vectors (up to the maximum size of a "long long" datatype (at least 9223372036854775807, on supported systems ). EXPORT None by default. Exportable constants HASHCNT PRIME_SIZ SIZ SEE ALSO libbbloom.so AUTHOR Peter Alvaro and Dmitriy Ryaboy,