Re^2: Hashes aren't being differently randomized

And yet, it seems very unlikely that effectively changing the hash function would result in identical output order. The output order is based on the hashed key values modulo the number of buckets, so changing the hash function is very likely to change the output order.

So test the effectiveness of the randomization directly: See which hash keys collide. If the same hash keys collide between runs, then the randomization isn't doing what it is supposed to:

my %hash;
my $key= 'aa';
my $count= @ARGV ? shift(@ARGV) : 20;
$hash{$key}= 'base';
while( 1 ) {
    $hash{++$key}= 'test';
    if( %hash =~ /^1/ ) {
        print "Collision: $key\n";
        exit   if  --$count <= 0;
    }
    delete $hash{$key};
}
[download]

I always get the same output, but this PERL_HASH_SEED_DEBUG environment variable doesn't work for me so I won't jump to any conclusions (my 5.008_007 is new enough that it should include it, however).

- tye

Comment on Re^2: Hashes aren't being differently randomized Download Code

Replies are listed 'Best First'.
Re^3: Hashes aren't being differently randomized by Hue-Bond (Priest) on Jun 27, 2006 at 14:41 UTC
I always get the same output Same here: `$ PERL_HASH_SEED_DEBUG=1 perl ./foo \| md5sum HASH_SEED = 4155558892 da7b0b6c9adb2fbc01be283dc16aed19 - $ PERL_HASH_SEED_DEBUG=1 perl ./foo \| md5sum HASH_SEED = 51798468 da7b0b6c9adb2fbc01be283dc16aed19 - $ PERL_HASH_SEED_DEBUG=1 perl ./foo \| md5sum HASH_SEED = 2930542944 da7b0b6c9adb2fbc01be283dc16aed19 -` [download] -- David Serrano	[reply] [d/l]
Re^4: Hashes aren't being differently randomized (colliders) by tye (Sage) on Jun 27, 2006 at 22:07 UTC
Ever wonder exactly which keys in your particular hash are sharing a bucket? Here is a destructive tool that will tell you that: my %hash= 0..299; for my $av ( getKeyCollisions( \%hash ) ) { print "@$av\n"; } sub getKeyCollisions { my( $hv )= @_; no warnings 'numeric'; my $buckets= 0 + %$hv; my @keys= keys %$hv; my( @clash, @return ); while( @keys ) { my $key= shift @keys; delete $hv->{$key}; my $b= 0 + %$hv; if( $b == $buckets ) { push @clash, $key; } elsif( @clash ) { push @return, [@clash,$key]; @clash= (); } $buckets= $b; } return @return; } __END__ 276 90 206 118 272 190 298 194 226 58 284 86 76 82 110 280 228 268 218 202 168 184 14 178 24 104 124 256 216 26 80 72 162 74 240 246 108 230 64 210 68 188 116 88 136 30 144 286 120 28 40 134 156 192 250 22 42 158 94 248 296 34 132 164 [download] - tye	[reply] [d/l]
Re^3: Hashes aren't being differently randomized (-DNO_HASH_SEED) by tye (Sage) on Jun 27, 2006 at 21:35 UTC
"perl -V" reports " ... `ccflags ='` ... `-DNO_HASH_SEED` ... ", which likely explains why /I/ always get the same results. My Perl was built by Randy Kobes. I'm a bit curious why that setting was chosen. - tye	[reply] [d/l] [select]


Just another Perl shrine
	PerlMonks