Re^2: Quantitative Change instead of Boolean

[titivillus]

I've taken a look at Digest::Nilsimsa. It is a cool thing. However, I've found a problem.

for my $d ( 30 .. 36 ) {
  my $this = $nil->text2digest( ( 'a' x $d ) . 'b' ) ;
  my $that = $nil->text2digest( ( 'a' x ( $d + 1 ) ) ) ;
  print nilcomp( $this , $that ) ;
  }

sub nilcomp {
  my $diff = 0 ;
  my $diff2 = 0 ;
  my @this = split /|/ , shift ;
  my @that = split /|/ , shift ;
  for my $a ( 0 .. scalar(@this)-1 ) {
    $diff++ if $this[$a] ne $that[$a] ;
    my $is = hex $this[$a] ;
    my $at = hex $that[$a] ;
    if ( $is != $at ) {
      $diff2 += abs $is - $at ;
      }
    }
  return ( join "" , @this) . qq(\n) .
         ( join "" , @that) . qq(\n) .
         $diff . qq( characters different\n) .
         ( abs $diff2 ) . qq( bits different\n\n);
  }

gives you

000000000000900000010021000008105000080010000004000c400000000008
0000000000009000000000200000080040000000000000040008400000000000
8 characters different
25 bits different

000000000000900000010021000008105000080010000004000c400000000008
0000000000009000000000200000080040000000000000040008400000000000
8 characters different
25 bits different

000000000000900000010021000008105000080010000004000c400000000008
0000000000009000000000200000080040000000000000040008400000000000
8 characters different
25 bits different

000000000000900000010021000008105000080010000004000c400000000008
0000000000009000000000200000080040000000000000040008400000000000
8 characters different
25 bits different

000000000000900000010021000008105000080010000004000c400000000008
0000000000009000000000200000080040000000000000040008400000000000
8 characters different
25 bits different

0000000000009000000000200000080040000000000000040008400000000000
0000000000009000000000200000080040000000000000040008400000000000
0 characters different
0 bits different

0000000000009000000000200000080040000000000000040008400000000000
0000000000009000000000200000080040000000000000040008400000000000
0 characters different
0 bits different

If all I was trying to use this on were 35-character data sets, that'd be cool, but I'm trying to run this on whole web pages. I pull out all markup and whitespace, I'll still be in the headers by the time the 35th character rolled around. I love it in theory, but in practice, the data's too big for the module. So, I could do this:  $output =~ s[(.{35})][$nilsimsa->text2digest($1)]ge ; or something of the sort, but that seems ... goofy. But it does point out that taking length $output and comparing it to last time should indicate a small change, if they're only a small number of characters apart.