Re^3: Determining uniqueness in a string.

[creamygoodness]

Precompiling regexes can make a dramatic difference. Compare the stats below for u1 vs u5 -- there's an 868% speedup on my box!

Nevertheless, the backreference is just a better algo, and it still wins.

use Benchmark qw( cmpthese ) ;

my $digits = 2491306578;
my @regexes = map { qr/$_/ } 0 .. 9; 

cmpthese( -5, {
    u1 => sub{
               for ( 0 .. 9 ) { 
                  return 0 if $digits !~ /$_/;
               } 
               return 1;
             },      
        
    u2 => sub{
                my @x = sort split //, $digits;
                my $x = join "", @x; 
                $x =~ tr/0-9//s;
                length $x == 10 ? return 1 : return 0;
              },      
         
    u3 => sub {
                return 0 if $digits =~ /(\d).*\1/; return 1;
              },      
         
    u4 => sub {
                return $digits !~ /(.).*\1/;
              },      
    u5 => sub {
                for (@regexes) { 
                     return 0 if $digits !~ $_; 
                } 
                return 1;
              }

       Rate    u1    u2    u5    u3    u4
u1  10555/s    --  -71%  -90%  -93%  -93%
u2  35955/s  241%    --  -65%  -76%  -76%
u5 102215/s  868%  184%    --  -32%  -32%
u3 149240/s 1314%  315%   46%    --   -1%
u4 150091/s 1322%  317%   47%    1%    --

[QM]

You're missing the closing }) (perhaps a copy/paste error?)

I also don't care for the assumption that it's exactly 10 characters when it gets to uniq. If that requirement changes, uniq becomes broken without being modified. While this is perhaps simplistic, the fact that the code is separated means it's easy to update one without seeing the other, and it quietly fails. uniq would be better named pannumeric.