Updated To correct error introduced whilst sanitising for public consumption.

It would be interesting to see how this would compare with merlyn's Inline::C version for performance. Unfortunately, I still don't have that ability (yet).

One point to note, this version should also work for unicode strings. If you don't need that, then swapping the unpack template from 'U*' to 'C*' should do the trick and might be a tad quicker.

#! perl -slw
use strict;

sub nUniqC{
      my @uniq; ## Updated. 
    @uniq[unpack 'U*',$_[0]] = (1)x length $_[0];
    scalar (grep defined $_, @uniq);
}

print nUniqC '1010101010';
print nUniqC '1234567812';
[download]

For speed in perl and assuming no utf-8, this seems a bit quicker.

sub nUniqC{
    my @uniq;
    scalar grep{ ++$uniq[$_] == 1 } unpack('C*',$_[0]);
}
[download]

Update Trimmed another 15% of fat. No I didn't. I misread bad data, the version above is quickest.

sub nUniqC{
    scalar grep{ ++$_[$_] == 1 } unpack('C*',$_[0]);
}
[download]