my $nth = 3;
my $rep = 1;
my $chr = 'e';
my $str = "Terence and Philip are sweet\n";

my $pos = 0;
$pos = index($str, $chr, $pos+1) for(1..$nth);
substr($str, $pos) =~ s/$chr/$rep/g;

print $str;

##</code><code>##

my $s = "Terence and Philip are sweet\n";
my $c = 'e';
my $r = 1;
my $n = 3;

sub mine {
  my ($str, $chr, $rep, $nth) = @_;
  my $pos = 0;
  $pos = index($str, $chr, $pos+1) for(1..$nth);
  substr($str, $pos) =~ s/$chr/$rep/g;
  return $str;
}

sub quents {
  my ($str, $chr, $rep, $nth) = @_;
  my $i = 1;
  $str =~ s/($chr)/$i++<$nth?$1:$rep/eg;
  return $str;
}

use Benchmark qw( timethese cmpthese );

cmpthese(500000, {
  "Mine"    => sub { mine   ($s, $c, $r, $n) },
  "Quent's" => sub { quents ($s, $c, $r, $n) },
});

__END__
Benchmark: timing 500000 iterations of Mine, Quent's...
      Mine: 14 wallclock secs (14.08 usr +  0.00 sys = 14.08 CPU) @ 35511.36/s (n=500000)
   Quent's: 19 wallclock secs (18.58 usr +  0.03 sys = 18.61 CPU) @ 26867.28/s (n=500000)
           Rate Quent's    Mine
Quent's 26867/s      --    -24%
Mine    35511/s     32%      --