use re 'eval';
my $str = "111111111";
my $times = 5;
my $re = qr/1/;
my $cnt;
$str =~ s/$re(?(?{++$cnt > $times})\A)/2/g;
print "$str\n";
[download]

IMO its a bad habit to use lexicals in such constructs. They will only work once. For instance your code will not work properly if naively converted to a subroutine. The $cnt var needs to be a package scoped var. Otherwise what happens is that the regex is compiled once and the first $cnt will be enclosed into the (?{}), on the second run the new $cnt is not used, rather the original will be used. So for instance

use re 'eval';
sub limited_re {
    my $str=shift;
    my $re=shift;
    my $repl=shift;
    my $times=shift;

    our $cnt;
    local $cnt=0;
    $str =~ s/$re(?(?{++$cnt > $times})\A)/eval $repl/ge;
    print "$str\n";
}

limited_re("111111111",qr/1/,5,2);
limited_re("000000000111111111",qr/1/,2,5);
[download]

Works as expected. Change the our $cnt; local $cnt=0; to a my $cnt; and it won't. I got bitten by this when working on a solution for QOTW 23 (which happens to be on my pad at the time of posting this.)



---
demerphq


First they ignore you, then they laugh at you, then they fight you, then you win.
-- Gandhi

Thanks, I did not realize the dangers of lexicals in this situation. Though in order to get the capture variable behavior of the OP's code, I made the following changes:

...
    $str =~ s/$re(?(?{++$cnt > $times})\A)/qq["$repl"]/gee;
...

limited_re("111111111",qr/(1)/,'${1}5',2);
[download]

Though it would have worked with the code as it was if you included double quotes inside the single quotes of the replacement string, e.g., '"${1}5"'.