Last weekend, diotalevi started an interesting discussion about regexp reentrancy, and everyone basically agreed that you cannot reenter regexp.

Now my curiosity drives me to think, what if I re-enter regexp on a new thread? Is regexp engine per thread? So I designed this set of three experiments. It was first designed with only two experiments:
  1. regexp reentrant on same thread;
  2. regexp reentrant on different threads.
  3. However, after I finished the first two parts, I realized the need to add a third one, which is a modified version of the second one, but without reentrant.
Now the details:
  1. First experiment
    Expected result: fail.
    Actual result: fail.
    Observation: No surprise, it cored. However before it cored, I already have seen all the printing, and they are correct (is this a surprise?)
    (As you can see there is an assignment, my $tmp = $2. It did not make any difference here, but it did make a difference in experiment two and three. I added it here just to make it fair for all three experiments.) 
    use strict;
use warnings;

$| ++;

sub parser {
    my $str = shift;
    $str =~ m/^\[(\d*),(.*?)\]$(?{print "1 = $1\n2 = $2\n";my $tmp = $
+2;parser($tmp) if ($tmp ne "[]")})/;
}

my $str = "[0,[1,[2,[3,[4,[5,[6,[7,[8,[9,[]]]]]]]]]]]";
parser $str;

    [download]
  2. Second experiment
    Expected result: to be frank, I didn't dare to guess.
    Actual result: fail.
    Observation: Surprise? yes and no. Just as experienment one, it cored, again before core, I have seen all printing already.
    If I remove that my $tmp = $2, lots of other problem are reported. Last time, we said that the reason regexp engine cannot reentered, is that regexp uses a set of global variables, I believe $1, $2 are among those. 
    use threads;

use strict;
use warnings;

$| ++;

#open(STDERR, ">", "error.txt");

sub parser {
    my $str = shift;
    $str =~ m/^\[(\d*),(.*?)\]$(??{print "1 = $1\n2 = $2\n";my $tmp = 
+$2;if ($tmp ne "[]") {my $t = threads->create("parser",$tmp); $t->joi
+n()}})/;}

my $str = "[0,[1,[2,[3,[4,[5,[6,[7,[8,[9,[]]]]]]]]]]]";
parser $str;

    [download]
  3. Third experiment
    Expected result: pass.
    Actual result: pass.
    Observation: But it came to me as a surprise that I have to do that $tmp = $2, otherwise it cores.

Replies are listed 'Best First'.
Re: revisit regexp reentrancy
by pg (Canon) on Dec 14, 2002 at 04:20 UTC
    Sorry, did it too quick, and didn't copy and paste the code for part 3, but I cannot update the original post... here comes it: 
    use threads;

use strict;
use warnings;

$| ++;

open(STDERR, ">", "error.txt");

sub parser {
    my $str = shift;
    $str =~ m/^\[(\d*),(.*?)\]$/;
    print "1 = $1\n2 = $2\n";
    my $tmp = $2;
    if ($2 ne "[]") {
        my $t = threads->create("parser",$2);
        $t->join()
    }
}

my $str = "[0,[1,[2,[3,[4,[5,[6,[7,[8,[9,[]]]]]]]]]]]";
parser $str;

    [download]
[reply]
[d/l]

Back to Meditations