in reply to Re^7: Parser Performance Question (updated)
in thread Parser Performance Question

Hu, how did I ever reach sainthood in our book? Eily => qr/ " (?: [^"\\] | \\. )* " /x); <= added the missing " in the character class

use Data::Dump qw( pp );
my $str = <<STR;
 " .. \\" .. 
STR

print $str;

my %re = (LanX => qr/ " (?: \\\\ | \\" | [^"] )* " /x,
          Eily => qr/ " (?: [^"\\] | \\. )* " /x);

$str =~ /$re{$_}/ and say "$_ found $&" or say "$_ found nothing" for 
+keys %re;

[download]
 " .. \" ..
Eily found nothing
LanX found " .. \"

[download]

Replies are listed 'Best First'.
Re^9: Parser Performance Question (Atomic grouping)
by LanX (Saint) on Oct 06, 2017 at 11:28 UTC
    Atomic grouping (?>...) seems to fix the problem 
    use strict;
use warnings;
use feature 'say';

use Data::Dump qw( pp );
my @strs =
  qw(
       "..\\"..
       "abc"
       "a\"bc"
       "a\\\\bc"
       "a\"
   );



my %re =
  (
   LanX => qr/ " (?> \\\\ | \\" | [^"] )* " /x,
   Eily => qr/ " (?: [^"\\] | \\. )* " /x

  );


for my $str (@strs) {
   say "\nTesting: <$str> = ", pp ($str);
   $str =~ /$re{$_}/ and say "$_ found $&" or say "$_ found nothing"
     for keys %re;
}

    [download]

    Testing: <"..\"..> = "\"..\\\".."
LanX found nothing
Eily found nothing

Testing: <"abc"> = "\"abc\""
LanX found "abc"
Eily found "abc"

Testing: <"a\"bc"> = "\"a\\\"bc\""
LanX found "a\"bc"
Eily found "a\"bc"

Testing: <"a\\bc"> = "\"a\\\\bc\""
LanX found "a\\bc"
Eily found "a\\bc"

Testing: <"a\"> = "\"a\\\""
LanX found nothing
Eily found nothing

    [download]

    Cheers Rolf
    (addicted to the Perl Programming Language and ☆☆☆☆ :)
    Je suis Charlie!

[reply]
[d/l]
[select]

      I had completly forgotten about the atomic match :). I reread the documentation (because I wasn't sure what it does exactly), and it shows that it is equivalent to the possessive quantifiers. So in the spirit of TIMTOWTDI: 

      use strict;
use warnings;
use feature 'say';

use Data::Dump qw( pp );
my @strs =
  qw(
       "..\\"..
       "abc"
       "a\"bc"
       "a\\\\bc"
       "a\"
   );



my %re =
  (
   LanX => qr/ " (?> \\\\ | \\" | [^"] )* " /x,
   Eily => qr/ " (?: [^"\\] | \\. )* " /x,
   Poss => qr/ " (?: \\\\ | \\" | [^"] )*+ " /x,

  );


for my $str (@strs) {
   say "\nTesting: <$str> = ", pp ($str);
   $str =~ /$re{$_}/ and say "$_ found $&" or say "$_ found nothing"
     for keys %re;
}

      [download]

[reply]
[d/l]
        Great, discussion we (re)learned a lot! :)

        edit

        I saw atomic grouping discussed in Friedl's Book, but this use example is very instructive.

        And inhibiting backtracking has a great performance benefit, think I have to revisist some older projects of mine again.

        Cheers Rolf
        (addicted to the Perl Programming Language and ☆☆☆☆ :)
        Je suis Charlie!

[reply]

        I appreciate reading the conversation you guys had, sorry I wasn't to able to take part. I'm now using a slightly modified version of Eily's regex (proven using the above framework and in my own tests):

        our $RXdqs =   qr/ " (?> \\. | [^"\\] )* " /x;

        [download]

        Note that all of my $RX... regex variables are used inside other regexes and surrounded on both sides by \s* and various specific characters like parentheses and commas (this is a parser for a formally defined syntax that proceeds through the input text serially, I'm not trying to find $needle inside some giant $haystack). I do have individual tests for these variables now, previously I was only testing the parser at a higher level.

        - Andrew
[reply]
[d/l]
[select]
Re^9: Parser Performance Question
by LanX (Saint) on Oct 06, 2017 at 10:48 UTC
[reply]
[d/l]

In Section Seekers of Perl Wisdom