codem/(?:-x pattern) #comment /x bug?/code

[dcorbin]

I'm at a loss to explain this behavior. Is this correct? Or is it a bug. This is version 5.005_03 built for i386-linux.

#!/usr/bin/perl

my $html = qq(  abcdef  );
sub check
{
        my $pattern = shift;

        my @br;

        eval "\@br = '  abcdef  ' =~ m/$pattern/x";

        die "$@\n" if $@;
        print "PATTERN=m/$pattern/x\n";
        print "ANSWER = ";
        print join("-",@br),"\n";
}

check('(?-x:(abcdef))');
check('(?-x:(abcdef)) # comment here');

executing this produces

PATTERN=m/(?-x:(abcdef))/x
ANSWER = abcdef
PATTERN=m/(?-x:(abcdef)) # comment here/x
ANSWER = 

I think both check's should produce the same results. Am I right or wrong?

[Shendal]

Tried your code with perl 5.6 (activestate build), and it works like a champ.



Output:
PATTERN=m/(?-x:(abcdef))/x
ANSWER = abcdef
PATTERN=m/(?-x:(abcdef)) # comment here/x
ANSWER = abcdef


Cheers,
Shendal

[Fastolfe]

I tried this with 5.00551 on Solaris 2.6 (egcs 2.91.57) and it failed on the 2nd pattern like the original poster. At first it seemed like the (?-x:...) bit was negating the x flag for the entire regexp, but I tried adding the comment to the pattern and it still failed to match, so I'm not sure. Very odd.

Like another poster, I was thinking that the end of the regexp being on the same line as the comment was causing the /x to be ignored (interpreted as part of the comment), but adding whitespace in there caused the pattern to fail in both cases (odd). Something isn't right.

[dcorbin]

Sounds like a bug that's been fixed then.

[Ovid]

First, your join statement on @br is useless as @br only contains one element.

Second, when you have a regex using a (?:...) modifier, anything between the question mark and colon acts as a local switch unless it's after a -, in which case, it serves to negate a switch that's been applied to the entire regex. In this case, the /x switch (or modifier, if you prefer) at the end of the regex has the regex ignore unescaped whitespace, which you don't have in your first regex. As a result, your /x switch serves no purpose in this regex, and would be negated by the ?-x: construct, in any event.

The reason why the second regex does not match is because it's trying to match "#commenthere" (note the lack of spaces). Since they don't appear in your target text, you get no match.

Hope this helps.

Cheers,
Ovid

Update: I'm a moron. I need to think about this one a bit more. It's a lot trickier than I first thought. Now that I have a better idea of what you are doing, this does look like a bug. Changing ?-x: to ?: makes it work, but (?-x:...) is supposed to be local.

P.S. I am using 5.005 (ActiveState on WinNT)

[dcorbin]

1) I know. It's code that's been simplified to show the problem.

2) That's what I thought. I know that /x is not relevant in the first case, but I was trying to show two very consistent cases, one of which works and one doesn't when I think they both should.

3) But the regex is applied with the x modifier at the expression level. Doesn't that mean that the comment should be treated as a comment?

[Hot Pastrami]

I believe you need the /x modifier for #2 to work... or use ?# for the comment

Update: Oops, you've got /x there already... sorry, didn't notice. Perhaps it's because the regexp ends on a line after the # for the comment? That may be crazy talk... I don't know a whole lot about these regular expressions yet.

Alan "Hot Pastrami" Bellows
-Sitting calmly with scissors-

[dcorbin]

I believe I HAVE the x modifier, applied to the entire expression.