Regex: Embedded pattern-match modifiers: Empty (?) modifier

[AnomalousMonk]

In code in my Re^2: creating qr from existing regex reply to bart, there is the possibility of creating an empty  (?) embedded pattern-match modifier construct. However, all the docs I've seen specify:

"(?adlupimsx-imsx)"
"(?^alupimsx)"
        One or more embedded pattern-match modifiers ...

(emphasis added), not zero. Note, though, that (all) the docs go on to say:

This is particularly useful for dynamic patterns, such as those read in from a configuration file, taken from an argument, or specified in a table somewhere.

To me, dynamic generation implies the possibility of zero modifiers: one should seek to support the most general case.

The code here and below works with an empty  (?) construct for Perl standard releases 5.8 - 5.14, all I can test with ATM. (I haven't tested with other modifier groupings than i or not-i.)

c:\@Work\Perl\monks>perl -wMstrict -le
"my $string = '/.*u\/ba$/';
 ;;
 my $flag = qr{ [msix] }xms;
 ;;
 my $convertable =
 my ($pattern, $modifiers) = $string =~ m{
   \A \s* / (.*?) / ($flag*) \z
   }xms;
 die qq{bad rx: '$string'} unless $convertable;
 ;;
 my $rx = qr{(?$modifiers)$pattern};
 print $rx;
 print 'A: match' if 'u/ba' =~ $rx;
 ;;
 my $ry = qr{ \A foo $rx }xms;
 print $ry
 ;;
 print 'B: match' if 'foolubatu/ba' =~ $ry;
 print 'C: match' if 'Foolubatu/ba' =~ $ry;
"
(?-xism:(?).*u\/ba$)
A: match
(?msx-i: \A foo (?-xism:(?).*u\/ba$) )
B: match

My Questions:

• Does the code behave identically with other, later Perl releases?
• Does anyone have experience with or insight concerning the "empty" variation of this construct?
• Is use of  (?) a good idea, or should it be considered "experimental and subject to change"?

[Athanasius]

Hello AnomalousMonk,

The output can’t be exactly the same across versions 5.8 to 5.14, because in 5.14.0 the (?^...) construct was added to signify default modifiers (see perl5140delta#Regular-Expressions). So, from 5.14 on, the output is:

14:13 >perl 1868_SoPW.pl
(?^:(?).*u\/ba$)
A: match
(?^msx: \A foo (?^:(?).*u\/ba$) )
B: match

14:13 >

But with that caveat, the behaviour is identical across my Strawberry Perl versions 5.12.3, 5.14.4, 5.18.2, 5.20.2, 5.22.0, 5.24.1, and 5.26.0.

My guess — and it’s no more than a guess — is that the documentation is wrong, and should read: Zero or more embedded pattern-match modifiers ...

Hope that helps,

[shmem]

My guess — and it’s no more than a guess — is that the documentation is wrong, and should read: Zero or more embedded pattern-match modifiers ...

Even if it is only a guess that the documentation is wrong: if it is underpinned by valid test cases, this warrants a bug report, so this can be evaluated and eventually be fixed.

[choroba]

Reported. Watch the thread for updates, currently it isn't clear whether the code will be fixed, or the docs.