Re^4: Feature Idea: qr//e

[TheDamian]

One extra tweak to the prototype would allow you to add trailing modifiers without requiring the pesky quotes around them:

sub qre (&;*) {
    my $re = shift->();
    eval 'qr/$re/'.(shift//'') || die $@
}

my $regex = qre{ join '|', qw/foo bar/ }i;

print "$regex\n";
__END__

or, from Perl 5.20 onwards:

sub qre :prototype(&;*) {
    my $re = shift->();
    eval 'qr/$re/'.(shift//'') || die $@
}                 

my $regex = qre{ join '|', qw/foo bar/ }i;

print "$regex\n";

Damian

[choroba]

It works only for some of the modifiers:

my $regex = qre{ join '|', qw/foo bar/ }m;

Returns:

Global symbol "$regex" requires explicit package name at ./1.pl line 1
+1.
Execution of ./1.pl aborted due to compilation errors.

Similarly, s yields

Global symbol "$regex" requires explicit package name at ./1.pl line 1
+1.
syntax error at ./1.pl line 14, at EOF
  (Might be a runaway multi-line ;; string starting on line 11)
Execution of ./1.pl aborted due to compilation errors.

[TheDamian]

++choroba

You're quite right: not a good tweak when it's not 100% applicable.

Of course, you could still write:

my $regex = qre{ join '|', qw/foo bar/ }m =>;

but that's probably just adding insult to injury.


And, yes, I was indeed caught out by this because I no longer even think about the m or s modifiers.
Especially now that I can add a standard:

use re '/xms';

at the top of every Perl file. ;-)

Damian

[haukex]

Hi choroba,

Hm, the * prototype is a really cool idea, too bad about that... here's a less elegant workaround (also incorporates AnomalousMonk's suggestion). As far as I can tell, for single letter modifiers it's only m and s that cause a problem, the others (/ixpodualn) are fine, and multi-letter modifiers are fine as far as they don't clash with any other subs (like main, min, or sum) or operators (like and, although that's not a valid combination of modifiers). With the letters "msixpodualn" you can actually spell a lot of words :-)

use List::Util qw/min sum/;
sub qre (&;*) {
    my $re = shift->();
    eval 'qr/$re/'.(@_ ? lc shift : '') || die $@
}

say qre{ join '|', qw/foo bar/ }M;
say qre{ join '|', qw/foo bar/ }msx;
say qre{ join '|', qw/foo bar/ }Min;
say qre{ join '|', qw/foo bar/ }Sum;

__END__

(?^m:foo|bar)
(?^msx:foo|bar)
(?^min:foo|bar)
(?^ums:foo|bar)

Regards,
-- Hauke D

[AnomalousMonk]

Oops...

Well, qr//s should always be written with a standard /xms tail anyway, so

c:\@Work\Perl>perl -wMstrict -le
"sub qre (&;*) {
   my $re = shift->();
   return eval 'qr/$re/xms' . (@_ ? shift : '') or die $@;
   }
 ;;
 my $regex = qre{ join '|', qw/foo bar/ }i;
 print $regex;
"
(?msix:foo|bar)

:-))

[LanX]

Most probably the parser gets confused because m and s are builtin commands on their own and can't be bare words.

In other words

sub m {} should be illegal.

[huck]

sub m {} should be illegal

I had considered saying something like that in And i learned never to name my subroutines s but then i realized that &s($n,$k) probably would not have thrown the error.

[AnomalousMonk]

Replacing the expression  (shift//'') with  (@_ ? shift : '') gives you 5.10 agnosticism.