How to dissect a Regexp object?

[kj]

Is there a way to tell if a given Regexp object, generated at
runtime, includes at least one pair of capture parentheses?

More generally, is there any documentation for the Regexp class?
(I'm referring to the class alluded to by the output of, e.g., ref
qr//). Running perldoc Regexp fails ("no docs found"), and perldoc
perlre does not say much at all about this class as such.

TIA!

Kynn

 

[C.DeRykus]

Is there a way to tell if a given Regexp object, generated at
runtime, includes at least one pair of capture parentheses?

More generally, is there any documentation for the Regexp class?
(I'm referring to the class alluded to by the output of, e.g., ref
qr//).  Running perldoc Regexp fails ("no docs found"), and perldoc
perlre does not say much at all about this class as such.

perldoc perlop (see: Regexp Quote-Like Operators)

The regex object is viewable as a string:

$ perl -le '$regex = qr/ab(\d+)/; print $regex'
(?-xism:ab(\d+))

 

[Ilya Zakharevich]

The Regexp 'class' doesn't have any methods[1], and isn't really useable
as a class at all. In Perl 5.12 it will be mostly replaced by a new
REGEXP svtype, which was what it should have been from the beginning.
(qr// will still return a ref blessed into Regexp, for compatibility.)

Is there any way to find that an object is of a REGEXP svtype (without
using overload::StrVal)? Without this, serialization is not possible;
witness failure of FreezeThaw...

Or, at least, get hints that it "might be REGEXP" (with no false
negatives), so that a call to overload::StrVal() is needed...

Yours,
Ilya

 

[C.DeRykus]

Is there a way to tell if a given Regexp object, generated at
runtime, includes at least one pair of capture parentheses?
...

If you have a recently current Perl version, you might be able
to leverage re::regexp_pattern in list context to check paren's.

On a Win32 5.10.1 strawberry distro for instance:

c:\strawberry\perl\bin\perl.exe -le "
use re 'regexp_pattern';
$r = qr/ab(\d+)/;
($pat) = regexp_pattern($r);
print $pat"
ab(\d+)

So you could parse $pat for capturing paren's. You'd need to
exclude certain assertions such as (? ... ) but that's left
as an exercise for the reader

 

[sln]

Is there a way to tell if a given Regexp object, generated at
runtime, includes at least one pair of capture parentheses?

More generally, is there any documentation for the Regexp class?
(I'm referring to the class alluded to by the output of, e.g., ref
qr//). Running perldoc Regexp fails ("no docs found"), and perldoc
perlre does not say much at all about this class as such.

TIA!

Kynn

Its not too hard to analyse the string returned by qr//
to get the start (and thereby the count) of capture groups.
To get the actual group text requires some recursion and thought.

use strict;
use warnings;

my $tmp = qr/\(\$th (i(s))(i(s))(i(s))(?i\(s)\)(i(s))(i(s))\))/x;
my @capt;

while ($tmp =~ /( (?<!\\)\((?!\?) )/xg ) {
push @capt, pos($tmp);
}
print "$tmp\n";
my ($i,$last) = (1,1);

for my $p (@capt) {
print (' 'x ($p - $last), $i++ % 10);
$last = $p+1;
}
print "\nFound ",scalar @capt, " capture groups\n";

__END__

(?x-ism:\(\$th (i(s))(i(s))(i(s))(?i\(s)\)(i(s))(i(s))\)))
1 2 3 4 5 6 7 8 9 0 1
Found 11 capture groups

 

[kj]

Its not too hard to analyse the string returned by qr//
to get the start (and thereby the count) of capture groups.
To get the actual group text requires some recursion and thought.

use strict;
use warnings;

my $tmp = qr/\(\$th (i(s))(i(s))(i(s))(?i\(s)\)(i(s))(i(s))\))/x;
my @capt;

while ($tmp =~ /( (?<!\\)\((?!\?) )/xg ) {
push @capt, pos($tmp);
}
print "$tmp\n";
my ($i,$last) = (1,1);

for my $p (@capt) {
print (' 'x ($p - $last), $i++ % 10);
$last = $p+1;
}
print "\nFound ",scalar @capt, " capture groups\n";

__END__

(?x-ism:\(\$th (i(s))(i(s))(i(s))(?i\(s)\)(i(s))(i(s))\)))
1 2 3 4 5 6 7 8 9 0 1
Found 11 capture groups

Thanks for this code! Now I must study it.

~K

 

[Martijn Lievaart]

Its not too hard to analyse the string returned by qr// to get the start
(and thereby the count) of capture groups. To get the actual group text
requires some recursion and thought.

use strict;
use warnings;

my $tmp = qr/\(\$th (i(s))(i(s))(i(s))(?i\(s)\)(i(s))(i(s))\))/x; my
@capt;

while ($tmp =~ /( (?<!\\)\((?!\?) )/xg ) {
push @capt, pos($tmp);
}
print "$tmp\n";
my ($i,$last) = (1,1);

for my $p (@capt) {
print (' 'x ($p - $last), $i++ % 10); $last = $p+1;
}
print "\nFound ",scalar @capt, " capture groups\n";

__END__

(?x-ism:\(\$th (i(s))(i(s))(i(s))(?i\(s)\)(i(s))(i(s))\)))
1 2 3 4 5 6 7 8 9 0 1
Found 11 capture groups

I think this will fail on the regexp /\\(.)/.

M4

 

[sln]

I think this will fail on the regexp /\\(.)/.

M4

Correct. Inserting (?:\\.)* should fix it.
See if this will fail on anything.

-sln

use strict;
use warnings;

my $tmp = qr/\(\$th(\\(?:.) \\(.\) \\\(.)(i(s))(i(s))(?i\(s)\)(i(s))(i(s))\)))/x;
my @capt;

# /(?<!\\)(?:\\.)*\((?!\?)/
# -------------------------
my $grprx = qr/
(?<!\\) # Not an escape behind us
(?:\\.)* # 0 or more escape + any char
\( # (
(?!\?) # Not a ? in front of us
/x;

while ($tmp =~ /($grprx)/g ) {
# print "'$1'\n";
push @capt, pos($tmp);
}
print "$tmp\n";
my ($i,$last) = (1,1);

for my $p (@capt) {
print (' 'x ($p - $last), $i++ % 10);
$last = $p+1;
}
print "\nFound ",scalar @capt, " capture groups\n";

__END__

(?x-ism:\(\$th(\\(?:.) \\(.\) \\\(.)(i(s))(i(s))(?i\(s)\)(i(s))(i(s))\))))
1 2 3 4 5 6 7 8 9 0 1
Found 11 capture groups