How can I store captures from a Perl regular expression into separate variables?

Question

I have a regex:

/abc(def)ghi(jkl)mno(pqr)/igs

How would I capture the results of each parentheses into 3 different variables, one for each parentheses? Right now I using one array to capture all the results, they come out sequential but then I have to parse them and the list could be huge.

@results = ($string =~ /abc(def)ghi(jkl)mno(pqr)/igs);

Answer 1

Your question is a bit ambiguous to me, but I think you want to do something like this:

my (@first, @second, @third);
while( my ($first, $second, $third) = $string =~ /abc(def)ghi(jkl)mno(pqr)/igs) {
    push @first, $first;
    push @second, $second;
    push @third, $third;
}

Answer 2

Starting with 5.10, you can use named capture buffers as well:

#!/usr/bin/perl

use strict; use warnings;

my %data;

my $s = 'abcdefghijklmnopqr';

if ($s =~ /abc (?<first>def) ghi (?<second>jkl) mno (?<third>pqr)/x ) {
    push @{ $data{$_} }, $+{$_} for keys %+;
}

use Data::Dumper;
print Dumper \%data;

Output:

$VAR1 = {
          'first' => [
                       'def'
                     ],
          'second' => [
                        'jkl'
                      ],
          'third' => [
                       'pqr'
                     ]
        };

For earlier versions, you can use the following which avoids having to add a line for each captured buffer:

#!/usr/bin/perl

use strict; use warnings;

my $s = 'abcdefghijklmnopqr';

my @arrays = \ my(@first, @second, @third);

if (my @captured = $s =~ /abc (def) ghi (jkl) mno (pqr) /x ) {
    push @{ $arrays[$_] }, $captured[$_] for 0 .. $#arrays;
}

use Data::Dumper;
print Dumper @arrays;

Output:

$VAR1 = [
          'def'
        ];
$VAR2 = [
          'jkl'
        ];
$VAR3 = [
          'pqr'
        ];

But I like keeping related data in a single data structure, so it is best to go back to using a hash. This does require an auxiliary array, however:

my %data;
my @keys = qw( first second third );

if (my @captured = $s =~ /abc (def) ghi (jkl) mno (pqr) /x ) {
    push @{ $data{$keys[$_]} }, $captured[$_] for 0 .. $#keys;
}

Or, if the names of the variables really are first, second etc, or if the names of the buffers don't matter but only order does, you can use:

my @data;
if ( my @captured = $s =~ /abc (def) ghi (jkl) mno (pqr) /x ) {
    push @{ $data[$_] }, $captured[$_] for 0 .. $#captured;
}

Answer 3

An alternate way of doing it would look like ghostdog74's answer, but using an array that stores hash references:

my @results;
while( $string =~ /abc(def)ghi(jkl)mno(pqr)/igs) {
    my ($key1, $key2, $key3) = ($1, $2, $3);
    push @results, { 
        key1 => $key1,
        key2 => $key2,
        key3 => $key3,
    };
}

# do something with it

foreach my $result (@results) {
    print "$result->{key1}, $result->{key2}, $result->{key3}\n";
}

with the main advantage here of using a single data structure, AND having a nice readable loop.

Answer 4

@OP, when parenthesis are captured, you can use the variables $1,$2....these are backreferences

$string="zzzabcdefghijklmnopqrsssszzzabcdefghijklmnopqrssss";
while ($string =~ /abc(def)ghi(jkl)mno(pqr)/isg) {
    print "$1 $2 $3\n";
}

output

$ perl perl.pl
def jkl pqr
def jkl pqr

Answer 5

You could have three different regex's each focusing on specific groups. Obviously, you would like to just assign different groups to different arrays in the regex, but I think your only option is to split the regex up.

Answer 6

You can write a regex containing named capture groups. You do this with the ?<myvar> construct at the beginning of the capture group:

/(?<myvar>[0-9]+)/

You may then refer to those named capture groups using a $+{myvar} form.

Here is a contrived example:

perl -ne '/^systemd-(?<myvar>[^:]+)/ && { print $+{myvar} . "\n"}' /etc/passwd

Given a typical password file, it pulls out the systemd users and returns the names less the systemd prefix. It uses a capture group named myvar. This is just an example thrown together to illustrate the use of capture group variables.