Matching n parentheses in perl regex

Question

Hi, I've got some data that I'm parsing in Perl, and will be adding more and more differently formatted data in the near future. What I would like to do is write an easy-to-use function, that I could pass a string and a regex to, and it would return anything in parentheses. It would work something like this (pseudocode):

sub parse {
  $data = shift;
  $regex = shift;

  $data =~ eval ("m/$regex/")
  foreach $x ($1...$n)
  {
    push (@ra, $x); 
  }
  return \@ra;
}

Then, I could call it like this:

@subs = parse ($data, '^"([0-9]+)",([^:]*):(\W+):([A-Z]{3}[0-9]{5}),ID=([0-9]+)');

As you can see, there's a couple of issues with this code. I don't know if the eval would work, the 'foreach' definitely wouldn't work, and without knowing how many parentheses there are, I don't know how many times to loop.

This is too complicated for split, so if there's another function or possibility that I'm overlooking, let me know.

Thanks for your help!

Answer 1

You are trying to parse a complex expression with a regex - which is an insufficient tool for the job. Recall that regular expressions cannot parse higher grammars. For intuition, any expression which might be nested cannot be parsed with regex.

Answer 2

In list context, a regular expression will return a list of all the parenthesized matches.

So all you have to do is:

my @matches = $string =~ /regex (with) (parens)/;

And assuming that it matched, @matches will be an array of the two capturing groups.

So using your regex:

my @subs = $data =~ /^"([0-9]+)",([^:]*):(\W+):([A-Z]{3}[0-9]{5}),ID=([0-9]+)/;

Also, when you have long regexes, Perl has the x modifier, which goes after the closing regex delimiter. The x modifier allows you to put white-space and newlines inside the regex for increased readability.

If you are worried about the capturing groups that might be zero length, you can pass the matches through @subs = grep {length} @subs to filter them out.

Answer 3

When you want to find text inside of pairs of parenthesis, you want to use Text::Balanced.

But, that is not what you want to do, so it will not help you.

Answer 4

Then, I could call it like this:

@subs = parse($data, 
          '^"([0-9]+)",([^:]*):(\W+):([A-Z]{3}[0-9]{5}),ID=([0-9]+)');

Instead, call it like:

parse($data, 
    qr/^"([0-9]+)",([^:]*):(\W+):([A-Z]{3}[0-9]{5}),ID=([0-9]+)/);

Further, your task would be made simpler if you can use named captures (i.e. Perl 5.10 and later). Here is an example:

#!/usr/bin/perl

use strict; use warnings;

my %re = (
    id => '(?<id> [0-9]+ )',
    name => '(?<name> \w+ )',
    value => '(?<value> [0-9]+ )',
);

my @this = (
    '123,one:12',
    '456,two:21',
);

my @that = (
    'one:[12],123',
    'two:[21],456',
);

my $this_re = qr/$re{id}   ,   $re{name}    : $re{value}/x;
my $that_re = qr/$re{name} : \[$re{value}\] , $re{id}   /x;

use YAML;

for my $d ( @this ) {
    print Dump [ parse($d, $this_re) ];
}

for my $d ( @that ) {
    print Dump [ parse($d, $that_re) ];
}

sub parse {
    my ($d, $re) = @_;
    return unless $d =~ $re;
    return my @result = @+{qw(id name value)};
}

Output:

---
- 123
- one
- 12
---
- 456
- two
- 21
---
- 123
- one
- 12
---
- 456
- two
- 21