How can I extract a string between matching braces in Perl?

Question

My input file is as below :

HEADER 
{ABC|*|DEF {GHI 0 1 0} {{Points {}}}}

{ABC|*|DEF {GHI 0 2 0} {{Points {}}}}

{ABC|*|XYZ:abc:def {GHI 0 22 0} {{Points {{F1 1.1} {F2 1.2} {F3 1.3} {F4 1.4}}}}}

{ABC|*|XYZ:ghi:jkl {JKL 0 372 0} {{Points {}}}}

{ABC|*|XYZ:mno:pqr {GHI 0 34 0} {{Points {}}}}

{
    ABC|*|XYZ:abc:pqr {GHI 0 68 0}
        {{Points {{F1 11.11} {F2 12.10} {F3 14.11} {F4 16.23}}}}
        }
TRAILER

I want to extract the file into an array as below :

$array[0] = "{ABC|*|DEF {GHI 0 1 0} {{Points {}}}}"

$array[1] = "{ABC|*|DEF {GHI 0 2 0} {{Points {}}}}"

$array[2] = "{ABC|*|XYZ:abc:def {GHI 0 22 0} {{Points {{F1 1.1} {F2 1.2} {F3 1.3} {F4 1.4}}}}}"

..
..

$array[5] = "{
    ABC|*|XYZ:abc:pqr {GHI 0 68 0}
        {{Points {{F1 11.11} {F2 12.10} {F3 14.11} {F4 16.23}}}}
        }"

Which means, I need to match the first opening curly brace with its closing curly brace and extract the string in between.

I have checked the below link, but this doesnt apply to my question. Regex to get string between curly braces "{I want what's between the curly braces}"

I am trying but would really help if someone can assist me with their expertise ...

Thanks Sri ...

Answer 1

This can certainly be done with regex at least in modern versions of Perl:

my @array = $str =~ /( \{ (?: [^{}]* | (?0) )* \} )/xg;

print join "\n" => @array;

The regex matches a curly brace block that contains either non curly brace characters, or a recursion into itself (matches nested braces)

Edit: the above code works in Perl 5.10+, for earlier versions the recursion is a bit more verbose:

my $re; $re = qr/ \{ (?: [^{}]* | (??{$re}) )* \} /x;

my @array = $str =~ /$re/xg;

Answer 2

Use Text::Balanced

Answer 3

I second ysth's suggestion to use the Text::Balanced module. A few lines will get you on your way.

use strict;
use warnings;
use Text::Balanced qw/extract_multiple extract_bracketed/;

my $file;
open my $fileHandle, '<', 'file.txt';

{ 
  local $/ = undef; # or use File::Slurp
  $file = <$fileHandle>;
}

close $fileHandle;

my @array = extract_multiple(
                               $file,
                               [ sub{extract_bracketed($_[0], '{}')},],
                               undef,
                               1
                            );

print $_,"\n" foreach @array;

---

OUTPUT

{ABC|*|DEF {GHI 0 1 0} {{Points {}}}}
{ABC|*|DEF {GHI 0 2 0} {{Points {}}}}
{ABC|*|XYZ:abc:def {GHI 0 22 0} {{Points {{F1 1.1} {F2 1.2} {F3 1.3} {F4 1.4}}}}}
{ABC|*|XYZ:ghi:jkl {JKL 0 372 0} {{Points {}}}}
{ABC|*|XYZ:mno:pqr {GHI 0 34 0} {{Points {}}}}
{
    ABC|*|XYZ:abc:pqr {GHI 0 68 0}
        {{Points {{F1 11.11} {F2 12.10} {F3 14.11} {F4 16.23}}}}
        }

Answer 4

You can always count braces:

my $depth = 0;
my $out = "";
my @list=();
foreach my $fr (split(/([{}])/,$data)) {
    $out .= $fr;
    if($fr eq '{') {
        $depth ++;
    }
    elsif($fr eq '}') {
        $depth --;
        if($depth ==0) {
            $out =~ s/^.*?({.*}).*$/$1/s; # trim
            push @list, $out;
            $out = "";
        }
    }
}
print join("\n==================\n",@list);

This is old, plain Perl style (and ugly, probably).

Answer 5

I don't think pure regular expressions are what you want to use here (IMHO this might not even be parsable using regex).

Instead, build a small parser, similar to what's shown here: http://www.perlmonks.org/?node_id=308039 (see the answer by shotgunefx (Parson) on Nov 18, 2003 at 18:29 UTC)

UPDATE It seems it might be doable with a regex - I saw a reference to matching nested parentheses in Mastering Regular Expressions (that's available on Google Books and thus can be googled for if you don't have the book - see Chapter 5, section "Matching balanced sets of parentheses")

Answer 6

You're much better off using a state machine than a regex for this type of parsing.

Answer 7

Regular expressions are actually pretty bad for matching braces. Depending how deep you want to go, you could write a full grammar (which is a lot easier than it sounds!) for Parse::RecDescent. Or, if you just want to get the blocks, search through for opening '{' marks and closing '}', and just keep count of how many are open at any given time.