Here is another attempt at expanding a one-liner into a big script, this one using a canned /regexp?/i:

use strict;
use warnings;
use Regexp::Common qw(balanced);

my @streams = (qw(
                  [ABC][DEF][HIJ]
                  ABC[DEF]HI[JK][LMNO]PQ
                  [A][B]C[DEF]HI[JK][LMNO]PQ
                  [A][B]C[DEF]HI[JK][LMNO]P[Q]
                  [ABCDEFHIJKLMNOPQ]
                  ABCDEF
                  ), q());
my $PATTERN = $RE{balanced}{-parens=>'[]'};

for my $input (@streams) {

    print qq($input => );
    my @pieces = ();

    while ($input =~ s/(\w*)($PATTERN)//) {
        my ($prematch, $match) = ($1, $2);

        push @pieces, split //, $prematch;
        push @pieces, $1 if $match =~ /(\w+)/;
    }
    push @pieces, split //, $input if length $input;

    printf qq(%s\n), join qq( ), @pieces;
}


__END__
[download]

which prints:

[ABC][DEF][HIJ] => ABC DEF HIJ
ABC[DEF]HI[JK][LMNO]PQ => A B C DEF H I JK LMNO P Q
[A][B]C[DEF]HI[JK][LMNO]PQ => A B C DEF H I JK LMNO P Q
[A][B]C[DEF]HI[JK][LMNO]P[Q] => A B C DEF H I JK LMNO P Q
[ABCDEFHIJKLMNOPQ] => ABCDEFHIJKLMNOPQ
ABCDEF => A B C D E F
 =>
[download]

Cheers.