Doing a proper job of rolling your own calls for character-by-character parsing. Parse::RecDescent is a worthwhile tool to learn, but if you'll never get more complicated than the OP task, this would probably suffice:
use strict;
while (<DATA>) {
chomp;
my $nested = 0;
my @out = ();
my $tkn = '';
for my $c ( split // ) {
if ( $c eq '(' ) {
$nested++;
$tkn .= $c;
}
elsif ( $c eq ')' ) {
if ( $nested ) {
$nested--;
} else {
warn "unmatched ')' in $_\n";
}
$tkn .= $c;
}
elsif ( $nested or $c ne ',' ) {
$tkn .= $c;
}
elsif ( $tkn ) {
push @out, $tkn;
$tkn = '';
}
}
if ( $nested ) {
warn "$nested unmatched '(' in $_\n";
} else {
push @out, $tkn if ( $tkn );
}
print join( " ;; ", @out ), "\n" if ( @out );
}
__DATA__
a,b,c,d
x,z,
a,(b,c(d,e)f)),h
(1,2)(345)
,(6,7),(789)
((345,6),567
,,m,,n,,(o,,p),,q,,
|