Well, with far less code, you could do the following:

my $string = q("one","two","three");
my @array = split( /"(?:,")?/, $string );
[download]

This will, however, create an empty element at the beginning of the list (the stuff before the opening quote of "one"). You could remove it afterwards with shift @array.

You could attempt to get rid of that extra element with a negative look-behind assertion:

 my @array = split( /(?<!^)"(?:,")?/, $string );
[download]

But that has the undesirable property of returning a string equivalent to:

 my @array = qw( "one two three );
[download]

Again, this is trivial to fix:

   $array[0] =~ s/^"/;
[download]

But, as you can see, when what appears as very simple code turns out to require some weird make-work code to solve the "almost-there" case, there's something wrong. This is what is known as code smell. I think what you really want is to load a module, such as Text::CSV_XS (or Text::CSV), and be done with it.

Even if you think it's overkill in this instance, it doesn't cost you much to learn how to use it. In the long run, you will have learnt how to use it in a simple scenario, and that will pay off later when you have some much hairier CSV problems, and you might risk trying to extend this code a bit more, rather than using a proven module... and that's when you'll pay the full price of not having learnt it now when it was easy.