Parsing can be tricky. Regexes aren't always the right answer. The main problem is that there are varying quoting rules and matching delimiter pairs that interact with each other. Coming up with a regex to handle it properly would be tricky.

Suppose you answered your immediate question. Then you'd encounter an expression like: f1, f2, substr('bizarro)',5,2)

You might lookover the Parse::RecDescent module to see if you can stand to use that.

...roboticus

Update: Regexes can be great for breaking the input stream into tokens, so perhaps you'll find it sufficient to tokenize your input, and then put some of the tokens back together to build the data statement you want. Once you tokenize your strings, then you can look for mismatched things like parens, and glue some tokens back together to get the results you want. (Poor man's parsing?) This approach probably has its own problems though...