The tokenizer breaks on strings with parens in them. You should be able to split on each character and implement a simple variant on the shunting yard algorith

The tokenizer breaks on strings with parens in them. You should be able to split on each character and implement a simple variant on the shunting yard algorithm in only a few more lines of code and get a robust tokenizer.