Now the wrapper uses a line-by-line csv parsing. Switching to pandas would make it more flexible, maintanable and easy for users.