SQL tends to be non-composable which makes complicated scripts really messy to refactor and modify (or read even). CTEs make it more sensible but they're also a

SQL tends to be non-composable which makes complicated scripts really messy to refactor and modify (or read even). CTEs make it more sensible but they're also a rather recent addition and don't fully solve the problem. Data Science tends to involve a lot of modifying of the same code rather than creating one off or static scripts.