Are there any situations when you would run pure Python on EMR, as opposed to PySpark? Does this always just mean that the master node executes the code, rather than distrib
