I just thought about the potential for this feature and checked to see if it existed yet as an idea. This could open a mind-blowing amount of analytical synergy between the high performance of analysis services and the endless statistical complexity of python. Imagine running a linear regression on a dynamic sample based on the filter context of the report or even the row context of the visual! The biggest problem I see would be performance issues. You would have to be very careful with your python code. I'm sure they could put in some processing power guardrails though, to keep measures using python from eating up all the resources. Not to mention they've already created similar functionality with Machine Learning Services in SQL Server. Great, great idea!
Provide DAX functions that execute R and Python scripts. Similar to how you can insert an R or Python step in Power Query, allow a construct similar to: Column = VAR __MyVar = PYTHON("python code ...
