The return type for foreachPartition is None, so this is expected. If you're looking to do arbitrary code execution and return a result, mapInPandas or Pandas UDFs are good choices - you'd want to combine those with something like a .toLocalIterator call to interact with results in Python