Could someone please explain why NoPI tables do not support row partitining alone like PPI tables.
Why it must be column partitioned, in order to support row partitioning?
Perhaps it is as simple as that row partitioning is not a stand-alone feature on itself, but as an ordering-method part of PI / Columnar?
Taking a shot in the dark here.
When you add row partitioning to a regular PI table, you are changing the sort order of the rows in the table. Without PPI rows are sorted by rowhash & uniqueness, with PPI rows are sorted by partition, rowhash and uniqueness.
Rows in a NOPI table are not sorted - this is the reason that loading such tables is quicker.
So if you added PPI (sorting) to a non-sorted table (NOPI) there is a mis-match in how the data is organised.
Having said that, I don't know about why you're allowed row partitioning for columnar tables...