•
Extend CARPENTER to handle datasets with
both large number of columns and rows
•
Switch dynamically between column and row
enumeration based on estimated cost of
processing