Hi, I'm writing a UDAF and I'm a little unclear about the PARTIAL1, PARTIAL2, FINAL and COMPLETE modes.
I've read the extent of the Javadoc ;) and looked at some of the built-in UDAFs in the Hive source tree and I'm still unclear about the properties of the input data in each aggregation step. Could anybody elaborate a little on the input data in each mode? Say, what are the safe assumptions for each mode assuming, e.g., CLUSTERED BY clause? thanks! alex