Details
-
Bug
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
Impala 1.2.4
-
None
-
None
Description
The average row size comes from the tuple descriptor's average serialized size, which adds up all the slots.
For base table, only referenced column will have a slot added to the tuple descriptor. However, for view, we "register" all the columns refs during analysis. Therefore, the average row size is the entire row.
There's a TODO in TupleDescriptor.java (ComputeMemLayout) saying that we don't need to count non-materialized slots into the average serialized size.