Details
-
Sub-task
-
Status: Open
-
Major
-
Resolution: Unresolved
-
None
-
None
Description
This task aims to revisit and extend the current "transitive operator selection" decisions, which allow us to pull operations, which receive their inputs from other spark operations and likely reduce data sizes into spark execution type even though these operations would fit in memory.
Examples with missing transitive operator selection are right indexing and grouped aggregates.