Details
-
Improvement
-
Status: Open
-
Major
-
Resolution: Unresolved
-
None
-
None
-
None
-
None
Description
Partition pruning results are cached during a query (SemanticAnalyzer and ParseContext are the scope).
We could also cache them between queries in MetaStore, which would be especially useful if metastore server is remote and thus long-lived/shared between clients.
It may be more complex than it seems due to OOM potential. Also the key would need to be changed since the same expression string that is currently used may mean different things for different queries.