13.48. Release 0.204
- Use distributed join if one side is naturally partitioned on join keys.
- Improve performance of correlated subqueries when filters from outer query can be propagated to the subquery.
- Add support for all geometry types in .
- Add function.
- Add
to_big_endian_32()
and functions. - Add cast between
VARBINARY
type and type. - Make
lpad()
and functions supportVARBINARY
in addition toVARCHAR
. - Allow using arrays of mismatched lengths with . The missing positions are filled with
NULL
. - Track execution statistics of
AddExchanges
andPredicatePushdown
optimizer rules.
Event Listener Changes
- Add resource estimates to query events.
- Display resource estimates in Web UI query details page.
Resource Group Changes
- Fix unnecessary queuing in deployments where no resource group configuration was specified.
- Fix over-estimation of memory usage for scan operators when reading ORC files.
- Fix memory accounting for sort buffer used for writing sorted bucketed tables.
- Disallow creating tables with unsupported partition types.
- Support overwriting partitions for insert queries. This behavior is controlled by session property
insert_existing_partitions_behavior
. - Prevent the optimized ORC writer from writing excessively large stripes for highly compressed, dictionary encoded columns.
- Enable optimized Parquet reader and predicate pushdown by default.
Cassandra Connector Changes
- Optimize partition list retrieval for Cassandra 2.2+.