20.82. Release 0.158
- Fix regression that could cause high CPU and heap usage on coordinator
when processing certain types of long running queries.
- Fix incorrect pruning of output columns in
- Fix ordering of
CHAR values so that trailing spaces are ordered after control characters.
- Fix query failures for connectors that produce non-remotely accessible splits.
- Fix non-linear performance issue when parsing certain SQL expressions.
- Fix case-sensitivity issues when operating on columns of
ROW data type.
- Fix failure when creating views for tables names that need quoting.
element_at() for out-of-range indices instead of failing.
- Remove redundancies in query plans, which can reduce data transfers over the network and reduce CPU requirements.
- Validate resource groups configuration file on startup to ensure that all
selectors reference a configured resource group.
- Add experimental on-disk merge sort for aggregations. This can be enabled with
experimental.spill-enabled configuration flag.
- Push down predicates for
REAL data types.
- Add hidden
$bucket column for bucketed tables that
contains the bucket number for the current row.
- Prevent inserting into non-managed (i.e., external) tables.
- Add configurable size limit to Hive metastore cache to avoid using too much
- Allow starting the server even if a contact point hostname cannot be resolved.