Release 0.158#
General#
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
EXPLAIN ANALYZE
.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.
Return
NULL
fromelement_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 the
experimental.spill-enabled
configuration flag.Push down predicates for
DECIMAL
,TINYINT
,SMALLINT
andREAL
data types.
Hive#
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 coordinator memory.
Cassandra#
Allow starting the server even if a contact point hostname cannot be resolved.