20.51. Release 0.198
Perform semantic analysis before enqueuing queries.
Add support for selective aggregates (
Parse decimal literals (e.g.
DECIMALby default. Previously, they were parsed as
DOUBLE. This behavior can be turned off via the
parse-decimal-literals-as-doubleconfig option or the
current_datefailure when the session time zone has a “gap” at
1970-01-01 00:00:00. The time zone
America/Bahia_Banderasis one such example.
Add variant of
DATEwith an implicit one-day step increment.
Increase the maximum number of arguments for the
zip()function from 4 to 5.
Add configuration property
task.max-local-exchange-buffer-sizefor setting local exchange buffer size.
Add trace token support to the scheduler and exchange HTTP clients. Each HTTP request sent by the scheduler and exchange HTTP clients will have a “trace token” (a unique ID) in their headers, which will be logged in the HTTP request logs. This information can be used to correlate the requests and responses during debugging.
Improve query performance when dynamic writer scaling is enabled.
Improve performance of
Improve query latency when tables are known to be empty during query planning.
array_agg()to avoid excessive object overhead and native memory usage with G1 GC.
Improve performance for high-cardinality aggregations with
DISTINCTargument qualifiers. This is an experimental optimization that can be activated by disabling the
use_mark_distinctsession property or the
Improve parallelism of queries that have an empty grouping set.
Improve performance of join queries involving the
Resource Groups Changes
Query Queues have been removed. Resource Groups are always enabled. The config property
experimental.resource-groups-enabledhas been removed.
WEIGHTED_FAIRscheduling policy to select oldest eligible sub group of groups where utilization and share are identical.
--enable-authenticationoption has been removed. Kerberos authentication is automatically enabled when
Kerberos authentication now requires HTTPS.
Add support for using AWS Glue as the metastore. Enable it by setting the
hive.metastoreconfig property to
Fix a bug in the ORC writer that will write incorrect data of type
Add wildcard character
*which allows querying several MBeans with a single query.
Add performance statistics to query plan in
Page.getBlocks(). This call was rarely used and performed an expensive copy. Instead, use
Page.getBlock(channel)or the new helper
Improve validation of