13.27. Release 0.234
- Fix query failure for cross-joining bucketed tables with when the bucket counts are compatible but mismatched.
- Fix scheduling regression by setting the default scheduler to legacy scheduler.
- Improve coordinator reliability by adding support to cap the number of total tasks running in a coordinator by pausing scheduling. This can be enabled by the configuration property
experimental.max-total-running-task-count-to-not-execute-new-query
. - Improve the scale writer heuristics by considering overall producer buffer utilization. This can be enabled by the session property
optimized_scale_writer_producer_buffer
and the configuration propertyoptimized-scale-writer-producer-buffer
. - Improve end point
v1/resourceGroupState
to supporting filtering of resource groups that are dynamically generated. - Improve connection pooling to avoid running out of sockets.
- Add KHyperLogLog type and related functions.
- Add support for forcing streaming exchange for Mark Distinct even if materialized exchange is enabled. This can be enabled by the session property
use_stream_exchange_for_mark_distinct
and the configuration propertyquery.use-streaming-exchange-for-mark-distinct
. ().
Geospatial Changes
- Add support for spatial joins for join condition
ST_Distance(p1, p2) < r
.
- Add
ZSTD
support for writingORC
andDWRF
files. This can be enabled by setting session propertyhive.compression_codec
toZSTD
.
Verifier Changes
- Add support for verifying
SELECT
queries that produce orUNKNOWN
(null) columns, or structured typed columns withDATE
orUNKNOWN
. - Add support for auto-resolving control check query failures due to
EXCEEDED_TIME_LIMIT
. - Add determinism analysis support for simple queries with top-level
ORDER BY LIMIT
clause. (#14181).
- Add parameter
AccessControlContext
to all methods inSystemAccessControl
. - Add to SelectionContext.
Druid Changes
- Add support for
LIMIT
evaluation pushdown.