This release contains a regression which may cause Presto to incorrectly
Fix incorrect results when dividing certain decimal numbers (#12930).
Fix planning failure for certain join queries caused by invalid distribution type (#12354).
Add support for automatically retrying failed buckets when using grouped execution. Currently this is supported for certain
INSERTqueries using the Hive connector. This can be enabled with the
recoverable_grouped_executionsession property or the
recoverable-grouped-execution-enabledconfiguration property (#12124).
Add support for grouped execution for queries with no joins or aggregations. This can be enabled with the
grouped_execution_for_eligible_table_scanssession property or the
experimental.grouped-execution-for-eligible-table-scans-enabledconfiguration property (#12934).
Add configuration property
max-concurrent-materializationsand session property
max_concurrent_materializationsto limit the number of plan sections that will run concurrently when using materialized exchanges.
Add support for computing
approx_distinct()over BingTile values.
merge_hll()to merge an array of HyperLogLogs.
Web UI Changes#
Add completed and total lifespans to the Presto Coordinator UI.
Hive Connector Changes#
Fix failures for
information_schemaqueries when a table has an invalid storage format.
Improve query execution time over bucketed table with large buckets.
Add config property
hive.metastore.glue.catalogidto configure the Glue catalog ID.
Add experimental interface
ConnectorPlanOptimizerto allow connectors to participate in query plan optimization (e.g., filter pushdown) (#13102).
LogicalRowExpressions::FALSE_CONSTANTrespectively to avoid collision with