Release 0.217¶

General Changes¶

Fix poor query planning performance for queries submitted when the cluster does not have the minimum required number of workers available.
Fix a bug that caused a union of a table with a VALUES statement to execute on a single machine, which may result in out of memory errors.
Improve performance of some queries that use window functions by eliminating redundant shuffles.
Add grouped execution support for window functions.
Add support for the BOOLEAN type for EXPLAIN IO statements.
Add support for choosing distribution type for semi joins based on estimated cost when the session property join_distribution_type='AUTOMATIC' is set.
Add support for collecting table statistics on demand with the ANALYZE statement.
Add a config option (query.stage-count-warning-threshold) to specify a per-query threshold for the number of stages. When this threshold is exceeded, a TOO_MANY_STAGES warning is raised.
Add per-task peak user memory usage to query statistics.
Add CLI support for showing the amount of data spilled during query execution.
Add ST_Points() function.
Add Elasticsearch Connector.
Remove the system memory pool and related configuration properties (resources.reserved-system-memory, deprecated.legacy-system-pool-enabled) entirely. System memory pool was deprecated in 0.201, and it was unused by default since that release. All memory allocations will now be served from the general/user memory pool.

Improve live plan to show data transfer statistics as edge labels.
Improve live plan to contain more information about each node. For example, table scan nodes now show the name of the tables that are being scanned.
Add the ability to zoom when viewing plans for completed queries.
Add support for showing the amount of data spilled during query execution.

Fix an issue where a partially successful rollback of a write could cause data loss and corrupt the metastore. The hive.skip-target-cleanup-on-rollback configuration property can be used to skip deleting target directories when partition creation is rolled back.
Fix an issue where creating a table on S3 could fail for S3 prefixes without any associated objects (e.g., empty S3 directories).
Add support for ANALYZE statement in the Hive connector. It’s possible to specify a list of partitions to collect statistics for using the WITH properties of the ANALYZE statement.
Add configuration property hive.temporary-staging-directory-enabled and session property temporary_staging_directory_enabled to control whether a temporary staging directory should be used for write operations.
Add configuration property hive.temporary-staging-directory-path and session property temporary_staging_directory_path to control the location of temporary staging directory that is used for write operations. The ${USER} placeholder can be used to use a different location for each user (e.g., /tmp/${USER}).

Add new SPIs for the ANALYZE statement. By default, running the ANALYZE statement for a connector that does not implement these SPIs results in an error o``USER_ERROR`` with error code NOT_SUPPORTED.