Release 0.197#

General Changes#

  • Fix query scheduling hang when the concurrent_lifespans_per_task session property is set.

  • Fix failure when a query contains a TIMESTAMP literal corresponding to a local time that does not occur in the default time zone of the Presto JVM. For example, if Presto was running in a CET zone (e.g., Europe/Brussels) and the client session was in UTC, an expression such as TIMESTAMP '2017-03-26 02:10:00' would cause a failure.

  • Extend predicate inference and pushdown for queries using a <symbol> IN <subquery> predicate.

  • Support predicate pushdown for the <column> IN <values list> predicate where values in the values list require casting to match the type of column.

  • Optimize min() and max() to avoid unnecessary object creation in order to reduce GC overhead.

  • Optimize the performance of ST_XMin(), ST_XMax(), ST_YMin(), and ST_YMax().

  • Add DATE variant for sequence() function.

  • Add ST_IsSimple() geospatial function.

  • Add support for broadcast spatial joins.

Resource Groups Changes#

  • Change configuration check for weights in resource group policy to validate that either all of the subgroups or none of the subgroups have a scheduling weight configured.

  • Add support for named variables in source and user regular expressions that can be used to parameterize resource group names.

  • Add support for optional fields in DB resource group exact match selectors.

Hive Changes#

  • Fix reading of Hive partition statistics with unset fields. Previously, unset fields were incorrectly interpreted as having a value of zero.

  • Fix integer overflow when writing a single file greater than 2GB with optimized ORC writer.

  • Fix system memory accounting to include stripe statistics size and writer validation size for the optimized ORC writer.

  • Dynamically allocate the compression buffer for the optimized ORC writer to avoid unnecessary memory allocation. Add config property hive.orc.writer.max-compression-buffer-size to limit the maximum size of the buffer.

  • Add session property orc_optimized_writer_max_stripe_size to tune the maximum stipe size for the optimized ORC writer.

  • Add session property orc_string_statistics_limit to drop the string statistics when writing ORC files if they exceed the limit.

  • Use the view owner returned from the metastore at the time of the query rather than always using the user who created the view. This allows changing the owner of a view.

CLI Changes#

  • Fix hang when CLI fails to communicate with Presto server.

SPI Changes#

  • Include connector session properties for the connector metadata calls made when running SHOW statements or querying information_schema.

  • Add count and time of full GC that occurred while query was running to QueryCompletedEvent.

  • Change the ResourceGroupManager interface to include a match() method and remove the getSelectors() method and the ResourceGroupSelector interface.

  • Rename the existing SelectionContext class to be SelectionCriteria and create a new SelectionContext class that is returned from the match() method and contains the resource group ID and a manager-defined context field.

  • Use the view owner from ConnectorViewDefinition when present.