Presto 337 Documentation

13.3. Release 335 (14 Jun 2020)

13.3. Release 335 (14 Jun 2020)#

General Changes#

  • Fix failure when reduce_agg() is used as a window function. (#3883)
  • Fix incorrect cast from TIMESTAMP (without time zone) to TIME type. (#3848)
  • Fix incorrect query results when converting very large TIMESTAMP values into TIMESTAMP WITH TIME ZONE, or when parsing very large TIMESTAMP WITH TIME ZONE values. (#3956)
  • Return VARCHAR type when substr() argument is CHAR type. (#3599, #3456)
  • Improve optimized local scheduling with regard to non-uniform data distribution. (#3922)
  • Add support for variable-precision TIMESTAMP (without time zone) type. (#3783)
  • Add a variant of substring() that takes a CHAR argument. (#3949)
  • Add information_schema.role_authorization_descriptors table that returns information about the roles granted to principals. (#3535)

Security Changes#

Web UI Changes#

  • Fix the value displayed in the worker memory pools bar. (#3920)

Accumulo Connector Changes#

  • The server-side iterators are now in a JAR file named presto-accumulo-iterators. (#3673)

Hive Connector Changes#

  • Collect column statistics for inserts into empty tables. (#2469)
  • Add support for information_schema.role_authorization_descriptors table when using the sql-standard security mode. (#3535)
  • Allow non-lowercase column names in system.sync_partition_metadata procedure. This can be enabled by passing case_sensitive=false when invoking the procedure. (#3431)
  • Support caching with secured coordinator. (#3874)
  • Prevent caching from becoming disabled due to intermittent network failures. (#3874)
  • Ensure HDFS impersonation is not enabled when caching is enabled. (#3913)
  • Add hive.cache.ttl and hive.cache.disk-usage-percentage cache properties. (#3840)
  • Improve query performance when caching is enabled by scheduling work on nodes with cached data. (#3922)
  • Add support for UNIONTYPE. This is mapped to ROW containing a tag field and a field for each data type in the union. For example, UNIONTYPE<INT, DOUBLE> is mapped to ROW(tag INTEGER, field0 INTEGER, field1 DOUBLE). (#3483)
  • Make partition_values argument to drop_stats procedure optional. (#3937)
  • Add support for dynamic partition pruning to improve performance of complex queries over partitioned data. (#1072)

Phoenix Connector Changes#

  • Allow configuring whether DROP TABLE is allowed. This is controlled by the new allow-drop-table catalog configuration property and defaults to true, compatible with the previous behavior. (#3953)

SPI Changes#

  • Add support for aggregation pushdown into connectors via the ConnectorMetadata.applyAggregation() method. (#3697)