Release 354 (19 Mar 2021)#

General#

Improve performance of certain queries involving LEFT, RIGHT and FULL JOIN when one side of the join is known to produce a single row. (#7090)
Preferred write partitioning is now triggered automatically when the estimated number of written partitions exceeds or is equal to preferred-write-partitioning-min-number-of-partitions. This estimation requires that the input tables of the query have statistics. To enable preferred partitioning unconditionally, set preferred-write-partitioning-min-number-of-partitions to 1. To disable preferred write partitioning, set use-preferred-write-partitioning to false. (#6920)
Fix incorrect results when multiple similar subqueries contain UNION clauses that differ only in the ALL vs DISTINCT qualifier. (#7345)
Fix DELETE and UPDATE for connectors that subsume filters. (#7302)
Fix failure when subqueries contain EXCEPT or INTERSECT. (#7342)
Fix failure of queries that contain RIGHT JOIN when late materialization is enabled. (#6784)

Respect partition filter for DELETE and UPDATE of ACID tables. Previously, the partition filter was ignored, resulting in the deletion or update of too many rows. (#7302)
Fix allocation of statement IDs for ACID tables, which could result in query failure or data loss due to creating multiple delta files with the same name. (#7302)
Fix incorrect query results when reading from an incorrectly bucketed table created and registered with the metastore by Spark. (#6848)
Avoid leaking file system connections or other resources when using the Avro file format. (#7178)
Fix query failure when columns of a CSV table are declared as a type other than varchar (string) in Glue metastore. Columns are now interpreted as varchar values, instead. (#7059)
Rename hive.parallel-partitioned-bucketed-inserts configuration property to hive.parallel-partitioned-bucketed-writes. (#7259)

Fix queries on empty tables without a snapshot ID that were created by Spark. (#7233)
Update to Iceberg 0.11.0 behavior for transforms of dates and timestamps before 1970. Data written by older versions of Trino and Iceberg will be read correctly. New data will be written correctly, but may be read incorrectly by older versions of Trino and Iceberg. (#7049)

Add support for MemSQL 3.2. (#7179)
Use correct case for name of the schema in CREATE SCHEMA. (#7239)
Improve performance of queries with ORDER BY ... LIMIT clause when the computation can be pushed down to the underlying database. (#7326)

Use proper column type (datetime(3)) in MySQL when creating a table with timestamp(3) column. Previously, the second fraction was being truncated. (#6909)
Use correct case for name of the schema in CREATE SCHEMA. (#7239)
Improve performance of queries with ORDER BY ... LIMIT clause when the computation can be pushed down to the underlying database. (#7326)

Fix incorrect query results for ORDER BY ... LIMIT clause when sorting on char or varchar columns and topn-pushdown.enabled configuration property is enabled. The optimization is now enabled by default. (#7170, #7314)
Use correct case for name of the schema in CREATE SCHEMA. (#7239)

Add support for parametric time type. (#7122)
Use correct case for name of the schema in CREATE SCHEMA. (#7239)
Improve performance of queries with ORDER BY ... LIMIT clause when the computation can be pushed down to the underlying database. (#7324)