Release 354 (19 Mar 2021)#
General#
Improve performance of certain queries involving
LEFT
,RIGHT
andFULL 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, setpreferred-write-partitioning-min-number-of-partitions
to1
. To disable preferred write partitioning, setuse-preferred-write-partitioning
tofalse
. (#6920)Fix incorrect results when multiple similar subqueries contain
UNION
clauses that differ only in theALL
vsDISTINCT
qualifier. (#7345)Fix
DELETE
andUPDATE
for connectors that subsume filters. (#7302)Fix failure when subqueries contain
EXCEPT
orINTERSECT
. (#7342)Fix failure of queries that contain
RIGHT JOIN
when late materialization is enabled. (#6784)
Security#
Fix retries for OAuth 2.0 authentication in case of token expiry. (#7172)
CLI#
Support OAuth 2.0 authentication. (#7054)
ClickHouse connector#
Use correct case for name of the schema in
CREATE SCHEMA
. (#7239)
Elasticsearch connector#
Fix failure when reading single-valued fields for array types. (#7012)
Hive connector#
Respect partition filter for
DELETE
andUPDATE
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 asvarchar
values, instead. (#7059)Rename
hive.parallel-partitioned-bucketed-inserts
configuration property tohive.parallel-partitioned-bucketed-writes
. (#7259)
Iceberg connector#
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)
MemSQL connector#
MySQL connector#
Use proper column type (
datetime(3)
) in MySQL when creating a table withtimestamp(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)