Release 370 (3 Feb 2022)#

General#

  • Add support for DEFAULT keyword in ALTER TABLE...SET PROPERTIES.... (#10331)

  • Improve performance of map and row types. (#10469)

  • Improve performance when evaluating expressions in WHERE and SELECT clauses. (#10322)

  • Prevent queries deadlock when using phased execution policy with dynamic filters in multi-join queries. (#10868)

  • Fix query scheduling regression introduced in Trino 360 that caused coordinator slowness in assigning splits to workers. (#10839)

  • Fix information_schema query failure when an IS NOT NULL predicate is used. (#10861)

  • Fix failure when nested subquery contains a TABLESAMPLE clause. (#10764)

Security#

  • Reduced the latency of successful OAuth 2.0 authentication. (#10929)

  • Fix server start failure when using JWT and OAuth 2.0 authentication together (http-server.authentication.type=jwt,oauth2). (#10811)

CLI#

  • Add support for ARM64 processors. (#10177)

  • Allow to choose the way how external authentication is handled with the --external-authentication-redirect-handler parameter. (#10248)

RPM package#

  • Fix failure when operating system open file count is set too low. (#8819)

Docker image#

  • Change base image to registry.access.redhat.com/ubi8/ubi, since CentOS 8 has reached end-of-life. (#10866)

Cassandra connector#

  • Fix query failure when pushing predicates on uuid partitioned columns. (#10799)

ClickHouse connector#

  • Support creating tables with Trino timestamp(0) type columns.

  • Drop support for ClickHouse servers older than version 20.7 to avoid using a deprecated driver. You can continue to use the deprecated driver with the clickhouse.legacy-driver flag when connecting to old servers. (#10541)

  • Remove the legacy allow-drop-table configuration property. This defaulted to false, which disallowed dropping tables, but other modifications were still allowed. Use Обзор контроля доступа в CedrusData instead, if desired. (#588)

Druid connector#

Hive connector#

  • Improve query performance when reading ORC data. (#10575)

  • Add configuration property hive.single-statement-writes to require auto-commit for writes. This can be used to disallow multi-statement write transactions. (#10820)

  • Fix sporadic query failure Partition no longer exists when working with wide tables using a AWS Glue catalog as metastore. (#10696)

  • Fix SHOW TABLES failure when hive.hide-delta-lake-tables is enabled, and Glue metastore references the table with no properties. (#10864)

Iceberg connector#

  • Fix query failure when reading from a table that underwent partitioning evolution. (#10770)

  • Fix data corruption when writing Parquet files. (#9749)

MySQL connector#

Oracle connector#

Phoenix connector#

  • Fix incorrect result when a date value is older than or equal to 1899-12-31. (#10749)

PostgreSQL connector#

Redshift connector#

SingleStore (MemSQL) connector#

SQL Server connector#

SPI#

  • Allow null property names in ConnetorMetadata#setTableProperties. (#10331)

  • Rename ConnectorNewTableLayout to ConnectorTableLayout. (#10587)

  • Connectors no longer need to explicitly declare handle classes. The ConnectorFactory.getHandleResolver and Connector.getHandleResolver methods are removed. (#10858, #10872)

  • Remove unnecessary Block.writePositionTo and BlockBuilder.appendStructure methods. Use of these methods can be replaced with the existing Type.appendTo or writeObject methods. (#10602)