General Changes

• Fix incorrect behavior of format() for char values. Previously, the function did not preserve trailing whitespace of the value being formatted. (#2629)
• Fix query failure in some cases when aggregation uses inputs from both sides of a join. (#2560)
• Fix query failure when dynamic filtering is enabled and the query contains complex multi-level joins. (#2659)
• Fix query failure for certain co-located joins when dynamic filtering is enabled. (#2685)
• Fix failure of SHOW statements or queries that access information_schema schema tables with an empty value used in a predicate. (#2575)
• Fix query failure when EXECUTE is used with an expression containing a function call. (#2675)
• Fix failure in SHOW CATALOGS when the user does not have permissions to see any catalogs. (#2593)
• Improve query performance for some join queries when Cost based optimizations are enabled. (#2722)
• Prevent uneven distribution of data that can occur when writing data with redistribution or writer scaling enabled. (#2788)
• Add support for CREATE VIEW with comment (#2557)
• Add support for all major geometry types to ST_Points(). (#2535)
• Add required_workers_count and required_workers_max_wait_time session properties to control the number of workers that must be present in the cluster before query processing starts. (#2484)
• Add physical_input_bytes column to system.runtime.tasks table. (#2803)
• Verify that the target schema exists for the USE statement. (#2764)
• Verify that the session catalog exists when executing SET ROLE. (#2768)

Server Changes

• Require running on Java 11. This requirement may be temporarily relaxed by adding -Dpresto-temporarily-allow-java8=true to the Presto JVM Config. This fallback will be removed in future versions of Presto after March 2020. (#2751)
• Add experimental support for running on Linux aarch64 (ARM64). (#2809)

Security Changes

• Principal Rules are deprecated and will be removed in a future release. These rules have been replaced with User Mapping, which specifies how a complex authentication user name is mapped to a simple user name for Presto, and Impersonation Rules which control the ability of a user to impersonate another user. (#2215)
• A shared secret is now required when using Secure Internal Communication. (#2202)
• Kerberos for Secure Internal Communication has been replaced with the new shared secret mechanism. The internal-communication.kerberos.enabled and internal-communication.kerberos.use-canonical-hostname configuration properties must be removed. (#2202)
• When authentication is disabled, the Presto user may now be set using standard HTTP basic authentication with an empty password. (#2653)

Web UI Changes

• Display physical read time in detailed query view. (#2805)

JDBC Driver Changes

• Fix a performance issue on JDK 11+ when connecting using HTTP/2. (#2633)
• Implement PreparedStatement.setTimestamp() variant that takes a Calendar. (#2732)
• Add roles property for catalog authorization roles. (#2780)
• Add sessionProperties property for setting system and catalog session properties. (#2780)
• Add clientTags property to set client tags for selecting resource groups. (#2468)
• Allow using the : character within an extra credential value specified via the extraCredentials property. (#2780)

CLI Changes

• Fix a performance issue on JDK 11+ when connecting using HTTP/2. (#2633)

Cassandra Connector Changes

• Fix query failure when identifiers should be quoted. (#2455)

Hive Connector Changes

• Fix reading symlinks from HDFS when using Kerberos. (#2720)
• Reduce Hive metastore load when updating partition statistics. (#2734)
• Allow redistributing writes for un-bucketed partitioned tables on the partition keys, which results in a single writer per partition. This reduces memory usage, results in a single file per partition, and allows writing a large number of partitions (without hitting the open writer limit). However, writing large partitions with a single writer can take substantially longer, so this feature should only be enabled when required. To enable this feature, set the use-preferred-write-partitioning system configuration property or the use_preferred_write_partitioning system session property to true. (#2358)
• Remove extra file status call after writing text-based, SequenceFile, or Avro file types. (#1748)
• Allow using writer scaling with all file formats. Previously, it was not supported for text-based, SequenceFile, or Avro formats. (#2657)
• Add support for symlink-based tables with Avro files. (#2720)
• Add support for ignoring partitions with a non-existent data directory. This can be configured using the hive.ignore-absent-partitions=true configuration property or the ignore_absent_partitions session property. (#2555)
• Allow creation of external tables with data via CREATE TABLE AS when both hive.non-managed-table-creates-enabled and hive.non-managed-table-writes-enabled are set to true. Previously this required executing CREATE TABLE and INSERT as separate statement (#2669)
• Add support for Azure WASB, ADLS Gen1 (ADL) and ADLS Gen2 (ABFS) file systems. (#2494)
• Add experimental support for executing basic Hive views. To enable this feature, the hive.views-execution.enabled configuration property must be set to true. (#2715)
• Add register_partition and unregister_partition procedures for adding partitions to and removing partitions from a partitioned table. (#2692)
• Allow running ANALYZE collecting only basic table statistics. (#2762)

Elasticsearch Connector Changes

• Improve performance of queries containing a LIMIT clause. (#2781)
• Add support for nested data type. (#754)

PostgreSQL Connector Changes

• Add read support for PostgreSQL money data type. The type is mapped to varchar in Presto. (#2601)

Other Connector Changes

These changes apply to the MySQL, PostgreSQL, Redshift, Phoenix and SQL Server connectors.

• Respect DEFAULT column clause when writing to a table. (#1185)

SPI Changes

• Allow procedures to have optional arguments with default values. (#2706)
• SystemAccessControl.checkCanSetUser() is is deprecated and has been replaced with User Mapping and SystemAccessControl.checkCanImpersonateUser(). (#2215)