Release 337 (25 Jun 2020)#

Примечание

This release fixes a potential security vulnerability when secure internal communication is enabled in a cluster. A malicious attacker can take advantage of this vulnerability to escalate privileges to internal APIs. We encourage everyone to upgrade as soon as possible.

General#

  • Fix incorrect results for inequality join involving NaN. (#4120)

  • Fix peak non-revocable memory metric in event listener. (#4096)

  • Fix queued query JMX stats. (#4129)

  • Fix rendering of types in the output of DESCRIBE INPUT. (#4023)

  • Improve performance of queries involving comparisons between DOUBLE or REAL values and integer values. (#3533)

  • Reduce idle CPU consumption in coordinator. (#3990)

  • Add peak non-revocable memory metric to query stats. (#4096)

  • Add support for variable-precision TIMESTAMP WITH TIME ZONE type (#3947)

  • Add support for IN predicate with subqueries in outer join condition. (#4151)

  • Add support for quantified comparisons (e.g., > ALL (...)) in aggregation queries. (#4128)

  • Add Druid коннектор. (#3522)

  • Add translate() function. (#4080)

  • Reduce worker graceful shutdown duration. (#4192)

Security#

  • Disable insecure authentication over HTTP by default when HTTPS with authentication is enabled. This can be overridden via the http-server.authentication.allow-insecure-over-http configuration property. (#4199)

  • Add support for insecure authentication over HTTPS to the Web UI. (#4199)

  • Add System information rules which control the ability of a user to access to read and write system management information. (#4199)

  • Disable user impersonation in default system security. (#4082)

Elasticsearch connector#

  • Add support for password authentication. (#4165)

Hive connector#

  • Fix reading CSV tables with separatorChar, quoteChar or escapeChar table property containing more than one character. For compatibility with Hive, only first character is considered and remaining are ignored. (#3891)

  • Improve performance of INSERT queries writing to bucketed tables when some buckets do not contain any data. (#1375)

  • Improve performance of queries reading Parquet data with predicates on timestamp columns. (#4104)

  • Improve performance for join queries over partitioned tables. (#4156)

  • Add support for null_format table property for tables using TextFile storage format (#4056)

  • Add support for null_format table property for tables using RCText and SequenceFile storage formats (#4143)

  • Add optimized Parquet writer. The new writer is disabled by default, and can be enabled with the parquet_optimized_writer_enabled session property or the hive.parquet.optimized-writer.enabled configuration property. (#3400)

  • Add support caching data in Azure Data Lake and AliyunOSS storage. (#4213)

  • Fix failures when caching data from Google Cloud Storage. (#4213)

  • Support ACID data files naming used when direct inserts are enabled in Hive (HIVE-21164). Direct inserts is an upcoming feature in Hive 4. (#4049)

PostgreSQL connector#

  • Improve performance of aggregation queries by computing aggregations within PostgreSQL database. Currently, the following aggregate functions are eligible for pushdown: count, min, max, sum and avg. (#3881)

Base-JDBC connector library#

  • Implement framework for aggregation pushdown. (#3881)