13.20. Release 0.237

Highlights

Notes

General Changes

  • Fix NPE in common sub-expression optimization triggered by CASE-WHEN expression.
  • Fix compiler failure with function type when common sub-expression optimization is enabled.
  • Improve coordinator RPC performance by removing unused entries from TaskStatus.
  • Add JSON format for EXPLAIN with type LOGICAL and DISTRIBUTED.
  • Add functions array_sum(), array_average(), map_normalize() and set_agg().
  • Add session property warning_handling to control how warnings are handled. The options are SUPPRESS, NORMAL and AS_ERROR. The default value is NORMAL.
  • Add support for defining SQL-invoked functions in plugins.
  • Add support to control which worker can receive tasks by implementing the NodeStatusService interface. See #14535.
  • Add warning to use approx_distinct() when using COUNT(DISTINCT x).
  • Add support for listing functions whose names match a specified pattern using the SHOW FUNCTION LIKE syntax.
  • Add flatten_geometry_collections() function to recursively flatten GeometryCollections.

Cassandra Connector Changes

  • Fix missing Netty library introduced in version 0.229 that causes the Cassandra connector to fail to load and queries to fail.

Druid Connector Changes

  • Add support for Druid’s basic and kerberos authentication.

Elasticsearch Connector Changes

  • Fix predicate pushdown for Elasticsearch.
  • Improve Elasticsearch (save an extra hop) by using shard primary host in Elasticsearch connector.
  • Add Elasticsearch array support using definitions in the _meta field.
  • Add support for AWS IAM authorization to Elasticsearch connector.
  • Add support for Elasticsearch query string syntax.
  • Add support for nested types in Elasticsearch.
  • Add support for querying Elasticsearch aliases.
  • Add system.nodes table to Elasticsearch.
  • Add handling for mixed-case columns in Elasticsearch.
  • Add support to load tables dynamically in Elasticsearch.
  • Add support to refresh Elasticsearch nodes periodically.

Hive Connector Changes

  • Add functionality to specify cache quota with respect to a scope. A scope could be at global, schema, table, or partition level. Cache quota prevents queries scanning too much data to disrupt cache locality. Such queries can only use the cache within their own scopes. Cache quota now only works with FILE_MERGE cache. Turn it on with config cache.cache-quota-scope and cache.default-cache-quota.

Verifier Changes

  • Fix an internal error when session properties of a control or a test query contains query_max_execution_time.
  • Fix auto-resolution of checksum query failure due to query complexity.
  • Add control query IDs, test query IDs, and peak total memory to verification outputs.
  • Add support to resubmit verification for CLUSTER_OUT_OF_MEMORY and ADMINISTRATIVELY_PREEMPTED errors.
  • Add support to retry DESCRIBE queries failed with TIME_LIMIT_EXCEEDED.

Resource Groups Changes

  • Fix existing resource group to handle transition of leaf resource group to internal and vice versa.