20.69. Release 0.180
Fix a rare bug where rows containing only
nullvalues are not returned to the client. This only occurs when an entire result page contains only
nullvalues. The only known case is a query over an ORC encoded Hive table that does not perform any transformation of the data.
Fix incorrect results when performing comparisons between values of approximate data types (
DOUBLE) and columns of certain exact numeric types (
Fix query failure due to
NoClassDefFoundErrorwhen scalar functions declared in plugins are implemented with instance methods.
Improve performance of map subscript from O(n) to O(1) in all cases. Previously, only maps produced by certain functions and readers could take advantage of this improvement.
Skip unknown costs in
Support Secure Internal Communication between Presto nodes.
Add initial support for
Add support for
Add Thrift Connector that makes it possible to use Presto with external systems without the need to implement a custom connector.
/v1/resourceGroupStateREST endpoint on coordinator.
Fix skipping short decimal values in the optimized Parquet reader when they are backed by the
Ignore partition bucketing if table is not bucketed. This allows dropping the bucketing from table metadata but leaving it for old partitions.
Improve error message for Hive partitions dropped during execution.
The optimized RCFile writer is enabled by default, but can be disabled with the
hive.rcfile-optimized-writer.enabledconfig option. The writer supports validation which reads back the entire file after writing. Validation is disabled by default, but can be enabled with the
Add support for
Add support for pushdown of non-equality predicates on clustering keys.
JDBC Driver Changes
Add support for authenticating using Kerberos.
Allow configuring SSL/TLS and Kerberos properties on a per-connection basis.
Add support for executing queries using a SOCKS or HTTP proxy.
Add support for executing queries using an HTTP proxy.
Add running time limit and queued time limit to