23.105. Release 0.108

General Changes

  • Fix incorrect query results when a window function follows a row_number() function and both are partitioned on the same column(s).
  • Fix planning issue where queries that apply a false predicate to the result of a non-grouped aggregation produce incorrect results.
  • Fix exception when ORDER BY clause contains duplicate columns.
  • Fix issue where a query (read or write) that should fail can instead complete successfully with zero rows.
  • Add normalize(), from_iso8601_timestamp(), from_iso8601_date() and to_iso8601() functions.
  • Add support for position() syntax.
  • Add Teradata compatibility functions: index(), char2hexint(), to_char(), to_date() and to_timestamp().
  • Make ctrl-C in CLI cancel the query (rather than a partial cancel).
  • Allow calling Connection.setReadOnly(false) in the JDBC driver. The read-only status for the connection is currently ignored.
  • Add missing CAST from VARCHAR to TIMESTAMP WITH TIME ZONE.
  • Allow optional time zone in CAST from VARCHAR to TIMESTAMP and TIMESTAMP WITH TIME ZONE.
  • Trim values when converting from VARCHAR to date/time types.
  • Add support for fixed time zones +00:00 and -00:00.
  • Properly account for query memory when using the row_number() function.
  • Skip execution of inner join when the join target is empty.
  • Improve query detail UI page.
  • Fix printing of table layouts in EXPLAIN.
  • Add Black Hole Connector.

Cassandra Changes

  • Randomly select Cassandra node for split generation.
  • Fix handling of UUID partition keys.