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.


  • 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.