Release 0.113

    The ORC reader in the Hive connector is broken in this release.

    The cluster resource manager announced in Release 0.103 is now enabled by default. You can disable it with the flag. Memory limits can now be configured via query.max-memory which controls the total distributed memory a query may use and query.max-memory-per-node which limits the amount of memory a query may use on any one node. On each worker, the config property controls how much memory is reserved for internal Presto data structures and temporary allocations.

    All session properties now have a SQL type, default value and description. The value for can now be any constant expression, and the SHOW SESSION command prints the current effective value and default value for all session properties.

    This type safety extends to the where properties can be validated and converted to any Java type using SessionPropertyMetadata. For an example, see HiveSessionProperties.

    Note

    • Allow using any type with value window functions first_value(), , nth_value(), and lag().

    • Add function.

    • Add url_encode() and functions.

    • concat() now allows arbitrary number of arguments.

    • Fix an issue that caused some specific queries to fail in planning.

    • Fix the Hive metadata cache to properly handle negative responses. This makes the background refresh work properly by clearing the cached metadata entries when an object is dropped outside of Presto. In particular, this fixes the common case where a table is dropped using Hive but Presto thinks it still exists.

    • Fix metastore socket leak when SOCKS connect fails.

    Note

    This is a backwards incompatible change with the previous connector SPI. If you have written a connector that uses structural types, you will need to update your code to the new APIs.