Release 0.181

    • Prevent queries from ever failing with “Table already exists”.

    • Fix excessive GC overhead caused by large arrays and maps containing VARCHAR elements.

    • Improve error handling when passing too many arguments to various functions or operators that take a variable number of arguments.

    • Improve performance of count(*) aggregations over subqueries with known constant cardinality.

    • Add per-task distribution information to the output of EXPLAIN ANALYZE.

    • Add support for DROP COLUMN in ALTER TABLE.

    • Change local scheduler to prevent starvation of long running queries when the cluster is under constant load from short queries. The new behavior is disabled by default and can be enabled by setting the config property task.level-absolute-priority=true.

    • Improve the fairness of the local scheduler such that long-running queries which spend more time on the CPU per scheduling quanta (e.g., due to slow connectors) do not get a disproportionate share of CPU. The new behavior is disabled by default and can be enabled by setting the config property task.legacy-scheduling-behavior=false.

    Hive Changes

    • Correctly categorize certain errors when writing tables using RCFile.

    • Decrease the number of file system metadata calls when reading tables.

    • Add support for dropping columns.

    • Add support for query cancellation using Statement.cancel().

    PostgreSQL Changes

    • Add support for operations on external tables.
    • Improve query performance by scanning index ranges in parallel.

    SPI Changes

    • Fix regression that broke serialization for SchemaTableName.