Release 0.174#

General Changes#

  • Fix correctness issue for correlated subqueries containing a LIMIT clause.

  • Fix query failure when reduce() function is used with lambda expressions containing array_sort(), shuffle(), reverse(), array_intersect(), arrays_overlap(), concat() (for arrays) or map_concat().

  • Fix a bug that causes underestimation of the amount of memory used by max_by(), min_by(), max(), min(), and arbitrary() aggregations over varchar/varbinary columns.

  • Fix a memory leak in the coordinator that causes long-running queries in highly loaded clusters to consume unnecessary memory.

  • Improve performance of aggregate window functions.

  • Improve parallelism of queries involving GROUPING SETS, CUBE or ROLLUP.

  • Improve parallelism of UNION queries.

  • Filter and projection operations are now always processed columnar if possible, and Presto will automatically take advantage of dictionary encodings where effective. The processing_optimization session property and optimizer.processing-optimization configuration option have been removed.

  • Add support for escaped unicode sequences in string literals.

  • Add SHOW GRANTS and information_schema.table_privileges table.

Hive Changes#

  • Change default value of hive.metastore-cache-ttl and hive.metastore-refresh-interval to 0 to disable cross-transaction metadata caching.

Web UI changes#

  • Fix ES6 compatibility issue with older browsers.

  • Display buffered bytes for every stage in the live plan UI.

SPI changes#

  • Add support for retrieving table grants.

  • Rename SPI access control check from checkCanShowTables to checkCanShowTablesMetadata, which is used for both SHOW TABLES and SHOW GRANTS.