Release 368 (11 Jan 2022)#

General#

  • Allow setting per task memory limits via query.max-total-memory-per-task config property or via query_max_total_memory_per_task session property. (#10308)

  • Improve wall time for query processing with the phased scheduling policy. The previous behavior can be restored by setting the query.execution-policy configuration property to legacy-phased. (#10350)

  • Enable phased scheduling policy by default. The previous behavior can be restored by setting the query.execution-policy configuration property to all-at-once. (#10455)

  • Improve performance of arithmetic operations involving decimals with precision larger than 18. (#10051)

  • Reduce risk of out-of-memory failure on congested clusters with high memory usage. (#10475)

  • Fix queries not being unblocked when placed in reserved memory pool. (#10475)

  • Prevent execution of REFRESH MATERIALIZED VIEW from getting stuck. (#10360)

  • Fix double reporting of scheduled time for scan operators in EXPLAIN ANALYZE. (#10472)

  • Fix issue where the length of log file names grow indefinitely upon log rotation. (#10394)

Hive connector#

  • Improve performance of decoding decimal values with precision larger than 18 in ORC, Parquet and RCFile data. (#10051)

  • Disallow querying the properties system table for Delta Lake tables, since Delta Lake tables are not supported. This fixes the previous behavior of silently returning incorrect values. (#10447)

  • Reduce risk of worker out-of-memory exception when scanning ORC files. (#9949)

Iceberg connector#

  • Fix Iceberg table creation with location when schema location inaccessible. (#9732)

  • Support file based access control. (#10493)

  • Display the Iceberg table location in SHOW CREATE TABLE output. (#10459)

SingleStore (MemSQL) connector#

  • Add support for time type. (#10332)

Oracle connector#

  • Fix incorrect result when a date value is older than or equal to 1582-10-14. (#10380)

Phoenix connector#

  • Add support for reading binary type. (#10539)

PostgreSQL connector#

  • Add support for accessing tables created with declarative partitioning in PostgreSQL. (#10400)

SPI#

  • Encode long decimal values using two’s complement representation and change their carrier type to io.trino.type.Int128 instead of io.airlift.slice.Slice. (#10051)

  • Fix ClassNotFoundException when using aggregation with a custom state type. (#10408)