Release 355 (8 Apr 2021)

General

Report tables that are directly referenced by a query in QueryCompletedEvent. (#7330)
Report columns that are the target of INSERT or UPDATE queries in QueryCompletedEvent. This includes information about which input columns they are derived from. (#7425, #7465)
Rename optimizer.plan-with-table-node-partitioning config property to optimizer.use-table-scan-node-partitioning. (#7257)
Improve query parallelism when table bucket count is small compared to number of nodes. This optimization is now triggered automatically when the ratio between table buckets and possible table scan tasks exceeds or is equal to optimizer.table-scan-node-partitioning-min-bucket-to-task-ratio. (#7257)
Include information about Spill to disk in EXPLAIN ANALYZE. (#7427)
Disallow inserting data into tables that have row filters. (#7346)
Improve performance of queries that can benefit from both Cost-based optimizations and join pushdown by giving precedence to cost-based optimizations. (#7331)
Fix inconsistent behavior for to_unixtime() with values of type timestamp(p). (#7450)
Change return type of from_unixtime() and from_unixtime_nanos() to timestamp(p) with time zone. (#7460)

Add assumeLiteralNamesInMetadataCallsForNonConformingClients parameter for use as a workaround when applications do not properly escape schema or table names in calls to DatabaseMetaData methods. (#7438)

Support Hive views containing LATERAL VIEW json_tuple(...) AS ... syntax. (#7242)
Fix incorrect results when reading from a Hive view that uses array subscript operators. (#7271)
Fix incorrect results when querying the $file_modified_time hidden column. (#7511)

Improve performance when fetching table metadata during query analysis. (#6975)
Improve performance of queries with ORDER BY ... LIMIT clause when the computation can be pushed down to the underlying database. (#7490)

Engine now uses ConnectorMaterializedViewDefinition#storageTable to determine materialized view storage table. (#7319)