Release 363 (6 Oct 2021)#

General#

  • Add HTTP event listener implementation which sends JSON serialized events to a remote HTTP endpoint. (#8821)

  • Improve performance of queries that group by bigint columns. (#9510)

  • Improve performance of queries that process row or array data. (#9402)

  • Improve query planning performance. (#9462)

  • Reduce query memory usage when spilling occurs. (#9270, #9275)

  • Reduce memory usage for processing JOIN clauses. (#9327)

  • Fix potential data loss in query results when clients retry requests to the coordinator. (#9453)

  • Fix incorrect result for comparisons between zero-valued decimals. (#8985)

  • Fix SHOW ROLES failure when there are no roles to display. (#9319)

  • Fix EXPLAIN ANALYZE to show estimates. (#9396)

  • Fix incorrect result for round() with precision set to 0. (#9371)

  • Respect deprecated.legacy-catalog-roles=true configuration property in SHOW ROLES, SHOW CURRENT ROLES and SHOW ROLE GRANTS statements. (#9496)

Python client#

  • Fix column type reported in cursor.description for time with time zone column. (#9460)

BigQuery connector#

  • Fix failure for queries where predicates on geography, array or struct column are pushed down to BigQuery. (#9391)

Cassandra connector#

  • Add support for Cassandra tuple type. (#8570)

Elasticsearch connector#

  • Add support for scaled_float type. (#9358)

Hive connector#

  • Support usage of avro_schema_url table property in partitioned tables. (#9370}

  • Add support for insert overwrite operations on S3-backed tables. (#9234)

  • Improve query performance when reading Parquet data with predicate on a decimal column. (#9338)

  • Fix Failed reading parquet data: Socket is closed by peer query failure when reading from Parquet table with a predicate. (#9097)

  • Fix query failure when updating or deleting from an ORC ACID transactional table that has some rows deleted since the last major compaction. (#9354)

  • Fix failure when reading large Parquet files. (#9469)

  • Fix failures for some UPDATE queries, such as those where the SET clause contains the same constant more than once. (#9295)

  • Fix incorrect results when filtering on Parquet columns containing a dot in their name. (#9516)

Iceberg connector#

  • Improve query performance when reading Parquet data with predicate on a decimal column. (#9338)

  • Fix support for comments when adding a new column. Previously, they were silently ignored. (#9123)

  • Fix page and block sizing when writing Parquet data. (#9326)

  • Fix failure when reading large Parquet files. (#9469)

MySQL connector#

  • Add support for variable precision time type. (#9339)

  • Support CREATE TABLE and CREATE TABLE AS statements for time type. (#9339)

Phoenix connector#

  • Allowing forcing the mapping of certain types to varchar. This can be enabled by setting the jdbc-types-mapped-to-varchar configuration property to a comma-separated list of type names. (#2084)

Pinot connector#

  • Fix failure when a column name is a reserved keyword. (#9373)

SQL Server connector#

  • Add support for SQL Server datetimeoffset type. (#9329)

  • Fix failure for queries where predicates on text or ntext typed columns are pushed down to SQL Server. (#9387)