Release 342 (24 Sep 2020)#

General#

  • Add from_iso8601_timestamp_nanos() function. (#5048)

  • Improve performance of queries that use the DECIMAL type. (#4886)

  • Improve performance of queries involving IN with subqueries by extending support for dynamic filtering. (#5017)

  • Improve performance and latency of queries leveraging dynamic filters. (#4988)

  • Improve performance of queries joining tables with missing or incomplete column statistics when cost based optimization is enabled (which is the default). (#5141)

  • Reduce latency for queries that perform a broadcast join of a large table. (#5237)

  • Allow collection of dynamic filters for joins with large build side using the enable-large-dynamic-filters configuration property or the enable_large_dynamic_filters session property. (#5262)

  • Fix query failure when lambda expression references a table column containing a dot. (#5087)

Atop connector#

  • Fix incorrect query results when query contains predicates on start_time or end_time column. (#5125)

Elasticsearch connector#

  • Allow reading boolean values stored as strings. (#5269)

Hive connector#

  • Add support for S3 encrypted files. (#2536)

  • Add support for ABFS OAuth authentication. (#5052)

  • Support reading timestamp with microsecond or nanosecond precision. This can be enabled with the hive.timestamp-precision configuration property. (#4953)

  • Allow overwrite on insert by default using the hive.insert-existing-partitions-behavior configuration property. (#4999)

  • Allow delaying table scans until dynamic filtering can be performed more efficiently. This can be enabled using the hive.dynamic-filtering-probe-blocking-timeout configuration property or the dynamic_filtering_probe_blocking_timeout session property. (#4991)

  • Disable matching the existing user and group of the table or partition when creating new files on HDFS. The functionality was added in 341 and is now disabled by default. It can be enabled using the hive.fs.new-file-inherit-ownership configuration property. (#5187)

  • Improve performance when reading small files in RCTEXT or RCBINARY format. (#2536)

  • Improve planning time for queries with non-equality filters on partition columns when using the Glue metastore. (#5060)

  • Improve performance when reading JSON and CSV file formats. (#5142)

Iceberg connector#

  • Fix partition transforms for temporal columns for dates before 1970. (#5273)

Kafka connector#

  • Expose message headers as a _headers column of MAP(VARCHAR, ARRAY(VARBINARY)) type. (#4462)

  • Add write support for TIME, TIME WITH TIME ZONE, TIMESTAMP and TIMESTAMP WITH TIME ZONE for Kafka connector when using the JSON encoder. (#4743)

  • Remove JSON decoder support for nonsensical combinations of input-format-type / data-type. The following combinations are no longer supported: (#4743)

    • rfc2822: DATE, TIME, TIME WITH TIME ZONE

    • milliseconds-since-epoch: TIME WITH TIME ZONE, TIMESTAMP WITH TIME ZONE

    • seconds-since-epoch: TIME WITH TIME ZONE, TIMESTAMP WITH TIME ZONE

MySQL connector#

  • Improve performance of INSERT queries when GTID mode is disabled in MySQL. (#4995)

PostgreSQL connector#

  • Add support for variable-precision TIMESTAMP and TIMESTAMP WITH TIME ZONE types. (#5124, #5105)

SQL Server connector#

  • Fix failure when inserting NULL into a VARBINARY column. (#4846)

  • Improve performance of aggregation queries by computing aggregations within SQL Server database. Currently, the following aggregate functions are eligible for pushdown: count, min, max, sum and avg. (#4139)

SPI#

  • Add DynamicFilter.isAwaitable() method that returns whether or not the dynamic filter is complete and can be awaited for using the isBlocked() method. (#5043)

  • Enable connectors to wait for dynamic filters derived from replicated joins before generating splits. (#4685)