Release 0.202#
General#
- Fix correctness issue for queries involving aggregations over the result of an outer join (#10592). 
- Fix - map()to raise an error on duplicate keys rather than silently producing a corrupted map.
- Fix - map_from_entries()to raise an error when input array contains a- nullentry.
- Fix out-of-memory error for bucketed execution by scheduling new splits on the same worker as the recently finished one. 
- Fix query failure when performing a - GROUP BYon- jsonor- ipaddresstypes.
- Fix correctness issue in - line_locate_point(),- ST_IsValid(), and- geometry_invalid_reason()functions to not return values outside of the expected range.
- Fix failure in - geometry_to_bing_tiles()and- ST_NumPoints()functions when processing geometry collections.
- Fix query failure in aggregation spilling (#10587). 
- Remove support for - SHOW PARTITIONSstatement.
- Improve support for correlated subqueries containing equality predicates. 
- Improve performance of correlated - EXISTSsubqueries.
- Limit the number of grouping sets in a - GROUP BYclause. The default limit is- 2048and can be set via the- analyzer.max-grouping-setsconfiguration property or the- max_grouping_setssession property.
- Allow coercion between row types regardless of field names. Previously, a row type is coercible to another only if the field name in the source type matches the target type, or when target type has anonymous field name. 
- Increase default value for - experimental.filter-and-project-min-output-page-sizeto- 500kB.
- Improve performance of equals operator on - array(bigint)and- array(double)types.
- Respect - X-Forwarded-Protoheader in client protocol responses.
- Add support for column-level access control. Connectors have not yet been updated to take advantage of this support. 
- Add support for correlated subqueries with correlated - ORpredicates.
- Add - multimap_from_entries()function.
- Add - bing_tiles_around(),- ST_NumGeometries(),- ST_GeometryN(), and- ST_ConvexHull()geospatial functions.
- Add - wilson_interval_lower()and- wilson_interval_upper()functions.
- Add - IS DISTINCT FROMfor- jsonand- ipaddresstype.
Hive#
- Fix optimized ORC writer encoding of - TIMESTAMPbefore- 1970-01-01. Previously, the written value was off by one second.
- Fix query failure when a Hive bucket has no splits. This commonly happens when a predicate filters some buckets out entirely. 
- Remove the - hive.bucket-writingconfig property.
- Add support for creating and writing bucketed sorted tables. The list of sorting columns may be specified using the - sorted_bytable property. Writing to sorted tables can be disabled using the- hive.sorted-writingconfig property or the- sorted_writing_enabledsession property. The maximum number of temporary files for can be controlled using the- hive.max-sort-files-per-bucketproperty.
- Collect and store basic table statistics ( - rowCount,- fileCount,- rawDataSize,- totalSize) when writing.
- Add - hive.orc.tiny-stripe-thresholdconfig property and- orc_tiny_stripe_thresholdsession property to control the stripe/file size threshold when ORC reader decides to read multiple consecutive stripes or entire fires at once. Previously, this feature piggybacks on other properties.
CLI#
- Add peak memory usage to - --debugoutput.
SPI#
- Make - PageSorterand- PageIndexersupported interfaces.