Release 0.202¶
General Changes¶
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 anull
entry.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 BY
onjson
oripaddress
types.Fix correctness issue in
line_locate_point()
,ST_IsValid()
, andgeometry_invalid_reason()
functions to not return values outside of the expected range.Fix failure in
geometry_to_bing_tiles()
andST_NumPoints()
functions when processing geometry collections.Fix query failure in aggregation spilling (#10587).
Remove support for
SHOW PARTITIONS
statement.Improve support for correlated subqueries containing equality predicates.
Improve performance of correlated
EXISTS
subqueries.Limit the number of grouping sets in a
GROUP BY
clause. The default limit is2048
and can be set via theanalyzer.max-grouping-sets
configuration property or themax_grouping_sets
session 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-size
to500kB
.Improve performance of equals operator on
array(bigint)
andarray(double)
types.Respect
X-Forwarded-Proto
header 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
OR
predicates.Add
multimap_from_entries()
function.Add
bing_tiles_around()
,ST_NumGeometries()
,ST_GeometryN()
, andST_ConvexHull()
geospatial functions.Add
wilson_interval_lower()
andwilson_interval_upper()
functions.Add
IS DISTINCT FROM
forjson
andipaddress
type.
Hive Changes¶
Fix optimized ORC writer encoding of
TIMESTAMP
before1970-01-01
. 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-writing
config property.Add support for creating and writing bucketed sorted tables. The list of sorting columns may be specified using the
sorted_by
table property. Writing to sorted tables can be disabled using thehive.sorted-writing
config property or thesorted_writing_enabled
session property. The maximum number of temporary files for can be controlled using thehive.max-sort-files-per-bucket
property.Collect and store basic table statistics (
rowCount
,fileCount
,rawDataSize
,totalSize
) when writing.Add
hive.orc.tiny-stripe-threshold
config property andorc_tiny_stripe_threshold
session 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 Changes¶
Add peak memory usage to
--debug
output.
SPI Changes¶
Make
PageSorter
andPageIndexer
supported interfaces.