Github
来源:Facebook
浏览 319
扫码
分享
2020-10-11 15:12:14
13.43. Release 0.209
General Changes
- Fix incorrect predicate pushdown when grouping sets contain the empty grouping set (#11296).
- Fix
X-Forwarded-Proto
header handling for requests to the /
path (#11168). - Fix a regression that results in execution failure when at least one of the arguments to
min_by()
or max_by()
is a constant NULL
. - Fix failure when some buckets are completely filtered out during bucket-by-bucket execution.
- Fix execution failure of queries due to a planning deficiency involving complex nested joins where a join that is not eligible for bucket-by-bucket execution feeds into the build side of a join that is eligible.
- Improve numerical stability for
corr()
, covar_samp()
, regr_intercept()
, and regr_slope()
. - Do not include column aliases when checking column access permissions.
- Eliminate unnecessary data redistribution for scalar correlated subqueries.
- Remove table scan original constraint information from
EXPLAIN
output. - Introduce distinct error codes for global and per-node memory limit errors.
- Include statistics and cost estimates for
EXPLAIN (TYPE DISTRIBUTED)
and EXPLAIN ANALYZE
. - Support equality checks for
ARRAY
, MAP
, and ROW
values containing nulls. - Improve statistics estimation and fix potential negative nulls fraction estimates for expressions that include
NOT
or OR
. - Completely remove the
SHOW PARTITIONS
statement. - Add
bing_tiles_around()
variant that takes a radius. - Add the
convex_hull_agg()
and geometry_union_agg()
geospatial aggregation functions. - Add
(TYPE IO, FORMAT JSON)
option for EXPLAIN that shows input tables with constraints and the output table in JSON format. - Add Kudu Connector.
- Raise required Java version to 8u151. This avoids correctness issues for map to map cast when running under some earlier JVM versions, including 8u92.
Web UI Changes
- Fix the kill query button on the live plan and stage performance pages.
CLI Changes
- Prevent spurious “No route to host” errors on macOS when using IPv6.
JDBC Driver Changes
- Prevent spurious “No route to host” errors on macOS when using IPv6.
Hive Connector Changes
- Fix data loss when writing bucketed sorted tables. Partitions would be missing arbitrary rows if any of the temporary files for a bucket had the same size. The
numRows
partition property contained the correct number of rows and can be used to detect if this occurred. - Fix cleanup of temporary files when writing bucketed sorted tables.
- Allow creating schemas when using
file
based security. - Reduce the number of cases where tiny ORC stripes will be written when some columns are highly dictionary compressed.
- Improve memory accounting when reading ORC files. Previously, buffer memory and object overhead was not tracked for stream readers.
- ORC struct columns are now mapped by name rather than ordinal. This correctly handles missing or extra struct fields in the ORC file.
- Add procedure
system.create_empty_partition()
for creating empty partitions.
Kafka Connector Changes
- Support Avro formatted Kafka messages.
- Support backward compatible Avro schema evolution.
SPI Changes
- Allow using
Object
as a parameter type or return type for SQL functions when the correponding SQL type is an unbounded generic.