Github
来源:Facebook
浏览 372
扫码
分享
2021-01-27 21:42:15
13.56. Release 0.206
General Changes
- Fix execution failure for certain queries containing a join followed by an aggregation when
dictionary_aggregation
is enabled. - Fix planning failure when a query contains a
GROUP BY
, but the cardinality of the grouping columns is one. For example: SELECT c1, sum(c2) FROM t WHERE c1 = 'foo' GROUP BY c1
- Fix high memory pressure on the coordinator during the execution of queries using bucketed execution.
- Add
ST_Union()
, ST_Geometries()
, ST_PointN()
, ST_InteriorRings()
, and ST_InteriorRingN()
geospatial functions. - Add
split_to_multimap()
function. - Expand the
approx_distinct()
function to support the following types: INTEGER
, SMALLINT
, TINYINT
, DECIMAL
, REAL
, DATE
, TIMESTAMP
, TIMESTAMP WITH TIME ZONE
, TIME
, TIME WITH TIME ZONE
, IPADDRESS
. - Add a resource group ID column to the
system.runtime.queries
table. - Add support for executing
ORDER BY
without LIMIT
in a distributed manner. This can be disabled with the distributed-sort
configuration property or the distributed_sort
session property. - Add implicit coercion from
VARCHAR(n)
to CHAR(n)
, and remove implicit coercion the other way around. As a result, comparing a CHAR
with a VARCHAR
will now follow trailing space insensitive CHAR
comparison semantics. - Improve query cost estimation by only including non-null rows when computing average row size.
- Improve query cost estimation to better account for overhead when estimating data size.
- Add new semantics that conform to the SQL standard for temporal types. It affects the
TIMESTAMP
(aka TIMESTAMP WITHOUT TIME ZONE
) type, TIME
(aka TIME WITHOUT TIME ZONE
) type, and TIME WITH TIME ZONE
type. The legacy behavior remains default. At this time, it is not recommended to enable the new semantics. For any connector that supports temporal types, code changes are required before the connector can work correctly with the new semantics. No connectors have been updated yet. In addition, the new semantics are not yet stable as more breaking changes are planned, particularly around the TIME WITH TIME ZONE
type.
JDBC Driver Changes
- Add
applicationNamePrefix
parameter, which is combined with the ApplicationName
property to construct the client source name.
Hive Connector Changes
- Reduce ORC reader memory usage by reducing unnecessarily large internal buffers.
- Support reading from tables with
skip.footer.line.count
and skip.header.line.count
when using HDFS authentication with Kerberos. - Add support for case-insensitive column lookup for Parquet readers.