Release 0.167
Fix planning failure for certain aggregation with both and
GROUP BY
.Fix incorrect aggregation of operator summary statistics.
Fix a join issue that could cause joins that produce and filter many rows to monopolize worker threads, even after the query has finished.
Expand plan predicate pushdown capabilities involving implicitly coerced types.
Optimize constant patterns for
LIKE
predicates that use an escape character.Validate escape sequences in
LIKE
predicates per the SQL standard.Reduce memory usage of min_by() and .
Add transform_keys(), and zip_with() lambda functions.
Add JMX stat for the elapsed time of the longest currently active split.
Add JMX stats for compiler caches.
Raise required Java version to 8u92.
Security Changes
The
http.server.authentication.enabled
config option that previously enabled Kerberos has been replaced withhttp-server.authentication.type=KERBEROS
.Add support for LDAP Authentication using username and password.
Add a read-only named
read-only
.Allow access controls to filter the results of listing catalogs, schemas and tables.
Add operator-level performance analysis.
Lots of minor improvements.
JDBC Driver Changes
- Allow escaping in patterns.
Fix write operations for
ViewFileSystem
by using a relative location.Remove support for the
hive-cdh4
andhive-hadoop1
connectors which support CDH 4 and Hadoop 1.x, respectively.Remove the
hive-cdh5
connector as an alias forhive-hadoop2
.Remove support for the legacy S3 block-based file system.
Add support for KMS-managed keys for S3 server-side encryption.
Cassandra Changes
- Add support for Cassandra 3.x by removing the deprecated Thrift interface used to connect to Cassandra. The following config options are now defunct and must be removed:
cassandra.thrift-port
, ,cassandra.transport-factory-options
andcassandra.partitioner
.
Add methods to
SystemAccessControl
andConnectorAccessControl
to filter the list of catalogs, schemas and tables.Add access control checks for SHOW SCHEMAS and .