Release 360 (30 Jul 2021)#
General#
Improve support for correlated subqueries with
GROUP BY
orLIMIT
and complex correlated filter conditions. (#8554)Report cumulative query system memory usage. (#8615)
Fix
DROP SCHEMA
andDROP SCHEMA RESTRICT
not to drop then schema if it is not empty. (#8660)Fix query failure when there is a constant predicate on some column
col
(e.gcol=1
), followed byORDER BY col
andLIMIT
. (#8535)Fix
SHOW CREATE SCHEMA
failure. (#8598)Fix query failure when running
SHOW CREATE SCHEMA information_schema
. (#8600)Improve performance of
WHERE
clause evaluation. (#8624)Reduce coordinator network load. (#8460)
Improve query performance by sending collected dynamic filters from coordinator to workers. (#5183)
Improve performance of inequality joins where join condition sides have different type. (#8500)
Improve performance of
IN (<subquery>)
expressions. (#8639)
Security#
Add support for automatic configuration of TLS for Secure internal communication. This removes the need to provision per-worker TLS certificates. (#7954)
CLI#
Fix auto completion when pressing the tab button. (#8529)
ClickHouse connector#
Add
write.batch-size
connector configuration property to control JDBC batch size used during writes. It can also be controlled using thewrite_batch_size
session property. (#8434)Add new configuration property
insert.non-transactional-insert.enabled
to allow INSERT queries to write directly to the target table. This can improve performance in some cases by sacrificing transactional behaviour. It can also be controlled usingnon_transactional_insert
session property. (#8496)
Elasticsearch connector#
Add support for assuming an IAM role. (#8714)
Hive connector#
Iceberg connector#
Add support for Trino views. (#8540)
Prevent incorrect query results by failing a query when Iceberg table has row-level deletes. (#8450)
Fix query failure when joining with a bucketed Iceberg table. (#7502)
Fix query failure when showing stats for a bucketed Iceberg table. (#8616)
Fix query failure when joining with a partitioned table that has structural columns (
array
,map
orrow
). (#8647)Fix failures for queries that write tables in Parquet format. (#5201)
Improve query planning time by reducing calls to the metastore. (#8676, #8689)
MemSQL connector#
Add
write.batch-size
connector configuration property to control JDBC batch size used during writes. It can also be controlled using thewrite_batch_size
session property. (#8434)Add new configuration property
insert.non-transactional-insert.enabled
to allow INSERT queries to write directly to the target table. This can improve performance in some cases by sacrificing transactional behaviour. It can also be controlled usingnon_transactional_insert
session property. (#8496)Partial support for
DELETE
statement where predicate can be fully pushed down to the remote datasource. (#6287)Fix performance regression of writes. (#8559)
MongoDB connector#
MySQL connector#
Add
write.batch-size
connector configuration property to control JDBC batch size used during writes. It can also be controlled using thewrite_batch_size
session property. (#8434)Add new configuration property
insert.non-transactional-insert.enabled
to allow INSERT queries to write directly to the target table. This can improve performance in some cases by sacrificing transactional behaviour. It can also be controlled usingnon_transactional_insert
session property. (#8496)Partial support for
DELETE
statement where predicate can be fully pushed down to the remote datasource. (#6287)Fix performance regression of writes. (#8559)
Oracle connector#
Add
write.batch-size
connector configuration property to control JDBC batch size used during writes. It can also be controlled using thewrite_batch_size
session property. (#8434)Add new configuration property
insert.non-transactional-insert.enabled
to allow INSERT queries to write directly to the target table. This can improve performance in some cases by sacrificing transactional behaviour. It can also be controlled usingnon_transactional_insert
session property. (#8496)Partial support for
DELETE
statement where predicate can be fully pushed down to the remote datasource. (#6287)
Phoenix connector#
Add
write.batch-size
connector configuration property to control JDBC batch size used during writes. It can also be controlled using thewrite_batch_size
session property. (#8434)Add new configuration property
insert.non-transactional-insert.enabled
to allow INSERT queries to write directly to the target table. This can improve performance in some cases by sacrificing transactional behaviour. It can also be controlled usingnon_transactional_insert
session property. (#8496)Improve performance of
ORDER BY ... LIMIT
queries on sorted data for Phoenix 5. (#8171)Partial support for
DELETE
statement where predicate can be fully pushed down to the remote datasource. (#6287)Fix performance regression of writes. (#8559)
PostgreSQL connector#
Add
write.batch-size
connector configuration property to control JDBC batch size used during writes. It can also be controlled using thewrite_batch_size
session property. (#8434)Add new configuration property
insert.non-transactional-insert.enabled
to allow INSERT queries to write directly to the target table. This can improve performance in some cases by sacrificing transactional behaviour. It can also be controlled usingnon_transactional_insert
session property. (#8496)Partial support for
DELETE
statement where predicate can be fully pushed down to the remote datasource. (#6287)
Redshift connector#
Add
write.batch-size
connector configuration property to control JDBC batch size used during writes. It can also be controlled using thewrite_batch_size
session property. (#8434)Add new configuration property
insert.non-transactional-insert.enabled
to allow INSERT queries to write directly to the target table. This can improve performance in some cases by sacrificing transactional behaviour. It can also be controlled usingnon_transactional_insert
session property. (#8496)Partial support for
DELETE
statement where predicate can be fully pushed down to the remote datasource. (#6287)
SQL Server connector#
Add
write.batch-size
connector configuration property to control JDBC batch size used during writes. It can also be controlled using thewrite_batch_size
session property. (#8434)Add new configuration property
insert.non-transactional-insert.enabled
to allow INSERT queries to write directly to the target table. This can improve performance in some cases by sacrificing transactional behaviour. It can also be controlled usingnon_transactional_insert
session property. (#8496)Partial support for
DELETE
statement where predicate can be fully pushed down to the remote datasource. (#6287)Fix performance regression of writes. (#8559)
SPI#
Cast materialized view storage table columns to match view column types. (#8408)
Remove deprecated
ConnectorSplitManager#getSplits
method overrides. (#8569)Introduce
ConnectorPageSource#getCompletedPositions
for tracing physically read positions. (#8524)Remove deprecated
TupleDomain.transform
. (#8056)