12.18. Release 0.218
This release has the potential to produce incorrect results for three way joins where one of the joins is a and one subqueryof the FULL OUTER JOIN
has a GROUP BY
on the join key and some expression that could be evaluated to a constant. For details, pleaserefer to #12577.
- Fix failures in regular expression functions for certain inputs where the pattern contains word boundaries (e.g.
\b
). - Fix a memory leak that occurs when a query fails with a semantic or permission error.
- Improve performance for queries with
FULL OUTER JOIN
where join keys have the :func:COALESCE
function applied. - Improve cost based optimizer to make decisions based on estimated query peak memory.
- Improve cost based optimizer for certain queries using
ORDER BY
. - Improve performance for queries with an followed by
LIMIT
. - Improve the error message for
INSERT
queries where columns do not match the target table. - Add support for using binary encoding for coordinator-to-worker communication.This feature is experimental, and it can be enabled with the
experimental.internal-communication.binary-transport-enabled
configuration property.Enabling this feature may help with coordinator scalability and reduces network, CPU, and memory usage on the coordinator. - Add for the
SphericalGeography
type. - Add a system table
system.metadata.analyze_properties
that shows the properties supported by the statement. - Add support for resolving key conflicts when using
split_to_map()
. - Add support for processing JSON protocol messages by generating bytecode on the coordinator.This feature is experimental, and it can be enabled with the
experimental.json-serde-codegen-enabled
configuration property.
Security Changes
- Change principal hostname to be configurable in Kerberos authenticator.
Hive Connector Changes
- Improve Parquet reader performance by reducing redundant footer reads.
- Add support for skipping Glacier files in Amazon S3. This feature can be enabled by setting the
hive.s3.skip-glacier-objects
configuration property. - Add support for Parquet files written with Parquet v1.9+ that use
DELTA_BINARY_PACKED
encoding with theINT64
type. - Add support for dictionary filtering for Parquet v2 files that use
RLE_DICTIONARY
encoding.
- Add support for Search Guard in Elasticsearch connector. Please refer to forthe relevant configuration properties.
MySQL Connector Changes
- Allow creating or renaming tables, and adding, renaming, or dropping columns.
PostgreSQL Connector Changes
- Allow creating or renaming tables, and adding, renaming, or dropping columns.
- Allow creating or renaming tables, and adding, renaming, or dropping columns.
SQL Server Connector Changes
SPI Changes
- Add to allow connectors to individually opt-in to connector-specific functionality.