Release 0.273¶

Highlights¶

• Add new Presto connector for Clickhouse.

Details¶

General Changes¶

• Fix disaggregated coordinator resource group aggregation issue related to WAITING_FOR_PREREQUISITE queries. When a query is moved to WAITING_FOR_PREREQUISITE, resource manager would end up counting them towards running queries. This gives a wrong impression to coordinator about running queries and it would end up running less queries than configured.

• Fix a bug where EXPLAIN (TYPE IO) did not include the index source from an index join.

• Fix a memory over-counting issue for using varchar types that could cause queries to fail with exceeded memory limit errors.

• Fix an off-by-one error during buffer size calculation in tdigest_agg().

• Fix memory tagging for UnnestOperator and TableWriterMergeOperator.

• Improve performance of :func: zip_with when used inside a reduce_agg().

• Add to_base32() and from_base32() varbinary functions.

• Add support for TRUNCATE TABLE statement and associated TRUNCATE.

• Add support to set timeout on the query analyzer runtime to prevent long running query analysis like complex regular expressions. The timeout duration can be set by the configuration property planner.query-analyzer-timeout or session property query_analyzer_timeout.

• Add trim_array() function that can be called to delete elements from the end of an ordinary array.

• Add a new optimization for showing results for (interactive) distinct limit N as they become available with no buffering. This can be enabled by setting the session property quick_distinct_limit_enabled to true.

• Add support for casting timestamp with micro/nanosecond precision.

• Add Read support for json & jsonb data types in postgresql.

• Add support for local round robin shuffle to reduce the partial distinct limit output size.

• Add support for column subfield access control checks in connectors. Connectors can specify subfield access control for row-type columns through checkCanSelectFromColumns() in SPI.

• Add warnings for approx_distinct() and approx_set() with low MAX_STANDARD_ERROR.

• Add support for retrying queries that failed with max requests queued exception.

• Update protoc and grpc version in presto-grpc-api module to adapt to linux aarch64.

• Upgrade zstandard compression to version 1.5.2.2. This improves the cpu used for data compressed with zstandard by about 2%.

• Upgrade postgresql driver to 42.3.3.

Clickhouse Connector Changes¶

• Add a Presto connector for Clickhouse with support of username/password authentication.

Hive Connector Changes¶

• Add a limit that prevents caching partitions with column counts greater than a configured threshold. This threshold can be set using the hive.partition-cache-column-count-limit configuration property.

• Add metastore configuration property hive.metastore.thrift.delete-files-on-table-drop to delete files on drop table.

• Add support for overwriting existing partitions with a Hive configuration property hive.insert-existing-partitions-behavior. This configuration property supersedes the legacy one hive.insert-overwrite-immutable-partitions-enabled. The new configuration property adds capability of overwriting new partitions for S3.

• Add support for timely Parquet metadata cache invalidation.

• Replace Hive session property streaming_aggregation_enabled with order_based_execution_enabled.

• Replace Hive configuration property hive.streaming-aggregation-enabled with hive.order-based-execution-enabled.

Iceberg Connector Changes¶

• Fix Iceberg $files table in case of column dropping.

• Add $properties system table.

• Add support for storing column comments for Iceberg tables.

• Upgrade Iceberg to 0.13.1.

Mongodb Connector Changes¶

• Add CAST(ObjectId() as STRING)().

Pinot Connector Changes¶

• Improve query performance by enabling pushdown of topn broker queries by default.

• Add double-quotes to Pinot generated queries to ensure that reserved keywords are escaped.

• Add TLS support in Pinot gRPC connection.

• Upgrade Pinot connector to support Pinot 0.10.0.

Spark Changes¶

• Improve handling of scenario where a query fails due to broadcast table violating node memory limit or broadcast limit. Previously the query failed while workers tried to load broadcast table causing container out of memory.

• Add a new configuration property spark.retry-on-out-of-memory-broadcast-join-enabled to disable broadcast join on broadcast OOM and retry the query again within the same spark session. This can be overridden by spark_retry_on_out_of_memory_broadcast_join_enabled session property.

Verifier Changes¶

• Add support to verifier for running control and test queries concurrently by setting concurrent-control-and-test configuration property to true.

SPI Changes¶

• Add getInfoMap method in ConnectorSplit which returns a Map<String, String>. This method should be preferred to the getInfo method which returns a raw object.

Credits¶

Ajay George, Amit Adhikari, Amit Dutta, Anuj Jamwal, Ariel Weisberg, Arjun Gupta, Arunachalam Thirupathi, Aryan, Asjad Syed, Beinan, Boris Verkhovskiy, Branimir Vujicic, Chen Li, Chunxu Tang, Derek Xia, Eduard Tudenhoefner, Fengpeng Yuan, Guy Moore, Harsha Rastogi, James Petty, James Sun, Jaromir Vanek, Jun, Ke Wang, Lin Liu, Maksim Dmitriyevich Podkorytov, Mandy Cho, Maria Basmanova, Michael Shang, Naveen Kumar Mahadevuni, Nikhil Collooru, Nirmit Shah, Ohm Patel, Pranjal Shankhdhar, Rebecca Schlussel, Reetika Agrawal, Rohit Jain, Rongrong Zhong, Ruslan Mardugalliamov, Sergey Pershin, Sergii Druzkin, Shashwat Arghode, Sreeni Viswanadha, Steve Chuck, Swapnil Tailor, Tim Meehan, Todd Gao, Valentin Touffet, Will Holen, Xiang Fu, Xiaoman Dong, Xinli shang, Yeikel, Zac, Zhenxiao Luo, abhiseksaikia, ellison840611, imjalpreet, shidayang, singcha, v-jizhang, xingmao.zheng, xyueji, yangyicheng, zhangbutao