Release 0.282
Details
General Changes
Fix
TEMPORARY
definition CREATE FUNCTION and DROP FUNCTION.Fix a bug where
cardinality(map_keys(x))
andcardinality(map_values(x))
would return wrong results.Improve performance of
Explain (TYPE VALIDATE)
by returning immediately after analysis and ACL checks complete without executing a dummy query. The output column is now calledresult
rather thanvalid
.Improve error handling when using custom
FunctionNamespaceManagers
.Improve null inferencing for join nodes.
optimize_nulls_in_join
session property is deprecated and is instead replaced with a newjoins_not_null_inference_strategy
session property (and corresponding configuration propertyoptimizer.joins-not-null-inference-strategy
) to control null inferencing.Add a new UDF
array_cum_sum()
to return an array whose elements are the cumulative sum of the input array.Add a query optimization to rewrite left join with null check on right join key with semi join. It’s controlled by session property
rewrite_left_join_null_filter_to_semi_join
and the default value istrue
Add an optimization for queries with empty input. The optimization is controlled by session property
simplify_plan_with_empty_input
and the default value istrue
Add an optimization to convert applicable cross join with an or filter to inner join. It’s controlled by session property
rewrite_cross_join_or_to_inner_join
and the default value istrue
Add an optimization to optimize cross join with array contains filter, it’s controlled by session parameter
rewrite_cross_join_array_contains_to_inner_join
and the default value istrue
Add an optimization to push down filter expression evaluation through cross join. It’s controlled by session property
push_down_filter_expression_evaluation_through_cross_join
and the default value isREWRITTEN_TO_INNER_JOIN
Add preprocessing for metadata calls required for analysis. This feature is disabled by default and controlled by session property
pre_process_metadata_calls
.Upgrade Joda version to 2.12.2. Note: a corresponding update to the Java runtime should also be made to ensure consistent timezone data.
Upgrade AWS SDK version to 1.12.261.
Hive Connector Changes
Remove the implementation of Alluxio’s metadata store because this feature won’t be supported in Alluxio 3.0.0.
Upgrade Alluxio version from 2.8.1 to 2.9.3.
Iceberg Connector Changes
- Add parquet metadata caching in Apache Iceberg.
Presto on Spark Changes
Add option to use
selectAuthorizedIdentity
API for Presto on Spark whenpermissions.authorized-identity-selection-enable
set totrue
.Add an additional field
attemptNumber
inTaskId
which is used to capture task retries in Presto on Spark. For presto classic this field will be set to0
by default.Remove CREATE FUNCTION and DROP FUNCTION support from Presto on Spark.
JDBC Connector Changes
- Fix
TRUNCATE TABLE
for JDBC connector.
Resource Group Changes
- Add schema support in resource group
StaticSelector
.
Credits
397090770, Adele Okoubo, Aditi Pandit, Ajay George, Ali Parsaei, Amit Dutta, Anant Aneja, Arpit Porwal, Arun Thirupathi, Beinan, Bin Fan, Chandrashekhar Kumar Singh, Christian Zentgraf, Chunxu Tang, Deepak Majeti, Deepak Majeti, Eduard Tudenhoefner, Facebook Community Bot, Ge Gao, George Wang, GuChangyang, HolyLow, Ivan Sadikov, Jalpreet Singh Nanda (:imjalpreet), Jaromir Vanek, Jialiang Tan, Jiayan Wei, Jon Janzen, Karteek Murthy Samba Murthy, Ke, Krishna Pai, Linsong Wang, Lisheng Guan, Lyublena Antova, Mahadevuni Naveen Kumar, Masha Basmanova, Matthew Peveler, Miaojiang (MJ) Deng, Michael Shang, Nikhil Collooru, Orri Erling, Pedro Eugenio Rocha Pedreira, Pranjal Shankhdhar, Pratyaksh Sharma, Pratyush Verma, Rebecca Schlussel, Reetika Agrawal, Rohan Pednekar, Rohit Jain, Sagar Sumit, Sanchit Kashyap, Sergey Pershin, Sergii Druzkin, Shrinidhi Joshi, Steve Burnett, Swapnil Tailor, Timothy Meehan, Vivek, Ying, Yiqun (Ethan) Zhang, Zac, Zhenxiao Luo, dnskr, ellison840611, feilong-liu, frankobe, jaystarshot, v-jizhang, xiaoxmeng