June 2025 Update

This update was generated with the assistance of AI. While we strive for accuracy, please note that AI-generated content may not always be error-free. We encourage you to verify any information that is important to you.

Core Library

• Add null key support for index join. #13891

• Add Async SpillMerger in LocalMerge. #13634

• Share filter among drivers for improved efficiency. #13784

• Make global arbitration consider query priority. #13827

• Fix anti and semi join result mismatch with filter and multiple matches. #13123

• Fix the server crash caused by remote exchange error. #13905

• Fix resource release for memoizing constant folding expression. #13755

• Fix Expr::isDeterministic for lambda functions. #13647

• Enable constant folding for lambda functions. #13642

Presto Functions

• Add noisy_avg_gaussian(), noisy_sum_gaussian(), noisy_count_gaussian(), noisy_count_if_gaussian() functions.

• Add tdigest_agg(), value_at_quatile(), values_at_quantiles(), quantile_at_value() functions.

• Add inverse_chi_squared_cdf(), inverse_f_cdf() functions.

• Add l2_squared() function.

• Add ST_Distance(), ST_GeometryType(), St_Centroid(), St_Boundary(), ST_XMin(), ST_XMax(), ST_YMin(), ST_YMax() functions.

• Add serialization and deserialization of Geometry type to/from WKT and WKB. #12771

• Fix values_at_quantiles() function to throw error on null input. #13810

• Optimize array_remove() to avoid unnecessary string copies.

Spark Functions

• Add unbase64(), trunc(), varchar_type_write_side_check(), cbrt() functions.

• Fix try_cast and cast function error case handling. #12993

• Fix lower() function on unicode character. #13158

• Fix handling of extreme floating-point values in from_json(). #13378

Connectors

• Add support for exist and list functions in HdfsFileSystem. #13813

• Add GCS filesystem operations: rmdir, mkdir, and rename. #13533, #13532, #13490

• Flatten complex-type vectors when writing to Parquet. #13338

• Fix multi range filter in timestamp reader for Parquet. #12926

• Fix NPE when reading complex type data from Parquet v2. #13512

Performance and Correctness

• Optimize getStringView performance. #13870

• Optimize streaming aggregation performance. #13812

• Reduce HashTable load factor from 0.875 to 0.7 for better performance. #13694

• Improve fillNewMemory tight loop performance. #13883

• Increase readBatchSize when the last batch is empty in TableScan. #13626

Credits

aditi-pandit, Anders Dellien, Andrii Rosa, Artem Selishchev, Bikramjeet Vig, Bowen Wu, Chandrashekhar Kumar Singh, Chengcheng Jin, Christian Zentgraf, David Reveman, Deepak Majeti, Devavret Makkar, Dharan Aditya, duanmeng, Eric Jia, Huameng (Michael) Jiang, iiFeung, Jacob Khaliqi, Jacob Wujciak-Jens, Jialiang Tan, Jim Meyering, Jimmy Lu, Jin Chengcheng, Joe Giardino, joey.ljy, Ke Jia, Ke Wang, Kent Yao, Kevin Wilfong, Konstantinos Karatsenidis, Kostas Xirogiannopoulos, Krishna Pai, lingbin, Luis Garcés-Erice, Mario Ruiz, Masha Basmanova, Natasha Sehgal, Oliver Xu, Orri Erling, Patrick Sullivan, Paul Meng, Peter Enescu, Ping Liu, Richard Barnes, Rui Mo, Shakyan Kushwaha, Wei He, wraymo, xhs7700, Xiao Du, xiaodou, Xiaoxuan Meng, Xin Zhang, Yabin Ma, Yi Cheng Lee, yingsu00, Yoav Helfman, yumwang@ebay.com, Zhichen Xu, zhli1142015, zml1206, Zoltan Arnold Nagy