Release 0.263#

Warning

There is a correctness bug around ORDER BY aggregations when spilling is enabled which is introduced by #16788

Details#

General Changes#

  • Fix a potential correctness bug for distinct aggregations with filter expressions containing null values.

  • Improve spilling CPU and memory efficiency for DISTINCT and ORDER BY aggregations.

  • Add support to create a heapdump when we exceed local memory limits for a query. This can be enabled with the heap_dump_on_exceeded_memory_limit_enabled session property and the directory to create the heapdump can be set by exceeded_memory_limit_heap_dump_file_directory session property.

  • Add key-based sampling feature that samples tables at query time. This feature can be enabled using session property key_based_sampling_enabled=true. Additionally, default sampling rate and sampling function name can be overridden by sessions properties key_based_sampling_percentage and key_based_sampling_function respectively.

Verifier Changes#

  • Improve false positives by skipping non-deterministic queries which have a LIMIT sub clause without ORDER BY.

Credits#

Ajay George, Arjun Gupta, Arunachalam Thirupathi, Ge Gao, James Petty, James Sun, Maria Basmanova, Neerad Somanchi, Nikhil Collooru, Pranjal Shankhdhar, Rongrong Zhong, SOURAV PAL, Sreeni Viswanadha, Swapnil Tailor, Tim Meehan, Zac Wen, Zhan Yuan, abhiseksaikia