[HIVE-24990] Support distinct in window aggregation in vectorized mode

• Type: Improvement
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: UDF, Vectorization
• Assignee: Unassigned
• Reporter: Krisztian Kasa
• Created: 08/Apr/21 04:54
• Updated: 08/Apr/21 04:54

Description

PTF operator can not be vectorized if query has windowing function with distinct because this version of the aggregate functions are not implemented yet.

SELECT sum(DISTINCT a) OVER (PARTITION BY b) FROM t1;

The only exception is count.

List of functions has vectorized version but does not have vectorized distinct version

row_number
rank
dense_rank
min
max
sum
avg
first_value
last_value