[HIVE-26447] Vectorization: wrong results when filter on repeating map key orc table

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 4.0.0-alpha-1
• Fix Version/s: 4.0.0-alpha-2
• Component/s: Hive
• Labels: pull-request-available
• Hadoop Flags: Reviewed
• Assignee: Yi Zhang
• Reporter: Yi Zhang
• Created: 02/Aug/22 23:49
• Updated: 22/May/23 20:09
• Resolved: 05/Aug/22 18:53

Description

Example reproducible case:

 

set hive.vectorized.execution.enabled=true;

set hive.fetch.task.conversion=none;

create temporary table foo (id int, x map<string,int>) stored as orc;
insert into foo values(1, map('ABC', 9)), (2, map('ABC', 7)), (3, map('ABC', 8)), (4, map('ABC', 9));

select id from foo where x['ABC']=9;

this only gives 1, when correct result should be 1,4

For every VectorizedRowBatch, only the first row is checked.  

This seems to be a corner case of ORC table have repeating string type key for map field in the MapColumnVector.

Issue Links

• is related to: HIVE-27368 Vectorization: ArrayIndexOutOfBoundsException when map key vector has repeating values
• links to: GitHub Pull Request #3492
• links to: GitHub Pull Request #3509