[ARROW-16646] [C++] HashJoin node can crash if a key column is a scalar

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 9.0.0
• Component/s: C++
• Labels: pull-request-available
• External issue URL: https://github.com/apache/arrow/issues/31996
• Assignee: Sasha Krassovsky
• Reporter: Weston Pace
• Created: 24/May/22 22:40
• Updated: 11/Jan/23 11:45
• Resolved: 26/May/22 01:18

Description

This only happens when the node has a bloom filter pushed down into it. In that case it will attempt to hash the key columns in arrow::compute::HashJoinBasicImpl::ApplyBloomFiltersToBatch by calling Hashing32::HashBatch on a batch made up only of key columns.

If one of those key columns happens to be a scalar, and not an array, then this method triggers a DCHECK and crashes.

Issue Links

• links to: GitHub Pull Request #13236