[ARROW-14182] [C++][Compute] Hash Join performance improvement

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 6.0.0
• Fix Version/s: 9.0.0
• Component/s: C++
• Labels: pull-request-available query-engine
• External issue URL: https://github.com/apache/arrow/issues/29768
• Assignee: Michal Nowakiewicz
• Reporter: Michal Nowakiewicz
• Created: 30/Sep/21 07:56
• Updated: 11/Jan/23 08:38
• Resolved: 13/Jul/22 19:52

Description

Add micro-benchmarks for hash join exec node.
Write a new implementation of the interface HashJoinImpl making sure that it is efficient for all types of join. Current implementation, based on unordered map, trades performance for a simpler code and is likely not as fast as it could be.

Issue Links

• is a child of: ARROW-12633 [C++] Query engine umbrella issue
• is a parent of: ARROW-14479 [C++][Compute] Hash Join microbenchmarks
• links to: GitHub Pull Request #12326
• links to: GitHub Pull Request #13493

Sub-Tasks

• [C++] Add row-based utilities for encoding a batch and merging row tables: Open