[ARROW-10796] [C++] Investigate RecordBatch sort performance

• Type: Wish
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 3.0.0
• Component/s: C++
• Labels: pull-request-available
• External issue URL: https://github.com/apache/arrow/issues/26736
• Assignee: Antoine Pitrou
• Reporter: Antoine Pitrou
• Created: 03/Dec/20 12:41
• Updated: 11/Jan/23 08:15
• Resolved: 14/Dec/20 08:24

Description

A RecordBatch is currently sorted as a one-chunk Table. It is suboptimal as it's paying the chunk resolution cost even though it's not necessary.

A dedicated RecordBatch sort may also form the basis for a faster Table sort (first sort individual batches, then use merge sort along the sorted batches).

Issue Links

• relates to: ARROW-10898 [C++] Investigate Table sort performance
• links to: GitHub Pull Request #8890