[ARROW-14042] [C++] Improve performance on dictionaries for 'case_when' kernel

• Type: Improvement
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: C++
• Labels: kernel
• External issue URL: https://github.com/apache/arrow/issues/29639
• Assignee: Unassigned
• Reporter: David Li
• Created: 20/Sep/21 15:15
• Updated: 11/Jan/23 08:37

Description

https://github.com/apache/arrow/pull/11022#discussion_r712162040

If we change the dictionary builders to allow separately encoding a new value (and getting back and index) and appending a dictionary item index, we could avoid re-encoding the same values over and over.

Issue Links

• relates to: ARROW-13573 [C++] Support dictionaries directly in case_when kernel