[ARROW-8381] [C++][Dataset] Dataset writing should require a writer schema

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Component/s: C++
• Labels: dataset
• External issue URL: https://github.com/apache/arrow/issues/24565
• Assignee: Ben Kietzman
• Reporter: Francois Saint-Jacques
• Created: 09/Apr/20 12:42
• Updated: 11/Jan/23 07:59
• Resolved: 05/Aug/20 19:02

Description

1. Dataset writing should always take an explicit writer schema instead of the first fragment's schema.
2. The MakeWritePlanImpl should not try removing columns that are found in the partition, this is left to the caller by passing an explicit schema.

Issue Links

• is superceded by: ARROW-8002 [C++][Dataset] Dataset writing should let you (re)partition the data