[ARROW-14008] [R][Compute] ExecPlan_run should return RecordBatchReader instead of Table

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 6.0.0
• Component/s: R
• Labels: pull-request-available query-engine
• External issue URL: https://github.com/apache/arrow/issues/29608
• Assignee: Ben Kietzman
• Reporter: Ben Kietzman
• Created: 15/Sep/21 20:55
• Updated: 11/Jan/23 08:36
• Resolved: 23/Sep/21 20:38

Description

Wrapping a sink node's output as a RecordBatchReader is already supported in c++. Currently a table is returned in R's ExecPlan_run to skirt difficult lifetime considerations, but it'd be more correct to allow streaming batches out of an ExecPlan instead of doing all work before ExecPlan_run returns.

Issue Links

• is depended upon by: ARROW-14029 [R] Repair map_batches()
• links to: GitHub Pull Request #11213