[ARROW-14470] [Python] Expose the use_threads option in Feather read functions

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 7.0.0
• Component/s: Python
• Labels: feather pull-request-available
• External issue URL: https://github.com/apache/arrow/issues/18910
• Assignee: Alenka Frim
• Reporter: Joris Van den Bossche
• Created: 26/Oct/21 12:06
• Updated: 11/Jan/23 08:40
• Resolved: 27/Oct/21 16:12

Description

On the C++ side, the Feather V2 Reader wraps the IPC RecordBatchFileReader, which accepts IpcReadOptions which can control the use of threads (and the default memory pool and some other options).

On the Python (cython) side, those options are not passed through. As a consequence the use_threads keyword only disables multithreading in the conversion from arrow table to pandas DataFrame, and not the actual reading. As a follow-up on ARROW-13317, we can actually make this keyword control both.p

Issue Links

• is related to: ARROW-13317 [Python] Improve documentation on what 'use_threads' does in 'read_feather'
• links to: GitHub Pull Request #11558