[ARROW-1377] [Python] Add function to assist with benchmarking Parquet scan performance

• Type: New Feature
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.7.0
• Component/s: Python
• External issue URL: https://github.com/apache/arrow/issues/17406
• Assignee: Wes McKinney
• Reporter: Wes McKinney
• Created: 19/Aug/17 18:50
• Updated: 11/Jan/23 07:14
• Resolved: 05/Sep/17 23:13

Description

It would be simpler to assess the raw performance of parquet-cpp Parquet scans to have something akin to https://github.com/apache/parquet-cpp/blob/master/tools/parquet-scan.cc available as a callable Python function. This way we can isolate the performance of scanning the file from converting it to Arrow (and to pandas)

Issue Links

• depends upon: PARQUET-1083 [C++] Refactor core logic in parquet-scan.cc so that it can be used as a library function for benchmarking