Hi all, I opened https://github.com/apache/arrow/pull/50146 almost four months ago. It has had no review from a committer yet, and I'd like to ask how best to get it reviewed and merged.
The PR lets the CSV reader use ConvertOptions::timestamp_parsers as a fallback, after ISO-8601, for columns explicitly typed as date32, date64, time32 or time64. Currently, those columns accept only strict ISO-8601, so values such as 15-OCT-15 with %d-%b-%y fail even when a matching parser is configured. Without timestamp_parsers the behavior is unchanged, and type inference is unchanged too. This PR addresses several long-standing issues, some of which have been open for years: - GH-28303 (open since 2021-04): unable to read date32/date64 in a specific format from CSV - GH-33357 (open since 2022-10): allow timestamp_parsers for date types - GH-31816 (open since 2022-05): strptime %b/%B fails on Windows - GH-31971 (open since 2022-05): strptime %p fails on Windows - GH-37180 (open since 2023-08): partly addressed, now works with timestamp_parsers=[ISO8601] - GH-41488 (open since 2024-05): the umbrella issue collecting the above What would be the best way to move this forward? Is there someone familiar with the CSV reader who could take a look? Best regards, Pearu
