after https://github.com/pandas-dev/pandas/pull/15838 - [ ] add dir support in pd.read_parquet - [ ] push down filtering of row groups - [x] The timestamps_to_ms used for pyarrow is planned for deprecation in https://issues.apache.org/jira/browse/ARROW-622 (#17438)