Hi there, I previously raised an issue regarding arrow timestamp values overflowing when reading parquet type INT96 ( https://issues.apache.org/jira/browse/ARROW-12096).
I would like to try and add a contribution to this try and fix this, but wanted to check the following: - If I should start with the code in "arrow/cpp/src/parquet/" to try and identify a fix? - If there is any existing process or functions to check this overflow that I should apply to the reader? - Or if I should write a PR to expose some settings in the pyarrow parquet reader settings to read INT96 types as a different timestamp than what arrow infers? Thanks, Karik
