Pandas Timestamp created from Arrow's don't behave correctly

[antoine-gallix]

I've encountered an issue while using Arrow to build panda's Timestamp data type. I've reported the bug on panda's repo but I've received an answer that suggest a solution might be on arrow's side so here I share the link to the issue: pandas-dev/pandas#31793

What is your opinion on where should be the problem resolved? On Arrow's side to make it behave more like a datetime, or on Panda's side to reject or transform Arrow instances when used to create a Timestamp?

[jadchaar]

Hmm it seems to be because we use dateutil timezone objects, but Pandas uses pytz. Thoughts @systemcatch?

[systemcatch]

Hey @antoine-gallix thanks for the report.

Jad has hit the nail on the head here in that arrow uses dateutil everywhere for timezones. It looks like pandas.Timestamp supports dateutil tzinfo's already which is good. I'll jump onto the pandas issue and provide some thoughts in a bit.

[systemcatch]

pandas-dev/pandas#37329 has been merged in to the pandas master branch. I believe this issue can be closed now.