Feature Type

  • [x] Adding new functionality to pandas

  • [ ] Changing existing functionality in pandas

  • [ ] Removing existing functionality in pandas

Problem Description

Currently df.convert_dtypes() will convert to nullable types (Int64, Boolean, etc), but it would be nice if there was an option to opt into the pyarrow-backed nullable types via pd.ArrowDtype

Feature Description

Maybe either

  • New keyword argument nullable_type="pandas"|"pyarrow"
  • Refactor io.nullable_backend global setting to mode.nullable_backend and allow that global config to apply to df.convert_dtypes

Alternative Solutions

For loop over the existing DataFrame object using astype and constructing pd.ArrowDtype?

Additional Context

No response

Comment From: phofl

I'd prefer the option solution I think, keeps us consistent with the I/O stuff

Comment From: seanjedi

take

Comment From: mroeschke

@seanjedi I would advise waiting until there are been a little more discussion regarding the direction of this issue.

Comment From: seanjedi

@mroeschke Hmm alright, let me know when I should work on it. I am a bit new here, and this issue seems to be doable for my skill level, but if I am committing too much than I can handle, then feel free to let me know. I also plan on working on this with a group of peers, so it won't just be me if I work on it. (A group of 3 to be exact)

Should I unassign myself?

Comment From: mroeschke

Yeah additionally I was hoping to work on this myself in a few days after some discussion and consensus.

May I recommend some of these older issues with a lot of community interest? https://github.com/pandas-dev/pandas/issues?q=is%3Aissue+is%3Aopen+sort%3Areactions-%2B1-desc

Comment From: seanjedi

Alright, I will take a look then, thanks for the recommendation