REGR: from_records not initializing subclasses properly

[phofl]

• closes REGR: DataFrame.from_records for subclasses no longer calls subclass constructor #57008 (Replace xxxx with the GitHub issue number)
• Tests added and passed if fixing a bug or adding a new feature
• All code checks passed.
• Added type annotations to new arguments/methods/functions.
• Added an entry in the latest doc/source/whatsnew/vX.X.X.rst file if fixing a bug or adding a new feature.

follow up from #58145

Are we back porting regression fixes or just into 3.0?

[mroeschke]

Are we back porting regression fixes or just into 3.0?

I think these should just go into 3.0 as 2.3 will just be new string functionality

[phofl]

Sounds good, I’ll move the whatsnew

[phofl]

Moved

[mroeschke]

Thanks @phofl

[TomNicholas]

I think these should just go into 3.0 as 2.3 will just be new string functionality

Hi guys, it's been 8 months and 3.0 hasn't materialized, meaning that this regression fix is not yet available. This is causing downstream problems in both geopandas and xarray.

The non-regressed version of pandas is now so old that it's no longer supported by xarray, which supports versions of dependencies for 12 months (which I think is pretty standard). This means that the latest versions of xarray are not compatible with any released version of pandas containing this fix. I believe geopandas has a similar problem (see geopandas/geopandas#3152 (comment)).

@phofl @mroeschke Any chance this fix could be backported, or released in 2.3.3? 🙏

[WillAyd]

@jorisvandenbossche

[ianhi]

I joined the dev meeting today and was helpfully suggested to open up a backport PR. I've done so here #62436