BUG: replace hits maximum recursion depth

[camold]

import pandas as pd
pd.DataFrame({"foo": [0,1,2]}).replace(0,pd.NA)

crashes by reaching maximum recursion depth.

Interestingly enough a replacement with None instead or numpy.nan works as expected.

Pandas version 1.5.2

[phofl]

Hi, thanks for your report. What you are doing is not a good idea in general, because your data are cast to object. It should work nevertheless though

[dicristina]

This was patched and then the patch was reverted. Issue #45725.

[phofl]

Not sure why this wasn't closed. This is fixed on 1.5.3 and tests were added

[phofl]

Sorry this was incorrect, we decided not to backport because the change was to invasive