BUG: nullable dtypes not preserved in Series.replace #44940
Conversation
jreback
added
Bug
Dtype Conversions
|
lgtm. let's get a few more eyes on this. cc @pandas-dev/pandas-core if any comments |
|
updated with tests for some more closed issues |
| # GH 24971, GH#23305 | ||
| ser = pd.Series(categorical) | ||
| result = ser.replace({"A": 1, "B": 2}) | ||
| expected = pd.Series(numeric).astype("category") |
There was a problem hiding this comment.
umm why is this a resulting categorical?
There was a problem hiding this comment.
bc that's what someone implemented for the Categorical.replace behavior. im on the fence about it, but for now i think we need to be consistent with it
There was a problem hiding this comment.
hmm ok let's open an issue about this, i dont' think we should raise (or coerce to object) rather than return a new categorical (but maybe others disagree)
There was a problem hiding this comment.
can you clarify? based on your earlier comment i expected the opposite opinion, so i think the "dont" may be an unintentional double-negative.
|
thanks followup comments |
replacewith nullable dtypes #40732Sits on top of #44932
test_replace_dtype is cribbed from #37512
Improves inconsistent Categorical behavior while removing a special-case kludge for Categorical in replace_list.
Fixes an inconsistent int->float behavior introduced (intentionally, the inconsistency wasn't noticed) in #15742