DEPR: Remove NumericIndex #51139
DEPR: Remove NumericIndex #51139
Conversation
| @@ -5375,6 +5362,7 @@ def identical(self, other) -> bool: | |||
| for c in self._comparables | |||
| ) | |||
| and type(self) == type(other) | |||
| and self.dtype == other.dtype | |||
There was a problem hiding this comment.
Previously the index type could differentiate e.g. Index([1, 2, 3], dtype=object) and Index([1, 2, 3], dtype="int64") (the latter actually being a NumericIndex), but we now need to be more precise.
| @@ -539,7 +538,6 @@ def __new__( | |||
|
|
|||
| klass = cls._dtype_to_subclass(arr.dtype) | |||
|
|
|||
| # _ensure_array _may_ be unnecessary once NumericIndex etc are gone | |||
| arr = klass._ensure_array(arr, arr.dtype, copy=False) | |||
There was a problem hiding this comment.
I had to remove “NumericIndex” and had to decide what to do with comment and opted to delete it. IMO _ensure_array could be refactored, but so could a lot of other things in pandas and _ensure_array is not that bad anymore compared to other things, so doesn’t merit a comment In the code base IMO.
This was my reasoning, others could of course have made a different judgement call…So it’s not a strong opinion, and I’m fine with it being refactored, but also it not being refactored :-)
There was a problem hiding this comment.
Yeah no worries, just wasn't sure if this was still being True or not
|
thx @topper-123 |
|
This removal is the partially the cause of a regression unpickling files from "pandas<2", #53300 |
|
This remove action caused a bug |
This finishes the removal of
NumericIndexfrom the code base. Note that this PR builds atop of #51132, so you might want to merge that first.Everything should be finished now after this PR, e.g. the doc updates have been merged (#51111) and the doc string has been updated (#51089). I'll have a final look once this is merged, but this should pretty much finish this issue.