BUG: fix Index's __iadd__ methods #5053
Conversation
|
@jreback are you sure this doesn't break back-compat? Seems like it would. |
|
@jtratner I think its ok; the problem is iadd was using the numpy iadd which effectively does the same thing, but doesn't do the asignment correctly (because its a numpy method and doesn't go thru the python assignment machinery) |
|
I just don't think you should block the non inplace methods without |
|
they don't work properly is the issue as they update the Index inplace which is semantically wrong (because its not calling the right method). |
|
I think it's fine to block inplace, but I thought you were blocking things |
| @@ -93,6 +93,9 @@ class Index(FrozenNDArray): | |||
|
|
|||
| _engine_type = _index.ObjectEngine | |||
|
|
|||
| __mul__ = __trudev__ = __floordiv__ = __pow__ = __imul__ = __itruediv__ = \ | |||
There was a problem hiding this comment.
ok..let's just limit these to the i methods you have here: imul,itruediv,ifloordiv,ipow,
There was a problem hiding this comment.
Not sure that _disabled is what you actually want, because it's not that it's immutable, it's that it returns the right type. /bikeshedding
|
@d10genes ok...this got too complicated! sorry.....let's just do the iadd; rest can do in a separate PR |
|
iadd...and isub? or just iadd |
|
Just |
ghost
assigned
|
I went ahead and updated the title of this PR to make it clearer what's going on. We definitely appreciate you being game on this - please just remove the If you want, you could open a new pull request with your proposed |
fixes pandas-dev#4996 alias Index's __isub__ to __sub__ document fixes to pandas-dev#4996 index isub test roll back everything non-iadd related
|
thanks you! |
closes #4996
Aliases iadd = add.
Not doing anything else for now until time to consider.